a performance figure supplied by the racecard provider for each past run: a speed rating on the Flat and a performance rating over jumps. Expert View shows the best and latest figure from the last twelve months.
How to read, research use and limitsHide detail
How to read it
Higher is better. "Spd best" or "Perf best" is the horse's top figure in the last year, "last" its most recent run, "vs Par" the gap to the typical winning figure for the course, distance and class, and "Spd rank" its place in today's field.
Research use
A quick way to rank a field on recent speed or performance rather than on finishing positions.
Limits
These are the provider's own figures, not Racing Post TS or RPR, which Furlong Lab does not licence. The method behind them is not published to the site, so they are shown as supplied and not combined with the site's own figures.
Also: Spd best · Perf best · speed figure · performance rating · vs Par
Furlong Lab's research score for a runner, 0 to 100, built from up to sixteen weighted factors (recent form, speed, weight, class, distance, surface, going, course, pace, draw, fitness, progression, trainer, jockey, consistency, equipment) scored from the horse's own prior runs known before the race.
How to read, research use and limitsHide detail
How to read it
50 is neutral; each factor is pulled towards 50 when its evidence is thin, so a high score needs both good evidence and plenty of it. Data confidence and a provisional, supported or insufficient status are shown beside the score.
Research use
A structured summary of the form book that shows which factors support a runner and which are unknown, so you can check the evidence behind each.
Limits
The score is a research and display figure, not a win probability and not a betting system; it uses no odds and books no bets. It is on its own scale and is not comparable with pounds ratings such as OR. Racecards show it only before the off.
Also: FL Form · FL Form score · form percentage · Form %
a time comparison that finds each runner's quickest past run (seconds per furlong) at today's exact distance and going, and ranks the field on it. The strict version uses only the last five starts.
How to read, research use and limitsHide detail
How to read it
The ranked list shows each horse's best matching time, how many runs ago it was set, and whether that figure is proven, quicker or slower than the horse's own average under those conditions. The betting angle version adds a weight check: the horse must carry at least 2 lb less than in the matched run, and only applies up to 5f.
Research use
Separates horses by what they have actually clocked under today's conditions rather than by finishing position.
Limits
Times come from the recorded race times. Non-winners' times are derived from the winner's time and the recorded beaten margins (five lengths per second), not timed individually. Times are not adjusted for pace, wind or daily track speed, so a figure from a fast-run race flatters the horse. Runs in restricted-rider and non-handicap races are excluded as anchors for the weight check.
Also: fastest · time ranking · fastest strict · seconds per furlong
Furlong Lab's own performance figure for a completed race: the official-rating-equivalent level each finisher ran to, worked out from the beaten margins and the official ratings of the opposition.
How to read, research use and limitsHide detail
How to read it
On the same pounds scale as an official rating, so "ran to 92" reads directly against a mark of 85. Each figure carries a high, medium or low confidence based on how many of the field were rated.
Research use
Shows whether a horse's recent runs were better or worse than its mark, and feeds the FL Rating Formula used on racecards and daily scans.
Limits
Built from beaten lengths and ORs with a distance- and code-aware pounds ladder; in fields with few rated horses the figure falls back to a low-confidence class prior. Figures are as-of-day only. The legacy form figure shown in some angle tools is a separate, unvalidated output kept apart from betting evidence.
Also: RanTo · ran to · ran-to · FL Rating Formula
racecard research signals built from the FL conditions scores: the gap between the top-rated horse and the second, joint leaders, and the daily stand outs where the leader is clear by at least 5 points with every runner rated.
How to read, research use and limitsHide detail
How to read it
"Gap 6.5" means the leader's displayed score is 6.5 points above the second horse. Stand outs list the reasons behind their confidence grade: how many of the field are rated, how many comparable runs the leader has, and whether declarations were checked within 24 hours.
Research use
Points you at races where the figures separate the field clearly, which is where checking the evidence is most worthwhile.
Limits
A gap is a descriptive ordering of displayed scores; it introduces no qualification threshold, stake or prediction. Stand outs are a research list with its own published record, not advice.
Also: FL gap · stand outs · standouts · FL Intelligence · clear top two
the three ground groups the site uses for statistics: Firm-side (firm, hard, fast, standard to fast), Good (good, good to firm, standard) and Soft-side (good to soft, soft, heavy, yielding, slow, standard to slow).
How to read, research use and limitsHide detail
How to read it
Record splits labelled by band pool similar goings so samples are usable. Good to Firm counts as Good; Good to Soft counts as Soft-side.
Research use
Lets a horse's ground record be read from a handful of runs instead of one or two on each exact description.
Limits
The grouping is a convention: Good to Firm and Good to Soft are different ground even though both neighbour Good. Exact descriptions are still shown where they are available.
Also: ground band · Firm-side · Soft-side · going group
how record tables group races by level: by UK Class 1 to 7 where the class is recorded, and otherwise by the handicap rating band (for example 0-75) or the race type (Maiden, Novice, Conditions, Listed, Group).
How to read, research use and limitsHide detail
How to read it
A row labelled "Class 4" holds every bet in Class 4 races; "0-75" holds bets in that handicap band; "Not recorded" means the race metadata was unavailable.
Research use
Shows whether a record or a horse performs at a particular level, and whether today's race is a step up or down.
Limits
Irish and overseas races carry no UK class, so they group by band or type. Historical rows without recorded metadata stay unrecorded rather than guessed.
Also: class grouping · band / race type · class move · class drop
whether stall position has mattered at this course and distance. Past Flat races are split into low, mid and high draw bands by each race's own field size and the win and place rates of each band are compared.
How to read, research use and limitsHide detail
How to read it
A notable bias is only flagged when at least 30 past races are available and the best band's strike rate beats the worst by 7 percentage points or more. Otherwise the fractions are shown plain, with no edge claimed.
Research use
Tells you whether a runner's stall is a real factor at this track before you weigh it.
Limits
Draw effects shift with ground, field size, pace and rail movements; the profile cannot see today's pace or rail position. A flagged bias is a historical tendency, not a prediction.
Also: draw advantage · stall bias · low draw · high draw
the racecard's picture of how the race may be run: each active runner's inferred run style laid out from front-runners to hold-up horses, with the projected pace scenario.
How to read, research use and limitsHide detail
How to read it
Front-runners on the left, hold-up horses on the right, unknown styles listed separately. Counts beside each style show how many recent runs supported it. The pace collapse flag appears when at least three front-runners and one hold-up horse are classified in a field of five or more.
Research use
A race with several leaders may fall apart for the closers; one with a lone leader may favour the front.
Limits
Styles are inferred from past race comments, not in-running positions, and unknown runners are never guessed. Jockeys change tactics; the map is a projection, not a plan.
Also: pace · pace collapse · pace setup · running styles
the trainer's or jockey's wins and runs over the fourteen completed days before the race date, from recorded GB and Irish results across all codes.
How to read, research use and limitsHide detail
How to read it
Shown as wins/runs and a percentage (Expert View "Trn 14d", race page "Last 14 days"). An "in form" badge appears only when the sample and rate clear the page's thresholds. N/A means no eligible recorded runs, not 0%.
Research use
A quick read on whether a yard or rider is firing; best used alongside the longer-term and at-course records on the same page.
Limits
Fourteen days is a small window and the figure swings on a handful of results. Result coverage may be incomplete and runs with unavailable outcomes are excluded. It is a win strike rate, separate from the provider's run-to-form figure.
Also: Trn 14d · trainer 14 days · jockey 14-day · last 14 days · 14-day form · in form
the racecard provider's figure for the share of a trainer's recent runners that ran to their expected form, as supplied on the card.
How to read, research use and limitsHide detail
How to read it
A percentage per trainer; higher means more of the yard's recent runners performed to expectation. Shown as supplied, with no site calculation.
Research use
A complementary read to the 14-day win rate: a yard can be running well without winning.
Limits
The provider's definition of "ran to form" and its window are not published to this site, so the figure cannot be reproduced or verified here. It is blank when the provider does not supply it.
Also: RTF · run to form · runs to form %
a trainer's, jockey's, sire's or horse's wins divided by runs, and placings (top three) divided by runs, over the stated period or at the stated course.
How to read, research use and limitsHide detail
How to read it
Shown as a percentage with the wins/runs or places/runs fraction beside it. Samples under five runs are shown as the fraction only and marked small.
Research use
Course and condition records show whether a connection or horse has a pattern worth noting; read them with A/E to see whether the wins beat the prices.
Limits
A raw win rate ignores the prices, so a 30% rate on short-priced favourites can be a losing record. Placings use the top-three convention regardless of field size.
Also: win % · place % · Win SR · Place SR · win rate · place rate
the one-to-five star grade on each Furlong Buddy selection. Stars order the model's confidence and set the notional stake in units (a 3★ pick is a 3U win bet).
How to read, research use and limitsHide detail
How to read it
More stars means the pick was graded higher and staked bigger; the record page shows settled picks and P/L by star.
Research use
Comparing the record by star shows whether higher grades have earned their larger stakes so far.
Limits
Stars are not probabilities: no win chance was declared when the picks were generated, so a star cannot be relabelled as a calibrated percentage. The overall record is loss-making on every price basis to date and no star or filter has been shown to be profitable on later picks.
Also: stars · star rating · Buddy stars · ★
the size of a market move measured on implied probability: the horse's current implied chance divided by its opening implied chance, minus one. A move from 10.00 to 5.00 is a shift of 1.00 (the chance doubled).
How to read, research use and limitsHide detail
How to read it
The Price Crash record groups selections by shift (0.25-0.49, 0.50-0.99, 1.00+). The stored signal is the relative change in implied probability, not a percentage cut in the odds.
Research use
Measuring on probability treats a move from 100/1 to 50/1 and from 2/1 to evens as what they are: the same doubling of chance, which a change in odds would hide.
Limits
The shift describes where money went, not why. The published Price Crash record shows how following such moves has fared; it is not a forecast for the next one.
Also: implied probability shift · price crash · shift · steamer shift
the FL Ratings model's chance of each runner winning, from a race-level model whose weights were learned on 2010 to 2021 and checked on 2022 to 2023 before being frozen. The field's chances sum to one.
How to read, research use and limitsHide detail
How to read it
Shown as a percentage beside the score and rank from the 10:00 capture, with a confidence range for how far the chance moves if the horse's ability were one standard deviation higher or lower.
Research use
A consistent, point-in-time chance for every runner that you can compare with the market and with your own view.
Limits
In research the model beats simple baselines but not the market: its probabilities sit well behind the eventual SP in log loss, and the value flags built from them lose money at obtainable prices. The 0-100 score beside it is a display scale, not a win percentage.
Also: win chance · FL Ratings probability · model probability
the price that matches a probability with no margin: 1 ÷ win probability. A 25% chance has fair odds of 4.00 (3/1).
How to read, research use and limitsHide detail
How to read it
On FL Ratings and Value Machine pages the fair odds sit beside the model's or calibrated probability; a quoted price above fair odds is bigger than the model thinks it should be.
Research use
The direct bridge between a probability and a price, used in every value comparison on the site.
Limits
Fair odds are only as good as the probability behind them. Furlong Lab's model-based fair odds have not beaten the market in testing, so a price above them is a research label, not value in the bank.
Also: fair price · true odds · fair value
the expected profit per unit staked at a quoted price: win probability × decimal odds − 1. A 25% chance at 5.00 has an edge of +25%; at 3.00 the edge is −25%.
How to read, research use and limitsHide detail
How to read it
FL Ratings shows the expected ROI at the quoted price and an uncertainty-adjusted version; a research flag is attached when the price clears the fair odds and the stricter ledger's rules.
Research use
Turns a probability and a price into one number you can compare across runners.
Limits
The research ledgers that follow these flags have lost between 16% and 21% of stakes at obtainable prices in every evaluation, so the flags are labelled research only and are not recommendations. The edge inherits every error in the probability.
Also: edge · expected ROI · EV · expected value · value flag
the Value Machine's comparison of its calibrated probability with the market's: calibrated probability ÷ market probability − 1, capped at +1.5, shown as a band (strong, fair, slight, none, negative).
How to read, research use and limitsHide detail
How to read it
An edge of +0.20 means the calibrated chance is 20% higher than the price implies. The absolute difference (calibrated minus market probability) is shown too, with an evidence note on the sample behind the calibration.
Research use
Highlights runners the model disagrees with the market on, which is where checking the evidence is most useful.
Limits
The calibration is applied at read time to stored model figures and is a presentation layer; the stored raw probabilities are the evaluation evidence. An edge is a disagreement with the market, not proof the model is right; the Value Machine's own record should be read before acting on any band.
Also: calibrated edge · VM edge · relative value · value band
tags on a runner's last runs showing a recorded excuse: a severe one (badly hampered, brought down, denied a clear run, stumbled, carried wide) or a softer one (slowly away, never nearer, ran on late, green), read from the official in-running comment, plus a backed-and-beaten tag where the horse was heavily supported and lost.
How to read, research use and limitsHide detail
How to read it
Each tag names the excuse and links to the run. The baseline beside the list shows how horses with that kind of excuse have actually fared next time, as wins against the wins their SPs expected.
Research use
Finds horses whose bare form figures may understate them, with evidence of how often such excuses have translated into a better run.
Limits
Nothing in the tag says the horse will improve; the baseline shows the historical rate and it is close to market expectation. Comments exist from August 2024 and price moves from February 2026, so older runs carry no tags.
Also: second chance · excuse · hampered · backed and beaten
an Angles Hub rule: a runner in a GB or Irish handicap racing today off an official mark lower than the mark of its most recent handicap win in the same code (Flat, hurdle and chase marks are never mixed).
How to read, research use and limitsHide detail
How to read it
Shown with the current and effective marks, the last winning mark and a tier (strong, positive, standard). Stale wins, long winless runs and declining form reduce the tier rather than being hidden.
Research use
A well-handicapped horse is one the handicapper has let down to a mark it has already won off; this rule finds them without a ratings calculation.
Limits
The mark alone does not say why the horse has slipped: age, injury or a change of yard may explain it. The rule's historical record and thresholds are documented in the research folder and it is a research angle, not a tip.
Also: BLWM · last winning mark · well handicapped · winning mark