Knowledge Base

Bet Finder saved lists don't email out at the moment — they are there to be run when you open them. If a daily email of runners matching your criteria is what you are after, the System Builder on RaceMetrics does exactly that: build the system, save it, and its qualifiers arrive by email each morning at around 8am, with follow-up notices if a non-runner changes the picture.

You can keep several systems active at once, and the email is deduplicated across them — a runner flagged by three of your systems appears once, with all three named.

Yes — that is exactly what My Ratings does (Proform Premium). You define your method once: what a beaten length costs in pounds, varying by race type, going and distance. RaceMetrics then rates every runner in every race with it — your recent history when you set the scale up, and every new result each morning after that. Nothing to type, ever.

How a figure is built

Each run is anchored on the class par (the mark a typical winner of that class, trip and going carries), credited with the weight the horse actually carried relative to the winner, then debited for its beaten distance on your own pounds-per-length curve. Because every race is anchored the same way, the figures compare across courses, classes and seasons — and sit on the same scale as the official BHA mark, the same idea as the Racing Post's RPR but built to your method.

Weight-for-age is applied by default, using the official unified European flat scale and the Anglo-Irish jumps scale, so a three-year-old running well against older horses is rated the way the handicapper would rate it.

Choosing a scale

Six presets in plain English: four beaten-lengths scales (Balanced, Sharp, Gentle, Soft-ground biased) and two speed-figure scales — Weight-adjusted speed figures (the classic timing-book rule, one point per 2.4 lbs either side of nine stone) and Speed figures on the OR scale, which re-expresses those figures in official-ratings points so they read like handicap marks.

Or choose Start from scratch and design your own: pick which race types and going each block covers, set your own pounds-per-length pins, weight-for-age on or off — with a live preview that scores a real race before anything is computed.

Where the figures appear

As a sortable My column on racecards (Ratings view and At-a-Glance), Form Expert, the A-Z Runners page, horse profiles, and as columns in the My Card builder. You can run up to three scales at once and compare them side by side on the same card.

Then prove it

The Scorecard shows how your pre-race top-rated has actually performed — strike rate, level-stakes return at Betfair SP and A/E against the market. Review winners replays any settled day: who won, and where your scale had it ranked before the off. And the System Builder carries a My Ratings filter group, so you can back-test your own method against real prices and run "my top-rated today" as a daily qualifiers email.

In the System Builder, note the difference between the pre-race fields (what your scale said before the off — the ones to back-test) and the earned fields (what the horse rated in the race — for studying your scale, not for reading as an edge).

Yes — that is exactly what My Ratings does (Proform Premium).

You define your method once: what a beaten length costs in pounds, varying by race type, going and distance. RaceMetrics then rates every runner in every race with it automatically — your whole recent history when you set the scale up, and every new result each morning after that. Nothing to type, ever.

How a figure is built

Each run is anchored on the class par — the mark a typical winner of that class, trip and going carries — then credited with the weight the horse actually carried relative to the winner, and debited for its beaten distance on your own pounds-per-length curve.

Because every race is anchored the same way, your figures are comparable across courses, classes and seasons, and they sit on the same scale as the official BHA mark — the same idea as the Racing Post's RPR, built to your method rather than someone else's.

Weight-for-age is handled for you

The official allowances are applied by default — the unified European flat scale and the Anglo-Irish jumps scale — so a three-year-old running well against older horses rates the way the handicapper would rate it. Same-age races are unaffected.

You can run up to three scales at once and compare them side by side on the same racecard.

Yes. My Ratings includes a speed-figures mode with two options.

• Weight-adjusted speed figures — the classic timing-book rule applied to the Proform speed figure: one point per 2.4 lbs carried either side of nine stone. The figures you already know, now crediting horses that ran their time under a burden.
• Speed figures on the OR scale — the same figures re-expressed in official-ratings points, so they read like handicap marks rather than clock numbers.

How the OR scale works

Each race is anchored on its class par (the same anchor the beaten-lengths scales use), and each horse's figure margin to the winner is converted into pounds — with the conversion calibrated separately for turf, all-weather, chase and hurdle from every officially-rated run since 2018.

Checked against 2026 racing, which sits outside the period it was calibrated on, the figures land within about a pound of the average official mark, with the same spread of values.

Individual runs will still differ from a horse's mark — and they should. One run is a single piece of evidence; a handicapper's mark is a considered view across several.

Yes — the System Builder preferences include extra fields for low in-running prices and the in-running percentage move.

On a settled racecard the My column shows both figures, written as 78 → 91:

• the muted first number is what your scale had the horse rated going into the race — its best from the last three rated runs, exactly the figure the card showed before the off;
• the second is the rating it earned in that race, worked out from the result.

On an upcoming race there is only the pre-race figure — the horse has not run yet, so there is nothing to earn.

Why the difference matters

Looking back through old cards, “My Rating is higher than the official mark” can look remarkable — a big strike rate and a big profit. But that is the earned figure, and a horse that ran well is above its mark by definition. The filter is describing what already happened, not something you could have acted on.

Test the same idea honestly with the System Builder field My Rating minus OR (pre-race): your figure before the off minus the horse's official BHA mark. A filter on that is a judgement you could actually have made, it works in the daily qualifiers, and the A/E will tell you whether the angle is real. The earned twin is labelled My Rating minus OR (earned) so the two are never confused.

RMP is a fair price for every runner in every UK and Irish race, struck by the RaceMetrics model each morning once the previous day's results are in and before the market has moved.

It is a 100% book: the prices across a race add up to exactly 100%, with no bookmaker margin in them, so 4/1 means the model gives the horse a one-in-five chance. Every runner also carries an RMP View — a one-line note written by the model at the same moment.

The model reads data you can see elsewhere on RaceMetrics: the RaceMetrics ratings for the horse and its connections, speed figures, the price it went off at last time, pace, Proform's compiled stats, the official rating, weight and days since a run, the horse's record at today's course, distance, going and class, trainer and jockey form, draw bias and headgear. How those are weighed together is the model itself and is not published.

RMP appears on the racecard Ratings view, At-a-Glance and Form Expert, in My Card (with RMP View as a full column), on A-Z Runners, and in the System Builder as columns on a system's qualifier and run listings — gold for the model's top-rated in each race. It is part of RM Pro and Proform Premium.

Not at the close — and we say so. RMP ran as a hidden forward test for weeks before release, scored every night against the closing Betfair market on races with no non-runners. Blended with the evening show it predicts where the market finishes about as well as the bookmakers' own evening prices do, and every attempt to turn “RMP bigger than the bookies” into a profit at the close has come back level. The market is usually right.

Where RMP earns its place is in disagreement: a horse the model rates far shorter than the early market is a horse worth a harder look, and the System Builder lets you test exactly those conditions against real results.

One thing to remember when comparing: a bookmaker's book carries a margin of 15–25% and RMP carries none, so the bookmaker's price will read shorter than RMP on most runners for that reason alone. RMP is an opinion to test, not a tip.

Because the withdrawn horse's chance has to go somewhere. RMP is struck once, in the morning, as a 100% book over the declared field. When a runner is withdrawn afterwards, the remaining prices are too big — the book no longer adds up to 100%.

RMP is therefore re-based over the runners actually left, so it sums to 100% again, and the rank adjusts with it. The figure you see is the one you could act on; hover it to see what was struck. Around four in ten races lose a runner after the morning strike, so this is routine rather than rare.

A horse that joins the field after the strike — a reserve called in — shows as not priced: the model never re-prices a race once struck.

RMP is a model price, struck once per runner before racing and then left alone. It is worked out from the racing record we already hold on every runner in the race, and the figure you see is the model's own opinion of the horse's chance, expressed as a price.

What goes into it

The model draws on the kinds of information you would expect a form student to weigh up, all of it data we hold on the card:

• Ratings and speed figures
• The official rating, the weight carried and days since the last run
• Course, distance, going and class records
• Pace and draw
• Last-time-out market position
• Recent trainer and jockey form
• Headgear and other declared changes

We do not publish the method or how the pieces are weighted against each other, in the same way we do not publish the workings behind the ratings themselves.

What we will tell you about how well it does

RMP was run as a hidden forward test for several weeks before it was shown to anyone, and it is measured against the Betfair close every day. The honest summary is that it predicts the closing market about as well as the evening show does, and every value-against-the-market test we have run has come back at roughly par.

So treat it as a considered opinion to test against your own, not as a tip and not as a licence to bet anything it prices bigger than the bookmakers. If you are comparing it with a bookmaker's book, remember to remove the bookmaker's margin first: on raw prices almost any bookmaker will look shorter than RMP simply because the book adds up to more than 100 per cent.

One thing that does change

An RMP price can move after it is struck, but only for one reason: if a runner is withdrawn, the remaining prices are rescaled so the field still adds up correctly. The underlying opinion is never restruck.

No — and the gap is measurable. RaceMetrics derives an actual going for every race after the meeting from the day's race times (the Going Adjustment), and since the start of 2023 it disagrees with the official pre-race description in roughly half of British turf races. In about one race in twelve the difference is two full grades or more — described Good, rode Soft — and the error runs in one direction: the ground rides firmer than described half again as often as softer. The official form book keeps the pre-meeting description forever, so form study built on it inherits those errors. Premium users can switch Form Expert to analyse a horse's record against actual going instead.

Yes. My Ratings (Proform Premium) includes imported rating sets: open My Ratings from any racecard — or Racecards → My Ratings — choose Add a scale, then Import your own ratings, name the set and upload a CSV. The file needs a header row naming date, time, course, horse and rating — a combined date/time column is fine, extra columns are ignored, and ratings can be whole numbers or decimals. Every row is checked first: a match report lists anything unmatched, with the line number and reason, before a single row is written — and re-importing a race simply replaces its figures, so correcting a file is always safe.

Ranks are worked out per race from your figures and adjust automatically for non-runners. Your ratings then appear on racecards, My Card, the A-Z and horse profiles, and the System Builder can filter, back-test and break them down — “My Rating Rank = 1” in Qualifiers means your top-rated on today’s card. History imports back to 2018 for honest back-testing, and imported sets share the three-active limit with computed scales. Full details in the My Ratings guide.

Ratings are updated daily after racing concludes. Results from each day's racing are processed overnight, so you'll see updated ratings reflecting yesterday's results each morning.

RaceMetrics ratings are our proprietary performance measure. They track form over time – 1500 is average, 1550+ indicates strong performers, and 1600+ represents elite level. Ratings are updated daily each morning based on the previous day's results.

We track seven connection types: Horses, Trainers, Jockeys, Owners, and bloodlines (Sires, Dams, Damsires). Each profile includes full form analysis, ratings history, P/L tracking, and performance breakdowns.

RaceMetrics applies Elo ratings to UK and Irish horse racing. The system was built by Simon Walton, founder of Proform Racing Limited (a British racing software company established in Halifax, West Yorkshire in 1995), and rates every race that has run in Britain and Ireland across seven connection types — horses, trainers, jockeys, owners, sires, dams and damsires. Ratings are recomputed daily on Proform's full 25+ year dataset and surfaced at racemetrics.co.uk.

Horse racing Elo ratings work by extending the original two-player chess formula to multi-runner races. For each horse in each race, RaceMetrics computes the Elo expectation against the entire field — what percentage of rivals the horse would be expected to beat given its rating versus the field's ratings — then updates the rating once based on how the horse's actual finishing position compared to that expectation. The same machinery runs in parallel across seven independent rating pools: horse, trainer, jockey, owner, sire, dam and damsire.

To use Elo ratings for horse racing, look at two figures together: the horse's individual H rating (its own ability) and the Combined Score (a weighted average of the trainer, jockey, owner, sire, dam and damsire ratings). A strong H signals individual horse ability; a strong Combined Score signals an above-average support package around the horse. Historically the horse with the highest Combined Score in a race wins 22.97% of the time — 2.15 times the random-chance rate — across 25+ years of UK and Irish racing data.

PRB stands for Percentage of Rivals Beaten — a finishing-position metric used in RaceMetrics' Elo adaptation. PRB normalises finish position by field size. For example, a horse finishing 3rd in a 12-runner race has a PRB of (12-3)/(12-1) × 100 = 81.8%, having beaten 9 of 11 rivals. PRB is preferred over raw finishing position because it accounts for field-size differences and produces a continuous 0-100 metric suitable for Bayesian rating updates.

The Combined Score is RaceMetrics' weighted average of six connection-type Elo ratings: Owner 20%, Trainer 20%, Jockey 20%, Dam 18%, Sire 12%, Damsire 10%. The horse's own H rating is reported separately and is deliberately not folded into the Combined Score. The two figures answer different questions: H is about the horse's own ability; Combined Score is about the strength of the support package around it.

The RaceMetrics Rankings are live performance rankings for British and Irish racing, covering horses, trainers, jockeys, sires, dams, damsires and owners. Each participant carries a RaceMetrics Rating that rises or falls after every race depending on how they perform against expectations. The top 25 in every category are free to view at racemetrics.co.uk/rankings, updated every morning after racing.

The top of the rankings changes with the results, so the honest answer is always the live table: racemetrics.co.uk/rankings shows the current top 25 horses, updated every morning after racing. Horses need at least three runs to qualify, so the leaders reflect current form rather than reputation.

The rankings page and the RM7 Stats Centre both show the biggest movers — the largest rating rises and falls over the past week — across all seven participant types. A fast-rising trainer or jockey is often the earliest sign of a yard hitting form.

Yes. The top 25 in every category are free for everyone at racemetrics.co.uk/rankings — no account needed. A free RaceMetrics account adds full racecards with every runner's rating, and entry to RaceMetrics 7, the free weekly game built on ranking changes.

Every RaceMetrics racecard shows a Form Temperature pill next to each trainer and jockey — a 0–100 score for their last 14 days, measured against their own five-season strike rate.

That last part is the point. Being hot is relative: two winners from six runners is a flying fortnight for a small yard and a quiet one for a champion stable. Scoring every trainer and jockey against their own record means big and small operations can be compared fairly.

Reading the pill

• Red — 95 or higher: a fortnight that good comes up by luck fewer than 1 time in 20
• Amber — 75 to 94: a better fortnight than usual
• Pale blue — 6 to 25: a quieter fortnight than usual
• Blue — 5 or lower: the same rarity on the quiet side
• No pill — 25 to 75: normal variance, which is most fortnights, so the card stays clean

Hover a pill — or press and hold it on a phone — for the fortnight's record, the five-season benchmark it is being judged against, and how often a run like it happens by chance.

Where it appears

On the Ratings view and At-a-Glance. On past racecards the score is worked out as it stood on the day of that race, so you can see the form a yard actually carried into a result. It is free on every racecard for all users.

For the underlying numbers as sortable columns — runners, wins, strike rate and A/E over the 14 and 28 days before the race — see the My Card Form Stats columns.

On our own testing, no — not on its own, and we think it is worth being straight about that.

Before building Form Temperature we back-tested the idea over two separate two-year periods of UK racing.

• The streak is real. Runners from statistically hot yards (a score of 95 or higher) won about 15% of the time against a baseline expectation of roughly 11%. Recent form does persist into the next run.
• The market has already read it. Actual-versus-expected against Betfair SP came out at roughly 1.0 in every temperature band, and backing hot-yard runners at level stakes lost more than backing runners at random — hot yards get over-bet.
• The tempting result was noise. Positive returns from significantly cold yards showed up in one two-year window and did not repeat in the other.

So the honest use of trainer and jockey form is as a lens for reading a race, not a selection method. It tells you which stables are running above or below their own standard, so you can ask why — the ground coming right, a new stable jockey, a change of target races — and then test that reason properly in the System Builder.

Patterns are saved from profile pages. When viewing statistics (e.g., a trainer's record at a specific course), look for the Save Pattern button. This creates a pattern with those conditions that will alert you when matching runners appear on racecards.

Yes. RaceMetrics records the In-Running High and In-Running Low for every run — the highest and lowest prices matched on Betfair while the race was being run — together with the percentage each sits above or below Betfair SP. These are the same fields shown in the Proform Race Guides and desktop software.

On horse form lines

RaceMetrics Pro and Proform Premium subscribers see BSP, IRLo (In-Running Low) and IRLo% (its move against BSP) on every form line by default — on racecard expansions, Form Expert drill-downs, profile drill-downs and My Card print. IRHi and IRHi% can be added with the Columns button, which also lets you show, hide and reorder every form-line column.

Reading the percentage

Negative means the horse traded that much shorter than its BSP in-play — a horse at -90% or more usually came very close to winning. Positive on the low means its price never dipped below BSP: it was never travelling like a winner.

In the System Builder

In-Running Low/High and both percentage moves are available as filters (for example, "traded 60%+ below BSP last time"), as breakdowns, and as columns on the Runs listings and CSV exports via the shared Drill columns chooser.

In-play price coverage runs from 2018. Form-line fields need RaceMetrics Pro or Proform Premium; the System Builder is part of Proform Premium.

Yes. My Card's Form Stats columns show point-in-time trainer form directly on the racecard — the trainer's runners, wins, win strike rate and A/E over both the 14 days and the 28 days before the race. Point-in-time means the numbers are computed strictly from data before the race date, so on a past racecard you see the form the yard had going into that race, not today's form.

Adding the columns

Open any racecard, switch to the My Card tab, click Customise and pick from the Form Stats group (FS chip) — T Runs 14d, T SR% 14d, T A/E 14d and their 28-day equivalents, plus the same two windows for jockeys, runners since the trainer's last winner, and the trainer–jockey pairing record.

Reading the numbers

The 14-day window is the classic "stable in form right now" yardstick; 28 days smooths out smaller yards. A/E above 1.00 means the yard is winning more often than the market expected — a high strike rate alone can already be priced in.

Form Stats columns need Proform Premium — they are the System Builder engine's own precomputed numbers, and the same 14-day and 28-day windows are available as System Builder filters for back-testing. The free Trainers in Form page shows today's in-form trainers over the last 14 days without a login.

The A/E (actual versus expected) figures in the System Builder are calculated using BSPs — Betfair Starting Prices. Expected wins are derived from each runner's actual BSP, so A/E measures value against the exchange market rather than the industry SP.

This applies to the System Builder on RaceMetrics, which is included with a Proform Premium subscription and goes further than the desktop software version — back-testing since 2018 with P/L at BSP, ISP and bookmaker prices side by side.

The A/E figures in the System Builder are calculated using BSPs (Betfair Starting Prices). The RaceMetrics web version of the System Builder can do more than the desktop software version and is included free with the subscription.

Build the selection exactly as you would for backing, then change the staking mode. The two are separate steps.

Step one: describe the runners

The conditions in the builder describe which runners you want to find, not what you intend to do with them. So a lay system for favourites in big-field handicaps is built by setting the conditions:

• Favourite
• Handicap only
• Field size of twelve or more

Do not phrase the condition itself as an instruction to lay. If you write the whole thing as a lay instruction, the parts that describe the runner get lost and the system comes back with conditions missing.

Step two: switch the staking to Lay

Once the system is matching the runners you want, open the Staking tab and choose Lay.

The results are recalculated with the outcome inverted, and stakes are set by liability rather than by stake, so the figures reflect what a lay bettor actually risks. Each-Way is available in the same place.

This means you can build one selection and look at it both ways — back and lay — without rebuilding anything.

All four sit in the System Builder together, so they combine as one system.

The finish: use LTO Finishing Speed %, or LTO Finishing Speed vs Par — par-relative is usually the sharper of the two, because it allows for how the race was run. LTO Finishing Speed % Rank asks instead for the fastest finisher in that race, which travels better across race types than an absolute figure.

Trouble in running: Hard-Luck Story LTO flags the horse that met trouble, and LTO In-Running Positions Gained (or LTO Change in Lengths) finds the one that was making ground through the field.

The market: BSP Rank for betting position — top 2, top 4 — and BSP (odds to 1) for a price band.

One thing worth knowing: the LTO sectional filters describe the horse's last run, so they work on today's card in qualifiers. Sectional filters for the race being analysed are post-race, so they are available in backtests but cannot qualify a runner in advance. Sectional filters require the sectionals subscription.

Yes. With the Sectionals Add-On the System Builder includes sectional position filters built from TPD per-section tracking: Early Position and Early Lengths Behind Leader (measured roughly 1-1.5 furlongs after the start), Position ~2f In and Lengths Behind ~2f In (the second section), and Position 2f Out and Lengths Behind 2f Out (where it was turning in). Each comes as a matched pair: the figure in the current race for researching results, and a Last Time Out version.

The one-rule front-runner finder

"Led early last time out" is a single rule: LTO Early Position is 1 (or LTO Early Lengths Behind Leader = 0). Because lengths behind the leader reads the same in a 5-runner race as a 16-runner one, "within a length of the lead early" is often a sharper sieve than raw position. The same rules project today's likely race pace: check how many of today's runners led early last time.

Works for today's qualifiers

A previous run's sectionals are settled history, so all the LTO versions work for daily qualifiers and qualifier emails as well as back-tests. Matching breakdowns — LTO Early Position bands, LTO Early Lengths Behind bands and LTO Position 2f Out — show strike rate, A/E and ROI by where horses raced last time before you commit to a rule.

Position data covers ARC all-weather courses across the whole back-test era and RMG courses from mid-2023, with no Irish coverage; runs without sectional data are excluded from the cohort rather than counted as zeros. Requires Proform Premium plus the TPD Sectionals Add-On.

Yes to both. Official ratings are available as filters, and handicap debuts are found with the career handicap-runs rule.

Official Rating

The Official Rating filter takes a minimum, a maximum or a range. It only includes horses that had a published mark at the time of the race, so unrated horses drop out of the cohort.

Finding handicap debutants

Set All Hcap Runs (career) to 0. That counts every handicap run the horse had before that day, so 0 means a genuine first-ever handicap start — not simply a horse that was not in a handicap last time out.

Use Flat Hcap Runs instead if you want to ignore any previous jumps handicap runs.

A worked example

Handicap debutants on the Flat in Ireland rated 55 or lower is five rules:

• Course Country: Ireland
• Race Type: Turf and All-Weather — the All-Weather brings in Dundalk, so remove it for turf only
• Handicap: Yes
• All Hcap Runs (career): 0
• Official Rating: 55 or lower

Over the full history from 2018 that returns roughly 1,500 runs, around 130 to 160 a year recently. The Runners tab lists every qualifier with dates, courses, prices and finishing positions and exports to CSV. The Qualifiers tab flags any of today's or tomorrow's declared runners that match, so the same rules work forward as well as back.

You can also type the description into the Ask box in plain English and it will build the rules for you.

That one wants Goal Seek rather than the Ask box — the two work in opposite directions.

The Ask box translates a system you already have in mind into rules: "3yo handicappers on soft ground, top 3 in the betting" becomes those filters, ready to run. It needs the conditions, not the outcome.

Goal Seek works the other way round. You set the targets — a minimum strike rate, a minimum number of qualifiers, a positive return — and it searches rule combinations for ones that reach them, then holdout-checks what it finds against a period it did not search, so a rule that only fitted the past does not survive the trip.

So: start from a broad system, open Goal Seek, and set your three targets there.

Yes, both.

Back-testing

Once you have a My Ratings scale, the System Builder gains a My Ratings filter group — your rating, its rank in the race, your rating minus the official mark, your rating last time out, and the change since. Test your own method against real Betfair prices with the same A/E, ROI, drawdown and holdout checks every other system gets.

Daily selections

My Rating Rank works in the daily qualifiers: today's declared field is ranked live on each horse's provisional figure — exactly the number the racecard shows — so My Rating Rank = 1 means your top-rated in the race, before the off. Save it as an active system and those runners arrive in the qualifiers hub and the qualifiers email each morning.

Two one-click starters sit in the My Ratings panel: Today's top-rated, and a Divergence Finder — the runners you rate number one that the RaceMetrics ratings don't.

Cross-checking two methods

Each rule can read a different one of your scales, so one system can say my speed scale rates it top and my lengths scale has it in the top three.

Did it actually work?

The Scorecard tracks how your pre-race top-rated has performed over 30, 90 or 365 days — strike rate, level-stakes return at Betfair SP, and A/E against the market. Review winners replays any settled day race by race: who won, where your scale had the winner ranked before the off, and how your own number one got on.

Pre-race or earned — pick the right field

The group contains both, deliberately. My Rating (pre-race) and My Rating Rank (pre-race) are what your scale said going into the race — each horse's best from its last three rated runs, the figure the racecard showed before the off. Back-test these: a filter on them is a selection you could have made, and they use the same basis as the daily qualifiers, so a back-test and the morning email agree.

My Rating and My Rating Rank (no label) are what the horse earned in the race, derived from the result. They are excellent for studying how your scale behaves — but they are hindsight, so their P/L is not an edge. A filter on earned rank 1 can show a strike rate above 60% purely because it is describing what already happened.

What counts as a selection

A pre-race back-test counts every runner your scale rated top, including the ones that did not complete. If your top-rated fell or was pulled up it still appears, as the losing bet it was. Leaving those out would quietly flatter every figure, because a horse that fails to finish never wins. Only genuine non-runners are removed — and those do change the ranking, exactly as they do on the racecard.

Rating your horse against the handicapper

My Rating minus OR (pre-race) is your figure before the off minus the horse's official BHA mark — positive means you rate it ahead of the handicapper. This is the honest “well handicapped” test, and it works in the daily qualifiers. Its earned twin, My Rating minus OR (earned), is derived from the result: a horse that ran well is above its mark by definition, so that one is for studying your scale, never for reading as an edge.

Yes, and there is nothing to switch on — it runs every time you calculate the staking figures.

The simulation reshuffles the running order of your matched bets 200 times, re-runs all ten staking plans against each ordering, and reports the outcome in the Bust-risk column. It runs on any system up to 25,000 priced bets, and the figure carries through to the CSV export.

Bust-risk is the column to watch on the aggressive plans. A plan can show a healthy final bank on the actual running order and still empty the bank in a third of the reshuffles. Same bets, same edge — only the order of arrival differs. A run of losers early enough in the sequence takes the bank down before the winners arrive, and the actual historical order is only one of the orders you might have met.

That makes Bust-risk the honest companion to the final-bank figure: the bank tells you what happened once, the Bust-risk tells you how often that outcome survives a reordering.

Around eight requests through the week combining sectional form with conventional rules, several leaving one condition unmapped. The recurring idea is a horse whose sectional figures say more than its finishing position, expressed several different ways. Sectional-based systems already have a knowledge item, but not one that starts from this framing, which is how customers actually phrase it.

Yes, in two places — one looking backwards and one looking forwards.

For back-tested bets: the all-bets table

In Portfolio, pick your systems, run Compare and Blend, then click Show the bets. You get every bet any of the selected systems flagged, newest first, with duplicates removed — a runner that two of your systems both like appears once, tagged with a coloured marker for each system that found it. Download CSV gives you the lot as a spreadsheet, including a column naming which systems picked each bet, and the Columns button lets you add fields from the runs drill-down before you export.

One thing to know: the table pulls each system's most recent bets across the shared date window. If a busy system fills that window you will see a note — narrowing the date range brings in the rest.

For today and tomorrow: the qualifiers hub

The same job, forward-looking. Every active system's selections on one page, deduplicated, tagged by the system that found each one, with its own CSV.

What about OR-ing rules together?

More of this exists than people expect. Any multi-select rule — courses, going, trainers, headgear and so on — is already an OR within that rule.

For OR across different rules, saving each variant as its own system and blending them is the cleaner route, and not just as a workaround. You get exactly "system A's rules or system B's rules", deduplicated, and you keep honest per-system figures, so Trim, Frontier and the holdout checks still mean something for each system on its own.

Because strike rate and value are different questions. Impact Value asks whether a selection wins more often than chance; A/E asks whether it wins more often than the market expected at the price. Good horses backed at bad prices post a strong IV with an A/E at or below 1.0 — the market already knew they were good and priced the edge away. Only A/E above 1.0, with a meaningful sample behind it, marks a genuinely backable angle. Read the two together on every System Builder back-test: IV for “is it real”, A/E for “is it mispriced”.

Betfair SP is formed by money matched in the seconds before the off — it does not exist until the race starts. A rule like “BSP between 5 and 10” therefore filters on information no bettor could have acted on, and the back-test quietly reads the future. Results built that way flatter the angle and cannot be reproduced with real bets. Filter instead on the evening, breakfast or morning prices — all recorded before the off and carried with live per-race ranks for qualifiers — and keep BSP as the settlement price your profit is measured at, not a selection rule.

Probably not on that evidence alone. Test a hundred combinations and roughly five will clear a conventional significance bar by pure chance — the multiple-comparisons problem, and the single biggest reason data-mined angles fail once followed with real money. Chi only answers “is this gap unlikely by chance”; it cannot know you went looking until something passed. That is exactly what the Robustness tab exists for: a time-based holdout and a random-split test on data the angle was not built on. Run it after Goal Seek or any long filter hunt, every time — an angle that holds on both splits is worth trusting, one that only exists on the data that found it is a description of noise.

Yes. In the System Builder's Staking tab, choose Double Or Bust (DOB) or Treble Or Bust (TROB) as the bet type. Every bet your system matched is settled as the in-running trade: back at Betfair SP, lay in-running at half (DOB) or a third (TROB) of the decimal odds — a hit locks in +1 or +2 points whatever the result, a miss loses the stake.

The tab shows your hit rate against the break-even line — 50.51% for DOB and 33.78% for TROB at 2% commission, recomputed from your own commission setting — using Betfair's recorded in-running lows for every runner back to 2018. Runners priced below decimal 2.02 (DOB) or 3.03 (TROB), where Betfair's 1.01 minimum price makes the lay leg impossible, are excluded from the count rather than scored as losers, and the excluded figure is always shown.

DOB and TROB are back-test only: an in-running price exists only after the race is run, so they never appear in qualifiers — they show how your selections would have settled as a trade, assuming the lay was matched at the target price. All ten staking plans run except Kelly (there is no quoted market for “will it trade at the target”, so no honest probability to feed it). Requires Proform Premium.

Because the lay never matched. Double Or Bust locks its profit by laying in-running at half the decimal odds — if the price never trades that low, there is nothing to match against, so the strategy as defined books the bet as a full loss even when the horse goes on to win.

It is rare — about one eligible winner in nine hundred never trades at half its BSP — and the Staking tab reports the count whenever it happens, so nothing is hidden.

In reality an unmatched lay would leave you holding a winning back bet, but commercial DOB trackers settle strictly, and the System Builder follows that convention so its hit rates are comparable with how the strategy is defined everywhere else.

Yes. The System Builder breaks any back-test down by Field Size (exact) — one line per exact runner count (5, 6, 7 and so on), not just broad bands — with runs, wins, win and place strike rate, P/L and A/E on every line. Place strike rate follows real place terms, which turn at exact field sizes: 2 places at 5–7 runners, 3 at 8–15, 4 in handicaps of 16 or more — so “exactly 7 vs exactly 8” is a question worth asking.

Pair it with the Race Age Band breakdown (2yo only, 3yo only, 3yo+, 4yo+) or put the two on the Grid — for example a sire’s place % by exact field size, in one view. Filters cover sire, runner sex, course country, Turf vs All-Weather, handicaps and race age limits, so a question like “a sire’s male runners in GB turf handicaps, place % from 5 to 15 runners” is one saved system. Back-tests run over every UK & Irish result since 2018 at real Betfair prices. The System Builder is included with RaceMetrics Pro and Proform Premium.

The System Builder is a betting-system back-tester. You build a system from a set of rules (filters), then test how those runners would have performed across UK and Irish racing back to 2018 - strike rate, A/E, ROI and profit/loss at Betfair SP - before refining it and tracking today's qualifiers.

The System Builder is included with RaceMetrics Pro and with the Proform Premium subscription. It runs in the browser on any device - nothing to install.

Back-tests run on resulted UK and Irish racing from 1 January 2018 onwards. Some specialist fields (sectionals, bookmaker prices, RaceMetrics ratings) start later and will raise the start date automatically with a warning.

P/L is shown to level stakes at Betfair Starting Price (with your chosen commission), alongside Industry SP and bookmaker prices. A/E (actual versus expected wins from the market) measures value independently of price.

You can save up to 50 systems (raised from 20 in July 2026). Up to 50 active saved systems feed the Today's Qualifiers hub and, if you opt in, the morning qualifier email - use the Active toggle on each saved card to choose which ones count. An imported system doesn't use a slot until you choose to save it.

Yes. The Export button (under Save) downloads the current system as a small file named after the system - its rules, date range, commission and follow-up setting. Send that file to another subscriber and they load it with the Import button. It arrives unsaved and is checked against the current filter catalogue, so they can inspect and back-test it before deciding to save. The System Builder needs Proform Premium, so shared files are only loadable by subscribers.

Yes, three ways. The Data split setting offers random Half A / Half B (build on one half, confirm on the other) and a machine-learning-style Train 80% / Test 20% holdout - both split by race with a deterministic hash, so a given race always lands in the same bucket. The Robustness tab adds a time-based check: it trains on older data and tests on the last 12 months. A system that passes both a random split and the time holdout is far less likely to be curve-fitted.

Yes. Use the Top-3 Finishes in Last 4 Runs filter and set it to 4 - it counts how many of the last four starts finished in the first three, so 4 means every one of them did. Set it to 3 or more instead for horses placed in at least three of the last four, which usually gives a more workable pool. There is a matching Top-3 Finishes in Last 3 Runs for a tighter, more recent version, and Wins in Last 4 Runs if you would rather count victories than placings. Two things to know: a horse needs at least four previous runs to qualify, so lightly-raced types are excluded rather than judged on a shorter record, and a run where the horse did not complete (fell, pulled up, unseated) never counts as a top-three finish.

Yes, by combining four filters. Set Class Change vs LTO to 0 and Class Change vs 2nd-LTO to 0, then set Distance Change vs LTO (furlongs) and Distance Change vs 2nd-LTO (furlongs) to a range of 0 to 0. Each of those measures today's race against that previous run, so setting all four to zero means today, the last run and the run before are all at the same trip and the same grade. Drop the two class filters if you only care about the trip, or the two distance filters if grade is all that matters. Note the distance filters work in exact yards, so a 7f race and a 7f 20y race count as different trips.

No. In follow-up mode your filters describe a trigger run - a race the horse has already had - and the mode backs that horse's next starts after it. The qualifying race itself is never bet.

That decides how you set a system up. To follow horses beaten at a particular meeting, describe that race as today's run in the filters and tick follow-up. If you describe it as the last-time-out run instead, the trigger becomes the horse's next start, and you would be following the runs after that - skipping the first run back.

You can follow 1 to 6 subsequent starts (3 by default). Only races that have actually been run count as a leg, so a declared but unrun race does not use one up. It stops at the first win unless you untick that option, and the winning run is included. Each follow-up run is counted once even if the horse triggers the system more than once, and the date range applies to the trigger run - the follow-up runs can fall after it. Follow-up gives the summary figures only; breakdowns, staking and qualifiers are unavailable while it is on.

They answer different questions, and confusing them is the most common mistake with these figures.

A/E compares actual wins against what the betting market expected, measured from Betfair Starting Price. It tells you whether something has been good value. 1.0 means it won exactly as often as its odds implied.

Impact Value (IV) compares actual wins against a fair share by chance. In a ten-runner race every runner's fair share is one win in ten, so an IV of 1.5 means that group won half as often again as chance alone would give it. IV ignores the price completely.

Something can have a strong Impact Value and still be poor value to back, because the market has already priced it in. That is exactly why the System Builder shows both side by side.

Yes. The Draw IV, Pace IV, Tactical Score and Draw × Pace IV columns shown on the racecard are also available as System Builder filters and breakdowns, so an idea you spot on a card can be tested over results going back to 2018 rather than taken on trust.

They are computed point-in-time — strictly from racing before each race — so the same figures work for back-testing history and for listing today's qualifiers. The same values are also available as they stood last time out, which supports angles such as “finished in the first three despite a bad draw”.

One caution: about 7% of flat runners have no running-style history at all and a further 19% only one to three runs. A Run-Style Sample filter is provided so you can set your own minimum before trusting any pace figure.

A first-season sire - also called a freshman sire - is a stallion whose first crop of offspring is racing for the first time. Because none of his stock has ever run, there is no progeny record for punters or the market to price from, and that uncertainty is why the angle is followed.

In System Builder the First-Season Sire filter matches runners from a sire's very first crop to reach the track. Sire's Crop Number gives you finer control - 1 is his first crop, 2 his second, and so on - so you can select the first two crops, or exclude freshmen entirely with 3 or more. A matching Sire's Crop breakdown shows how each crop performs before you commit to a rule.

One thing worth knowing: you cannot identify a first-season sire from his age. When a stallion's first foals reach the track depends on when he retired from racing, not on how old he is - a horse retired after his three-year-old season has his first two-year-olds out when he is seven, a year later and it is eight, and one who raced on until five is a freshman at nine. RaceMetrics therefore counts crops from when each sire's progeny actually started running, so the filter is exact rather than an estimate.

Across every two-year-old run since 2018, the first crop is the only one showing a positive return, with each later crop negative. The sample is not large enough to treat that as a reliable edge, so it is better used as a starting point for your own testing than as a finished system.

Yes - that is what the Grid tab does. A normal breakdown splits your system one way at a time; the Grid splits it two ways at once. You choose what goes down the side and what goes across the top, and every cell is the answer for that pair.

The question it was built for

A customer asked for each trainer's strike rate in August, over the last five seasons, for yards with more than five runners, split by class of race. Before the Grid you could answer half of that at a time - one trainer across all the classes, or one class across all the trainers. Now it is one screen: set the date range, add the rule Month is August, run the system, then open Grid and pick Rows: Trainer, Columns: Class, Show: Strike rate, Min runs: 6.

What you can change

• Rows and Columns - any two of the breakdown dimensions. Trainers against courses, sires against distance, run style against field size.
• Show - which figure fills the cell: strike rate, A/E, ROI, P/L, PRB, Impact Value, Chi, place strike rate, runs or wins. Switching it redraws straight away.
• Min runs - lines with fewer runs than this are left out, and thin cells are greyed rather than shown as a percentage. A 100% strike rate off one run is noise and should look like it.

Reading it

Colour is relative to your own system, not an absolute: blue means that cell does better than the system does overall, amber and red worse. Click a cell to see the runs behind it, or shift-click to add both rules - the trainer and the class - to your system at once. The grid exports to CSV.

The totals are labelled honestly. The line above the grid tells you how many lines qualified in total, so "50 of 853" never reads as the whole picture, and the totals row says "These 50" rather than "All" when you are looking at part of the list.

The Grid is part of the System Builder and needs a Proform Premium subscription.

Yes. The System Builder's Last Time Out filters include the Betfair in-running trading from the horse's previous run: LTO In-Running Low and LTO In-Running High (decimal odds), plus LTO In-Running Low % Below BSP and LTO In-Running High % Above BSP.

The one-rule version

"Traded at least 60% below its Betfair SP in-running last time" is a single rule: LTO In-Running Low % Below BSP is at least 60. Positive numbers mean the horse shortened in running — it travelled like a big danger mid-race. Pair it with LTO Finishing Position 2 or worse to keep only the ones that got beaten — the classic "travelled like a winner, worth another chance" angle.

It works for today's qualifiers

The this-race in-running fields are post-race only, but a previous run's trading is settled history — so these LTO rules work for daily qualifiers and qualifier emails as well as for back-tests.

Deeper form and breakdowns

The same four fields are available 2, 3 and 4 runs back, so you can demand a pattern — for example, traded 60%+ below BSP in each of its last two runs. A matching LTO In-Running Low vs BSP breakdown shows how each trading band has actually performed before you commit to a rule.

In-running price coverage runs from 2018. The System Builder is part of Proform Premium.

Yes. The RaceMetrics Ratings tab includes two Yes/No filters — No Collateral Score and No Future Form Score — that select exactly the runners those scores cannot reach. About 19% of runners carry no Collateral score (no shared-rival form lines to score from) and about 8% no Future Form score (none of their recent races could be scored) — typically lightly-raced types.

From the breakdown to a rule

In the Collateral Score Rank and Future Form Score Rank breakdowns (and the matching value bands), unscored runners appear as the Unknown row — its + rule button adds the No-Score filter for you. Set the filter to No instead and a system takes only runners that do carry a score.

Qualifiers and the back-test window

Whether a runner has a score is known before the race, so both filters work for daily qualifiers — automated systems can deliberately include (or exclude) the unscored runners. RaceMetrics ratings begin in 2023, so adding a No-Score rule starts the back-test there; earlier races carry no ratings at all. There is deliberately no AI Rating version: AI coverage now spans essentially every declared runner on the day, so there is no unknown group left to select.

The System Builder is part of Proform Premium.

Yes. The 2 & 3 Runs Back breakdown tab includes Distance Move vs 2-Back and Distance Move vs 3-Back splits — today's trip minus the trip the horse ran that day, in real furlongs — with buckets from 2f+ shorter today through Same trip (within half a furlong of today's distance) to 2f+ longer today. The Last Time Out tab has the same split against the most recent run.

Position at today's distance in one screen

Open the Grid and put 2-Back Position down the side with Distance Move vs 2-Back across the top: every cell is that pairing's P&L, so "placed 2 back at today's distance" is a single cell, with the shorter- and longer-trip versions alongside for comparison. The same works for 3-back.

From bucket to rule to qualifiers

Each bucket's + rule button adds the matching Distance Change filter with the same definition, and because a previous run's distance is settled history the rules work for daily qualifiers as well as back-tests. About a third of runners with a 2-back run are at that same trip today.

The System Builder is part of Proform Premium.

Goal Seek answers "what single rule could I add to make this system better?". Pick a target — A/E, Betfair SP ROI, win strike rate or profit per run — click Find, and it scans your system's breakdowns for the one rule that would lift that target the most.

Significance-checked, not fluke-chasing

Every suggestion is checked with a statistical significance test (chi-squared), so a lucky-looking line from a handful of runs is not promoted as an improvement. Each suggestion also offers a one-click robustness check against the last 12 months, so you can see whether the improvement is likely to last.

Use it sparingly

Hit Apply to add a suggested rule, then re-run to find the next improvement on the now-tighter system. Every rule you add shrinks your sample — add one or two strong, sensible suggestions rather than ten weak ones.

Goal Seek is part of the System Builder, included with Proform Premium.

Auto-build is Goal Seek on autopilot. Tell it what you want to reach — for example "Betfair SP ROI of at least 5%, keeping at least 500 runs" — and it builds a system toward that spec one rule at a time, running Goal Seek repeatedly and adding the best rule each round until it hits your target or cannot safely go further.

Holdout-guarded at every step

Piling on rules is exactly how over-fitting happens, so Auto-build checks every step against the future: it trains on the older data, tests on the held-out last 12 months, and only keeps a rule that holds up out-of-sample. When the best remaining rules only work in the training data it stops and says so — "stopped to avoid over-fitting" is a feature, not a failure.

Nothing is applied automatically

When it finishes, click Load this system into the builder to review, tweak and run it like any other system. It needs at least 24 months in your date range (so the final 12 can be held back). Treat the result as a strong first draft — always sanity-check that the rules make racing sense.

Auto-build is part of the System Builder, included with Proform Premium.

Yes — the Weight Carried (Provisional) filter on the Horse tab. It is the allotted weight minus any jockey’s claim: a horse set to carry 9-2 (128 lbs) with a 5 lb claimer aboard actually carries 8-11 (123 lbs), and that 123 is the figure this filter works on.

Why it exists

Racecards quote the allotted weight with the claim shown separately, but once a race results, the recorded weight carried is the net figure. The plain Weight (lbs) filter follows the card, so before a race it reads the allotted weight and after a race it reads the net one. Weight Carried (Provisional) puts both on the same basis: back-tests read the recorded weight carried, and today’s qualifiers subtract the claim from the allotted weight — so a rule like “carries 9st or less” selects tomorrow’s runners on exactly the basis the back-test was built on.

Using it

It sits next to Weight (lbs) on the Horse tab and takes a range in pounds. The Weight Carried breakdown’s one-click “+ rule” buttons add this filter, so a band spotted in a breakdown carries straight through to live qualifiers. It is called provisional because a rider change before the off can still move the figure — rare, and settled by the time results are recorded.

Rank as well as value

There is a rank twin, Weight Carried Rank (Provisional): the horse’s rank within the race on the weight actually carried, with 1 = top weight. For today’s runners the claim is deducted before ranking — so a 5 lb claimer allotted the same weight as another horse ranks below it — while on resulted races the recorded weight carried is ranked as-is. It is computed live for every race (nothing stored), and in back-tests it matches the existing Weight Rank exactly, because recorded weights already have the claim taken off; what it changes is today’s and tomorrow’s qualifiers.

Part of the System Builder; needs a Proform Premium subscription.

Trim is the mirror image of Goal Seek. Goal Seek looks for a rule worth adding; Trim takes each rule you already have, re-runs the back-test with that rule removed, and tells you what the rule is really contributing.

The four verdicts

• Drop - the runners this rule excludes actually carry an edge, so removing it improves your target. Cut it.
• Loosen - the rule barely moves your target either way. It is a volume lever you can relax for more bets.
• Keep - genuinely load-bearing. Removing it weakens your target, so leave it alone.
• Redundant - the rule excludes nothing your other rules do not already exclude. It is doing no work.

Why removing rules often helps

Systems accumulate rules, and every rule you add shrinks the sample behind every figure in your results. Cut a rule marked Drop, re-run, and you will often find the leaner system makes more money from more bets - with a bigger sample supporting each statistic. Each suggestion carries the same one-click robustness check against the last 12 months used elsewhere in the builder, so you can see whether the change is likely to hold up.

Trim is part of the System Builder and needs a Proform Premium subscription.

Yes, on both counts.

Lay P/L, win and place markets

Every back-test has always shown the win-market lay P/L (laying each pick at Betfair SP). The results now also show lay P/L in the Betfair place market — laying each pick at the Betfair place price, so the bet wins when the horse misses the places. It appears in the summary’s “Other angles” strip and as Lay Plc P/L and ROI columns in every breakdown, all after commission.

Robustness for lay systems

The Robustness tab now reads both ways. The Holdout Check — train on the older data, test on the held-back last 12 months — carries Lay ROI and Lay Place ROI rows with their own consistent / weaker / flipped verdicts, so a lay angle gets the same over-fitting defence a back angle does. The Streaks & Drawdown card adds Max Drawdown (Lay) and Max Drawdown (Lay Place): the worst peak-to-trough fall of the lay equity curves at level stakes. Remember a lay’s losing run is the back side’s winning run — a lay loses when the horse wins.

Staking too

The Staking tab’s bet modes are now Back, Lay, Lay the Place and Each-Way — so all ten staking plans, the drawdown figures and the Monte-Carlo bust-risk simulation can be run on a place-market lay strategy, liability-staked, with commission applied.

Part of the System Builder; needs a Proform Premium subscription.

Most systems have one number doing the heavy lifting - odds under 8, a rating rank of 3, a field of at most 12. Frontier asks the question you should ask yourself: why that number?

What it does

Pick a numeric rule and Frontier re-runs your entire back-test once for every value on a ladder - under 4, under 5, under 6, and so on - charting the trade-off at every step: how many bets you get, against what they return. It also plots the rule-off point, which is the same system with that rule removed entirely, so you can see what the rule is buying you overall.

Reading the curve

• A smooth slope means the edge fades gently as you tighten. Your exact threshold is a matter of preference, so you can take the extra volume.
• A sudden cliff means one value is carrying the whole result. That is a classic over-fitting warning - worth a robustness check before you trust it.

Found the value you want? Click Use and the rule updates to it, then re-run the full analysis to refresh the other tabs. Sweeps run one back-test per value (up to 12), tightest first, and stop early if the cohort gets too large - narrow your date range if you hit that.

Frontier is part of the System Builder and needs a Proform Premium subscription.

Yes — two ways, both from any date window (a fortnight is the classic).

As research: the Trainer breakdown

Set the back-test dates to the last two weeks, run it, then open Breakdowns → Trainer. The PRB column is the average percentage of rivals beaten by each yard's runners — above 50 means the stable is running better than mid-field even in defeat. Sort by PRB and look for yards with 0 in the Wins column: horses finishing close without getting their heads in front. PRB² alongside gives extra credit for near-misses in big fields.

As a live rule: the Trainer PRB % filters

The Connections group carries Trainer PRB % (14d) and Trainer PRB % (28d) — the same yard-wide average as a pre-race filter, computed strictly from racing before each race day. The classic “due a winner” recipe combines Trainer PRB % (14d) of 55 or more, Trainer SR % (14d) of 0, and Trainer Runs (14d) of 4 or more — stables whose form the results column has not caught up with yet. Because the figures are point-in-time, the same rule back-tests honestly over history and lists today's qualifiers.

Part of the System Builder; needs a Proform Premium subscription.

Yes. In the System Builder, click Add rule and type winning into the Search filters box. Two rules come up:

• Lbs Below Highest Winning OR — the highest handicap mark the horse has won off, minus today's mark. A value of 5 or more means the horse runs today off a mark at least 5lb below one it has already won from. Set it to -5 or lower for the reverse angle, where the horse is above its winning mark.
• Highest Winning Handicap OR — the raw mark on its own, if you would rather filter on the level than the difference.

One point to be precise about: this measures against the highest mark the horse has won off, not its most recent win. For four horses in five those are the same number, because a win pushes the mark up. They part company on a horse coming back down the weights that won off a higher mark some time ago — which is exactly the horse the filter is there to find.

Both rules are scoped to handicap wins at today's race type — not simply flat or jumps. Turf, All-Weather, Chase, Hurdle and National Hunt Flat count separately, so a turf handicap win does not count towards an All-Weather run today. The horse needs a previous handicap win at today's race type for there to be anything to measure against.

Worth knowing before you lean on it: on its own the angle is market par — 35,370 runners since 2023, impact value 0.95, A/E 1.00 at Betfair SP. Use it as a condition inside rules you already trust rather than a system on its own.

Worth pairing with: search since last win for Runs Since Last Win and Days Since Last Win.

Full walkthrough with a worked example: Has it won off a mark this high before?

Sectional timing analysis is a separate add-on, not part of the base subscription. You need a RaceMetrics Pro or Proform Premium subscription, plus the TPD Sectional Timing Add-On at £50/month — arranged by emailing admin@racemetrics.co.uk and paid by bank transfer (the full amount is passed on to TPD, our sectional data partner).

Two things worth knowing: the add-on must be set up through Proform — a subscription taken directly with TPD doesn't unlock it here — and the data provided is historical sectional data for analysis, not TPD's live in-race feed. Proform Premium desktop users who also want to export the historical data have a separate TPD option with a first-year upfront payment; email us for details.

The sectionals add-on is a separate extra (the TPD add-on) and is not included in the base subscription. It has to be purchased in addition.

Sectional data needs a Proform Premium subscription (£97.50/month, with 6-for-5 and 12-for-9 multi-month options) plus a TPD add-on arranged through Proform, not directly with TPD. Two TPD options, both £50/month: full software access with historical export (first year paid upfront by bank transfer, passed to TPD in full) or RaceMetrics-only access with no export and no upfront fee. The data provided is historical sectional data, not TPD's live feed.

Sectionals are processed around lunchtime each day and output the next morning alongside the results. Exact timing partly depends on when TPD finalise their side — some races take longer than others.

Worth restating: the sectional add-on provides post-race historical data for analysis, not TPD's live in-race feed. If you want live in-running data, that's a separate subscription taken directly with TPD.

In the desktop software's System Builder: choose the horse, add all the sectional fields in System Builder preferences, and run the system. Double-click the single-line output to drill into every run and every section for that horse, then right-click the column heading to export.

That gives you the horse's complete sectional record — every section of every run — in one exportable grid. (Requires the TPD sectional add-on with export, i.e. the software option.)

Sectionals are an add-on, not included. On top of Proform Premium there are two options: sectionals in the Windows software with full data export at £50/month but with the first year paid upfront (£600, required by TPD because the data can be exported), or £50/month for sectional access on RaceMetrics only — full history included but no bulk export and no upfront year.

No — the £50/month RaceMetrics option includes ALL past sectional history, you just can't bulk-export it. The upfront-year option exists specifically because the Windows software version allows exporting the entire dataset.

Choose the horse in the System Builder, add all the sectional fields in System Builder preferences, run the system, double-click the single-line output to drill into all runs and sections for that horse, then right-click the heading to export.

Start at www.racemetrics.co.uk/sectionals.html — every view is walked through in an 11-video series with a full written guide behind it in the help centre.

Sectional access requires both the desktop-inclusive subscription and the third-party timing add-on, which is arranged through us rather than directly with the data supplier.

Sectionals require the desktop-inclusive subscription plus the timing add-on. Payment for the add-on is by bank transfer only, using the customer's own name as the reference.

Yes, RaceMetrics 7 is completely free to enter for all registered users. There is no purchase or subscription requirement - even free tier users can compete for the weekly cash prizes. It's a skill-based competition, not gambling.

RaceMetrics 7 is our free weekly skill-based leaderboard competition. Pick 7 participants within a budget and accumulate points based on how their performance rankings change during the week. It's free to enter for all registered users – no purchase required.

User Rating 1 (UR1) is the Power Ratings — its rank, value and odds fields are the Power Rating rank, value and price. The race-guide Forecast Price is the UR1 odds: the Power Rating values converted to a 100% book (a Power Rating is roughly percentage chance × 10, so a 500-rated horse ≈ 50% chance).

The speed figure shown on the race guides is the static LTO speed figure (stLTO in the software): same race code, no zero figures, within the last year. Dynamic (dy) speed figures follow whatever filters you have set in the software at the time — filter the form book to Soft ground and the dy figures compute on Soft-ground runs only.

User Ratings 2–6 are yours to fill: import your own ratings into them (and UR1 too, if you'd rather not download the Power Ratings).

The desktop software holds a longer archive. The RaceMetrics System Builder holds less history but is where development is happening.

How far each goes back

• Proform desktop: 1997 for Britain, 2009 for Ireland.
• RaceMetrics System Builder: 2018 onwards.

Some RaceMetrics fields start later still. The RaceMetrics ratings filters and the bookmaker price fields begin in 2023, and the System Builder raises your start date with a warning if a rule you have chosen needs it.

Which to use

Use the desktop software when the question genuinely needs the deeper archive — long-run course or trainer records that predate 2018.

Use the RaceMetrics System Builder for everything else. The desktop System Builder has not been developed alongside it, so the online version carries filters, breakdowns, staking analysis and forward-looking qualifiers that the desktop tool does not have.

Yes. Every run since 2018 carries the Betfair in-running high and low — the highest and lowest prices matched during the race — and the percentage against Betfair SP.

User Rating 1 (UR1) IS the Power Ratings — rank, value and odds. The race-guide Forecast Price is the UR1 odds (the power-rating values converted to a 100% book; a power rating is roughly percentage chance × 10). The guides' speed rating is the static LTO speed figure (stLTO — same race code, no zero figures, within the last year). Dynamic (dy) speed figures follow whatever filters are set in the software at the time.

Three bookmaker price snapshots are recorded for every runner in every race:

• Overnight — 8.30pm the night before racing
• Breakfast — 7.30am on the day
• Morning — 10.30am on the day

The prices are BoyleSports. For the earlier part of the history they were bet365 — the same snapshots at the same times, with a change of firm along the way.

After the race, the industry SP, the Betfair win SP, the Betfair place SP and the Betfair in-running low and high are stored as well, so the whole price journey from the night before through to the off and beyond is held.

Every price is stored against the individual horse in the individual race and kept permanently. The desktop database holds them from March 2016 and the RaceMetrics database from 2018, covering effectively every runner.

What you can do with them

In the System Builder each snapshot is a filter in its own right. You can also filter on the price rank within the race at each point, on the percentage move between any two points to isolate steamers and drifters, and on the price as a ratio against what the Power Rating makes the horse.

Back-test results are returned at SP, at Betfair SP and at the bookmaker prices themselves including best odds guaranteed, so the profit and loss reflects the price you would actually have taken rather than SP alone. The three prices and their moves can also be added as columns on your racecard view.

This needs a distinction, because two different kinds of figure are involved.

The summary tables are not point-in-time

The course-and-distance pace statistics and the per-stall draw statistics exist as ready-made tables, with runs, wins, Impact Value, percentage of rivals beaten and the underlying sample sizes. They are refreshed periodically with the current state of play. Earlier states are not archived, so these tables cannot tell you what the draw or pace figures looked like at the time of a past race.

Using them in a back-test measures your rule against figures that include races run after the one you are testing.

The per-run rows are

Every individual run's draw, finishing position, field size and running-style classification is stored on the run itself, back to 1997. From those you can compute a fully point-in-time draw or pace figure using only races that had been run at the time.

Per-run rating aggregates are safe

The rating aggregates stored on each run — highest and median speed figures for career and last year, and the figures from the horse's previous four runs — are recorded as they stood at the time of that race. Unlike the summary tables, these are safe for point-in-time work. The same is true of the Official Rating on each run, which is the mark the horse actually carried into the race rather than a backfilled figure.

Racecards are available on the website two days before racing.

Yes. The database stores each runner's ratings and statistics as they stood at the time of that race. A row for a race run in October 2025 contains nothing that happened after it, so historical exports can be used for modelling without back-filled information leaking into them.

Prices are the obvious exception in principle rather than in practice: a starting price cannot be known before the off, and is stored as what it returned.

The one field that is not point-in-time

Draw impact values are not stored at a fixed point in time. They are rebuilt periodically against the full record for each stall, so the figure attached to an old race reflects the most recent rebuild rather than what the figure was on the day.

The draw itself is historical — the stall number a horse raced from is recorded on the row — so a point-in-time impact value can be recalculated outside the software from exported data, using only races that had already been run at the date you are testing.

What you can export

The Windows desktop software, included with Proform Premium, exports both historical data and forward racecard data, and the ratings go with it. You choose which fields are included, and the export can be scheduled to run daily on your own machine so tomorrow's card is waiting for you each morning.

Yes. A Proform Premium subscription includes the Windows desktop software, and the data behind it sits in a local database on your own machine. Direct access is available: we supply the basic schema design and the connection details.

It is worth being straight about who it suits. There is no query builder and no guided interface — you are writing SQL against the tables yourself, so it assumes you already know SQL well. If you want the data without that, the export route gives you the same information: you pick the fields, and the export can be scheduled to run daily.

What this is not

The plain-English query tool inside RaceMetrics is a different thing. It answers racing questions and deliberately refuses requests to list tables and columns or to run arbitrary SQL, because it runs against our servers rather than your machine. Direct SQL access applies to your own local database, not to ours.

If you want to go this route, get in touch after subscribing and we will send the schema and access details.

RaceMetrics uses official racing data covering UK and Irish racing. Our database includes comprehensive historical performance data. P/L calculations use Betfair Starting Prices (BSP) for accurate profit tracking.

P/L shows the theoretical profit or loss if you had backed every runner to £1 stakes at Betfair Starting Price (BSP). This gives a standardised measure of value performance, removing the subjectivity of individual price judgements.

RaceMetrics uses official racing data covering UK and Irish racing. Results are updated daily and include Betfair SP prices for P/L calculations. Our database includes comprehensive historical performance data.

Yes. When a horse is withdrawn from a flat race, the stalls handlers on the racecourse physically close the stalls up, so the remaining runners load from renumbered positions — a horse declared in stall 14 of 14 races from stall 12 if two horses drawn below it come out. The official result records that renumbered draw. The only exception is a very late withdrawal at the start, when there is no time to re-draw.

How RaceMetrics handles it

RaceMetrics applies the same renumbering live: the moment a non-runner is declared, the draw on the racecard is re-shuffled and Draw IV is recomputed for the new stalls — so the card, System Builder qualifiers and the eventual back-test all agree. Many racing sites carry the declaration draw unchanged all day, which is why their draw data looks wrong whenever there are non-runners.

For System Builder users this means a runner can move into or out of a draw-based system when the stalls re-shuffle. With qualifier emails switched on, the intraday follow-up email lists both the new qualifiers and any morning picks that no longer qualify. The Draw Group (low/high half) filter is more robust to re-shuffles than a narrow Draw IV band, since a renumbering rarely moves a horse from one half of the stalls to the other.

No — and there's an honest reason. With a data product, the value is the ratings, form analysis and system results themselves: everything could be read, noted down and taken away in a single afternoon, so a "trial" is really just the product, free. To keep prices fair for paying members, we don't do it.

What we do instead: the RaceMetrics free account unlocks one full meeting every day with real ratings and racecards — the closest thing to a trial there is. Our YouTube walkthroughs show every tool being used on real races, and both help centres document every feature with screenshots. And every subscription is a monthly rolling plan with no lock-in — if it's not for you, cancel any time and you've risked one month, not a contract.

There is no free trial, for an honest reason: with a data product the value is the ratings, form analysis and system results themselves, which can be read and taken away in an afternoon — a trial would just be the product, free. Instead: the RaceMetrics free account unlocks one full meeting every day with real ratings and racecards (the closest thing to a trial), YouTube walkthrough videos show the tools on real races, both help centres document every feature, and every subscription is a monthly rolling plan with no lock-in — cancel any time.

If this is your first renewal on the current Proform website: wait until your subscription expires, then log in and renew from your account — you can't renew early that first time. After that first renewal, your subscription becomes recurring automatically.

Your access and your saved work aren't lost during a short gap at first renewal. If anything looks wrong after renewing, email admin@racemetrics.co.uk.

They are two separate products, sold on two separate subscriptions. Most people only need one.

RaceMetrics

The website. It runs in a browser on anything — Mac, PC, phone or tablet — with nothing to install.

• RaceMetrics Ratings for horses, trainers, jockeys, sires, dams, damsires and owners
• Racecards, form study and the Form Expert
• The System Builder, for building and back-testing your own systems
• Saved patterns and daily alerts

It is sold on a monthly ladder, from a free tier up to RaceMetrics Pro. Current prices are on the pricing page.

Proform Premium

A separate subscription that adds the Windows desktop software on top of full RaceMetrics access.

• The Proform desktop software, with the full database and bulk CSV export
• Race Guides
• The Bet, Lay and Trade Finder
• Daily stats alerts

The desktop software is Windows only. If you are on a Mac, RaceMetrics is the one that will work for you.

Which one

Choose RaceMetrics if you want ratings, racecards and system building on any device. Choose Proform Premium if you also need the desktop software, the Race Guides and the Finder tools, or if you need bulk data export.

Two things worth knowing before you subscribe

Sectional data is not included in either subscription. It comes from a third-party timing supplier and is a separate paid add-on, arranged through us rather than directly with the supplier.

There is no in-place upgrade between the two products. If you are on one and want the other, the first subscription is cancelled and refunded pro-rata, and you subscribe to the second.

For the first renewal on the new Proform website you need to wait until the subscription expires, then log in and renew. After that first renewal it becomes recurring automatically.

Yes, you can upgrade or downgrade your subscription at any time from your account settings. Upgrades take effect immediately with pro-rated billing. Downgrades take effect at the end of your current billing period.

Yes, you can cancel your subscription at any time from your account settings. When you cancel, you'll retain access until the end of your current billing period, then your account will revert to the free tier.

Yes. RaceMetrics partners earn 30% of every referred subscriber's first payment on any paid tier, and their audience gets 20% off for their first 3 months via a unique partner code. Apply by emailing admin@racemetrics.co.uk.

Each partner receives a unique discount code. Every checkout where the code is redeemed is attributed to that partner — no cookies or tracking links required, so attribution survives ad-blockers and cross-device journeys.

Yes — editorial sites, bloggers and YouTubers producing a genuine review can request 1 month of complimentary Proform Premium access (the top subscription, including sectional timing analysis on RaceMetrics and the System Builder) by emailing admin@racemetrics.co.uk.

Create a free account at www.racemetrics.co.uk using the same email address that's registered to your Proform software. You can check which email that is inside the software: open the options menu and look at the registration screen.

When the emails match, the system recognises your subscription automatically and unlocks the features included with your tier — Proform Premium subscribers get RaceMetrics access included. If you've accidentally registered with a different email, contact admin@racemetrics.co.uk and we'll link the accounts.

You may run the software on two computers per subscription.

Moving to a new machine

Do these in order. Most problems come from skipping the first step.

• Update the existing installation first. Log in to the Proform website, go to the account page and download the latest build. An old build cannot produce a backup that the current version will restore.
• Take a backup from within the software.
• Install the current build on the new machine.
• Restore the backup.

Registration details

Registration details are issued on request — email us and quote your Proform key.

The registration fields and the security checks are case sensitive. Enter them exactly as supplied, including capitalisation, or the check will fail.

Create an account at www.racemetrics.co.uk using the SAME email address registered to your Proform software. You can verify your registered email in the software under the options menu, registration screen. Using the matching email lets the system recognise the subscription automatically.

Daily limits (profile views) reset at midnight UK time. If you reach your limit, you can either wait for the reset or upgrade to a higher tier. Pattern limits are total (not daily) and can be increased by upgrading your subscription.

Daily limits reset at midnight UK time. If you hit your limit, you can either wait for the reset or upgrade to a higher tier for more capacity. Pattern limits are total (not daily) and can be increased by upgrading.

Yes and no — it depends which of the two products you mean.

RaceMetrics works on a Mac

RaceMetrics runs in the browser. It works on a Mac, PC, phone or tablet, and there is nothing to install. That includes the RaceMetrics Ratings, racecards, form study, the System Builder and alerts.

The desktop software is Windows only

The Proform desktop software is a Windows application. There is no Mac build.

Two things are only available through it:

• Bulk CSV export of the database
• Export of a horse's full sectional history

If you need either of those, you need a Windows machine.

Some customers run the desktop software on a Mac through a virtualisation layer. That does work, but the virtualisation software is a separate licence at your own cost, and several customers have moved to a Windows laptop instead once they weighed the two.

Yes. It is a false positive, and there is nothing wrong with your download or your installation.

Proform is a specialist application with a small user base. Antivirus engines score unfamiliar executables on how many machines have run them, so a program that only a few thousand people install can be flagged purely for being uncommon, with no malicious code involved.

What to do

• Allow the file, or add the Proform installation folder to your antivirus exclusions.
• Only ever install from the download button on your account page on the Proform website. Do not install a copy sent to you by anyone else.
• If the antivirus has already quarantined the file, restore it from quarantine and add the exclusion before running the installer again.

If the installation still will not complete after that, contact support and a remote-support session can be arranged.

No. Microsoft SQL Server does not run on ARM processors, and the Proform desktop database needs SQL Server, so the software cannot be installed on an ARM machine.

The usual symptom is error 1067 from the database service on startup, after what looks like a successful install.

Check before you buy

Machines sold with Snapdragon X, Snapdragon X Plus or Snapdragon X Elite processors are ARM based. So are Copilot+ PCs built on those chips. If the specification lists an Intel or AMD processor, it is not ARM and will be fine.

If you already have one

RaceMetrics runs entirely in the browser and needs no local database, so an ARM laptop will run RaceMetrics without any difficulty. Only the Windows desktop software is affected.

The Proform Key is generated by your machine when the software first runs. It identifies that installation, and support uses it to issue your registration details.

What to do

1. Note the Proform Key shown on the registration screen.
2. Send it to support, from the email address your subscription is held under.
3. You are sent back the details to type into the registration screen.
4. Enter them exactly as supplied and restart the software.

The common mistake

The registration fields and the security checks are case sensitive. Most failed registrations are a capital letter typed in lower case, or a trailing space picked up when copying and pasting.

If you install on a second machine, that machine generates its own Proform Key and needs its own registration details.

Almost always one of three things, and none of them means the installation is broken.

1. Filters are applied to exports

Exports honour whatever filters are set in the data view. An active filter set can exclude nearly everything, leaving a file with headers and no rows. The filter button shows red when filters are applied — clear them and export again.

2. The local database has not been updated

If the database has not been brought up to date since installation there is nothing to export. On the welcome screen, tick both boxes and click update. On a new installation this is the usual cause.

3. Sectional fields use a separate export

There are two export buttons. Sectional fields are exported through their own button, separately from the standard export, so selecting sectional columns in the standard export returns nothing for them. Export the two separately.

On a large export, a temporary not responding message while the file is being written is normal — leave it to finish.

This is an installation fault rather than a registration problem — a registration code will not fix it, and it is not something you have done wrong. It usually means the local database engine did not finish installing or is not starting with Windows.

If it happened on a fresh install

Get in touch and we will arrange a live support session. These faults are quick to fix with a look at the machine and slow to fix by email, so it is worth going straight to that rather than working through it blind.

If it happened after restoring a backup

There is a recovery order that works, and the first step matters most.

• Do not uninstall. Uninstalling makes this harder to recover, not easier.
• Download the desktop application installer from your account page — the application itself, not the database engine installer.
• Install it, reboot, and check the application opens and runs.
• If you get a provider error at that point, delete the version file in the application's help folder and run it again.
• Once a clean copy runs, copy your backup file onto the machine's own drive and restore from there — not directly from a USB drive.

If it still fails after that, the backup file itself is either damaged or from a version the machine cannot read.

The version rule for backups

A backup can be restored onto the same version or a newer one, but never onto an older one. Restoring a backup taken from an up-to-date machine onto a laptop running an older build will fail, and this is the most common cause. Update both machines to the same build before you copy a database between them.