UK & Ireland Horse Racing Analysis & Value-Ranked Picks, The Overlay

Today's call is our single best play of the day, explained in plain English: how much bigger we make it than the market, and why. The two cards beside it are our next-best calls. These count toward our public record.

Speculative picks are longer-shot ideas we like the look of. We are not backing these and they never touch our record or ROI: they are just for interest.

It is all observation mode: nothing is staked. We publish, settle at Betfair SP, and show the result, win or lose. Tap any pick, row or meeting to open that race in Race Lab.

Everything is free. The Overlay is a free research and insights platform. Every view and every tool, the full Ratings Board (every runner's A/E), the complete top-rated list, the daily selections, the calibration record and the day's meetings, is open to everyone with no account and no charge. We are a data-driven analysis product, never a tipping service: we publish, settle at Betfair SP and show the result, win or lose.

Our whole promise is calibration: when we say a horse has an X% chance, it should win about X% of the time. The Model calibration page now opens with a reliability chart that proves it. Each dot is a band of our ratings, plotted as our predicted win chance (across) against the real win rate (up). The closer a dot sits to the dotted line, the more honest that rating is.

It is measured forward and leak-free: every rating was stamped before the race ran, so nothing is fitted after the result. We also plot the market's chance for the same horses, and show our probability score (Brier) next to the market's. We sit a whisker behind it: we read chances about as well as the market does, and we never claim to beat it. That is the ceiling, and we would rather show it than sell past it. The chart refreshes daily and settles at Betfair SP.

On the Today page and the Ratings Board, each marker tells you why a runner is flagged:

• TOP RATED / Today's standouts: our highest-rated runners today. The number is our model rating (0–100). We select here on rating, not on beating a single price: over a 365-day backtest these ran to A/E ~1.12 at Betfair SP (above break-even), whereas odds-on shots and the old connection-only gate did not.
• ★ TOP RATED pill: one of today's higher-rated runners in the race.
• This type wins ~X% chip: a live calibration cue on the suggested horse. It shows how often runners in that horse's Verdict tier have actually won over our recent rolling sample (the same numbers as the calibration strip on Performance). It describes how the type performs, not a guarantee for that individual horse, and it is a win rate, not a profit claim.
• Confidence tiers (the pill on each runner): our rating sorted into calibrated win-rate bands, re-based July 2026 to what actually happens. TOP RATED (score 60+, the genuine top slice, wins ~25% over our rolling sample), SHORTLIST (45–59, ~20%), WATCH (30–44, ~16%), PASS below. A rare STANDOUT pill marks the elite 70+ runners. These are win-rate bands, not profit claims,: on the exchange our rated runners in the 2/1–7/1 range have held up, whereas odds-on shots do not pay their way.
• VALUE +Xpp chip: our sharpest honest signal. We compare our calibrated win probability (what runners on this score actually win at) with the probability a fair-book price implies, with the bookmaker's overround stripped out so we compare like with like. When our read is materially higher (a +4pp edge or more) and the price sits in the value window (2/1–10/1), the chip shows the gap, e.g. VALUE +7pp means we make it a 7-point bigger chance than a fair market does. It is a calibration read, not a profit promise.
• Green angles (e.g. 3/7): how many of the seven A/E angles beat the market for this runner. More green = the data likes more about it.
• A/E heat colours: each angle cell is shaded by its A/E, deep green 1.25+, light green 1.05–1.25, neutral 0.95–1.05, then pink/red below. The small number beside each figure is the sample size.
• Angle snippets under a horse's name (e.g. Trainer here 5/12, Jockey here 3/8, Won here): the connections' record at this course and their recent form, read straight from our 1.8M-run history, with the trainer's A/E at the track where we have it. Green = a strong record, pink = a poor one, grey = neutral. These are reminders to look closer, context not a tip.
• ★ nap / ★ N naps badge (gold, beside a horse's name): this runner is today's NAP for one or more named public tipsters, taken from the Sporting Life naps table. Hover to see which tipsters. It flags where the wider tipping crowd lands, shown as context alongside our own rating, not as our pick. The full picture is on the Tipster naps board.

On a phone the racecard reads lean by design. The model's selection gets a hero row labelled Top rated: a slim bar showing our rated win chance (gold) against the market’s implied chance (the tick), plus a one-line read. Every other runner is a field row showing one red→green ladder of where it sits in the race, its tier (Top rated / Shortlist / Watch / Pass) and, where it applies, the single ★ Standout flag. The race read at the top carries a colour kicker chip: red for a sit-out (“watch, not a bet”), green for a backable top-rated pick. Everything else (the score, the market-favourite and bounceback notes, the connection angles) now lives one tap down, under Flags & angles in the runner's drop-down, so the surface stays clean and the detail is always a tap away.

Results & prices: as each race runs its card shows the outcome: green ✓ WON, gold placed, red LOST, or slate N/R (non-runner). Prices are shown in your chosen format (set Fractions / Decimals / US in the footer); a green ↑ means it shortened (you beat the price), a red ↓ means it drifted.

Legacy badges. You may still see ★ NAP, ● STRONG, ⚖ WELL-IN or ◷ UNLUCKY LAST TIME on the older scoring surfaces (the Verdict tiers and the Performance history). These belong to the retired Sporting-Life-tuned staking model, which is paused: they are kept for reference and are no longer staked. The live product is the top-rated model above.

On the Racecards page, the Today's selections panel in the sidebar is our committed daily call. We run one model, our own runner-rating, and put up its top-rated runner at each meeting, provided that horse also clears the day's quality bar (the top third of all our ratings that day), so we never put up the best of a poor card. Some meetings therefore carry no selection.

Each row shows the time, course, horse, our rating for it, and the morning price. The Open record in the corner is the honest running total: how many selections have run, how many won, the strike rate and the return.

How we "back" it: each selection is a 1pt notional level stake, settled honestly at Starting Price (Betfair SP where available) and tracked openly. It is our model's call, led by how well-calibrated our ratings are, not a promise of profit: our own research finds no reliable edge over the exchange on this data, so we show the real record rather than a sales pitch. Nothing here is a real bet placed on your behalf.

This lane replaces the paused Sporting-Life staking picks. It is watched daily by our data monitor, so if it ever stops putting up a pick, we know at once rather than weeks later.

The US Racing page (in the Browse menu) shows today's and tomorrow's cards, plus results, from the major American tracks (Churchill Downs, Gulfstream, Santa Anita, Saratoga and the like), pulled from the same free Sporting Life feed as our UK and Ireland cards. Use the Today / Tomorrow toggle to switch day. There are also dedicated pages for the major tracks (Churchill Downs, Keeneland, Gulfstream Park, Santa Anita, Saratoga, Del Mar) and a Breeders' Cup 2026 guide, linked from the top of the US Racing page.

You can back any US runner with play money too: tap Bet on a card, pick a stake, and your virtual wallet (the same play balance as the UK cards) auto-settles it from the result. Purely for fun, never real cash, and still no Overlay score involved.

The US Read (beta). Our full UK and Ireland model does not carry across to US racing, so we do not run it there. Instead each US card carries a Read: a within-race strength number built only from the official rating, the one signal that beat the market when we backtested it against real US results. 50 is the middle of the field, higher is better rated, and a star marks the top. It is not our UK model and not a betting edge: US racing gives us no per-runner times or sectionals, so we cannot build the speed and pace figures that make the UK ratings earn. We show it transparently, label it beta, and track how it runs.

Reading a US card: No is the program (saddlecloth) number and PP is the post position (the stall). A chip marks the surface (dirt or turf), the class is read from the race title (Maiden Claiming, Allowance, Stakes and so on), and any medication flag (L, B) sits by the horse. Weights are in stones and pounds, and times are shown UK. As each race runs the card flips to a result: the finishing order with returned prices, winner highlighted.

Build a bet (Top-N finish). On any UK or Ireland race in the Race Lab, the header carries a Build a Top-N bet button. It opens a builder where you can back a runner to finish in the Top 2, Top 3 or Top 4, a play-money market we settle from our own results (win if it finishes within the places you chose, void if it is a non-runner, dead-heats on the boundary paid on a reduced stake). The price next to each runner is a fair price built only from our own calibrated chances (the win chances turned into a place probability, then into decimal odds): no bookmaker margin, no claimed edge, never a market quote. We only price UK and Ireland cards, where we rate every runner. Where our rating cannot separate several runners they will show the same fair price, which is honest: it means we rate their chances the same.

US racing runs from the afternoon into the small hours UK time, so the page fills through the evening. It is watched by our data monitor like every other feed, so if the US cards stop updating we know at once.

Betting US races. You can place win or each-way bets on US cards exactly like the UK: tap a price to open the same slip. We settle them from our own US results, win and place, with place terms set by the field size (win only up to 4 runners, 2 places for 5 to 7, 3 places for 8 or more), non-runners voided with your stake back and dead-heats paid on a reduced stake. US tracks themselves run win / place / show rather than UK-style each-way, so this is our UK-style each-way applied to the American card, in dollars.

US best bets (tracked). The US best bets page (in the Browse menu) lists Timeform's US racing best bets for the day, published on Sporting Life, with the advised price and the result once each race is run: won, placed or lost. Winners are highlighted, and a running record shows how the day is going. We do not author these tips and, as above, we do not rate US racing ourselves: this is the named expert's call, and an honest record of how it runs. Because the US column is only posted for the evening and weekend cards, the board is empty earlier in the day rather than inventing picks.

Today's call leads with a single bold panel, Our Top Pick. The gold dial on the left is the model's rating (1.00 means level with the market; higher is bigger value), next to it the horse, the race and the price, a one-line plain-English reason with the A/E evidence behind it, and an "Our edge vs the market" bar that fills with the size of the overlay. Beneath it sit the Next-best calls and the Speculative wildcards (fun only, never staked, never in the record).

Under the calls, the From the blog banner links to our Racing data & insights guides. It opens in a new tab.

On wide screens (roughly 1620px and up) a slim panel rides down the right margin linking to our World Cup 2026 hub. It follows you across the site and opens in a new tab. On narrower screens it tucks away so it never crowds the content.

The Top-Rated Runners panel on the Today page lists the model's strongest-rated runners. These come from a combination model built on 12 months of licensed data: it blends a runner's trainer, jockey, trainer-and-jockey combination and sire/damsire records into a single rating (shown as the number on each row). Roughly the strongest 10% of runners each day make the list.

These are ratings, not tips. The number is our model's read on each runner, published for research. Our wider testing finds no reliable betting edge over the market, so treat these as a starting point for your own analysis, not as bets.

On an individual race, a runner the model rates highly carries a green ★ TOP RATED pill on the Ratings Board.

Shown, never staked. We publish every rated runner and its result, win or lose, but nothing here is a staked bet.

The board locks at 10:30 BST once the final field is set, so it won't shuffle as the market moves. After that only results change, and any late non-runner drops off automatically.

A small red dot on a race means no Top Rated runner: the data has no strong opinion, so sit it out. Chasing the best-of-the-rest in those races has been a reliable way to lose money. The runners stay visible for research; tap Why? for the backtest behind it.

An amber 🔒 MARKET-BACKED FAVOURITE chip means the favourite won last time and is heavily backed, a shape where the market is usually right, so the model steps back rather than oppose it.

Top nav:

• Race Lab (the home page): the app opens here, on the next race off. A per-race breakdown with a Verdict rating on every runner: how well it fits today's race on form, conditions and connections (Standout / Top rated / Shortlist / Watch / Pass), with the model's working in the tabs below. The Velocity tab shows a Pace map of the field's likely early shape (front-runners through to hold-up horses). The race header also shows Avg OR, the average official rating of the runners, a quick read on the race's class level.
• Yesterday / Today / Tomorrow: the day's card and the model's plain-English standout.
• Velocity: cross-card pace map with a one-line take per race, plus a speed check on the race shape.
• Angles: today's runners that match a notable angle, course winners, in-form yards, trainers who fly at the track, headgear changes, class droppers and likely front-runners. Each drills through to the race. Context, never a tip.
• Angle Lab: build an angle, see its honest forward record, back today's qualifiers.
• Performance: full pick history, ROI and calibration.

The Overlay's read sits at the top of every race: one labelled panel that states the single answer, either the top-rated pick (with the value angle, the dominance gap and a pace-fit check) or a clear "sit this race out" with the reason why.

Below that one-liner, the read now spells out the why in plain English: a short, honest explanation of what the model likes about the top-rated runner (the factors doing the most work, ranked by their real weight in the score), how clearly it tops the field, and an honest caveat where the current form or the price says be careful. Every read ends with the calibration reminder, for example "horses we rate this way win only about 18 in 100, so this is a rating of chance, not a tip". It is written from the same numbers behind the Verdict donut, never names a bet to back and never claims a price is value: it is there to show our reasoning, not to tip. On a quiet race it will say plainly that nothing stands out.

Inside a Race Lab race the tabs are: Racecard (the one runner table, our old Card and Verdict tabs are now merged: each runner shows the Verdict score, recent form, the connections and weight, our strike-rate read on its record at the course, distance and going, plus price; tap a horse's name for its full profile, or tap the row to drop down its detail (on a phone this opens a Form / Record / Read switch with the model's value edge, model win probability vs the market, up top); the Advanced toggle by the tabs reveals five strike-rate donuts per runner (hidden by default for a cleaner card); each donut is the horse's lifetime strike rate at that condition, a greyed donut with a dot means its most recent run under that condition was 3+ years ago (a stale signal, with the date in the tooltip), and fewer than 3 runs shows the raw wins/runs instead of a percentage), Edge (the per-runner Ratings Board for that race, see below), Velocity (pace map with a one-line shape read), Draw (stall bias, Flat races only, so it's hidden on jumps cards) and Results (finish order, SP and the in-running comment shown under each runner). Every tab leads with the plain-English answer first, then the detail beneath.

On the Racecard you may see a small teal or red underline under a runner's number: teal means the horse is drawn on the favoured side, red the wrong side. It only appears on Flat races where the course shows a strong draw bias for that distance and field size, so most runners carry no dot at all. Hover it for the reason. It is context only, never folded into the score, because the market already prices the draw.

The Edge tab is the heart of the new model. It scores every runner in the race on A/E across seven angles (trainer, jockey, trainer×jockey, trainer at the course, sire, damsire and first-time-out), each on 12 months of licensed data. A/E above 1.00 beats the market; deeper green is a bigger edge, and the small number beside each figure is the sample size. The Green angles column counts how many of the seven beat the market, so you can see at a glance which runners the data likes. The old Patterns tab (10-year fixture winner profile) was retired in the revamp to make room for it.

The day selector in Race Lab spans a rolling window: the last 7 days as results to study, plus the days ahead as advanced racecards (fully scored once declarations are in).

On a wide screen Race Lab uses a two-column layout built around the card. The day's meeting nav sits in a strip across the top: one group per meeting, a button for each race time, and a small dot on each coloured by that race's best Verdict (green strong to red weak). A Today / Tomorrow switch sits at the left of the strip, and clicking any time opens that race; the page opens on the day's first race automatically. The racecard itself is the wide column on the left, so the runners are the first thing you see. A slim companion rail on the right holds this race's Overlay's read at the top (moved out of the card so it no longer pushes the runners down), then the live Standouts today panels that read across every card at once: Top rated of the day (our single highest-rated runner), Top rated today (the highest Verdict scores anywhere on the card) Biggest overlays (runners we rate well above the bigger price the market gives them) and Market movers (today's biggest steamers, the runners whose odds the market has cut the most since it opened, shown as factual market activity, not our picks: backing steamers is poor value on our record). Click any name to jump straight to that race. A Hide panel toggle on the right of the top strip collapses the companion rail so the card takes the full width, and remembers your choice. The Race Filter above the card narrows everything (the strip and the rail included) by number of runners, race code (Flat, Hurdle, Chase, AW, NHF), handicap or non-handicap, country and class. On a phone Race Lab opens on the Glance home instead: the day's races as a single scannable list, grouped Clear form edge, Tight then Wide open by how clearly our numbers separate the field (a clear form edge means one runner is well ahead on our figures, not that it is a confident bet), each race led by the pick with its tier chip and score (Standout / Top rated / Shortlist / Watch), a plain-English line, the price and a green-to-red bar of the field's shape (a one-line field-strength key sits at the top of the list). A By rating / By time toggle switches the list between that rating order and a plain chronological order. Inside a race, each runner row carries its tier and calibrated score alongside the field-position ladder, with Track and Bet on every runner. The Fast results button sits as a pill at the bottom-left, clear of the cards. Tap a race to open it (with prev/next arrows to move through the card), then use "All races" to come back. Tapping a runner opens a clean detail that leads with the read and shows only the facts and strike-rates that exist, rather than a wall of empty figures. The "Calibrated, not a tipster" chip on the home opens the Performance page, where you can check how the model's confidence has held up. If you ever prefer the old single-column layout, add ?ui=old to the address.

The speed check on Velocity: the pace verdict tells you the race shape (who leads, who closes, whether the front end overheats). Under it we add one line crossing that shape with each runner's speed figure (our own par-adjusted figure, or our data provider's speed rating, where available) to answer the question shape alone cannot: is the horse the pace favours actually fast enough to use it? A green line means the pace and the speed point the same way (for example a lone leader who is also the fastest); an amber line is a caution (for example a lone leader who is outpaced on the clock by horses sitting off him). It is a reading aid only and does not change any score.

☆ My stable (free): tap the star on any horse to follow it; you'll get one email at 20:00 BST on a day a stabled horse runs or is declared.

👛 Wallet: tap ＋ Bet on any runner and a slip opens pre-filled with the horse, price and race time. We auto-settle singles from our own results (win / place / lose / void) using proper bookmaker rules: each-way place terms set by field size and race type (handicap or not), non-runners voided with your stake returned, dead-heats paid on a reduced stake, and a Rule 4 deduction applied when a rival is withdrawn. We keep a running P&L, ROI and strike-rate, with Bets and Summary views. Your Wallet sits at the top: a play-money virtual bankroll (no real cash, ever) showing your balance, total deposited, what is riding on open bets and your profit or loss. Use the Add funds chips to top up any time: a top-up adds to your balance and keeps your running P&L counting, while Reset starts a fresh balance instead. When a bankroll is set, the bet slip shows your current balance and what would be left after the stake you've entered (it turns red if a stake is more than you hold), so you always know how much you have to play with. Betting closes once a race is off, tap ＋ Bet on a race that's underway and you'll get a "race underway" notice rather than a slip. The floating My Overlay button (bottom-right, on every page) opens your Slip, Wallet, My Stable and the yards & riders you follow in one tap; the slip shows the day's bets and clears settled ones after 21:30, and everything stays in your Wallet. You can place a bet from anywhere, not just the Race Lab: the ＋ Add a bet button at the foot of the slip opens the same bet slip on any page, and you can add an off-card bet by typing the horse, course and date yourself. To build a multiple, just tap a runner's odds on any card (homepage, racecards, US cards, or the 🎲 Add to multiple button in the Race Lab bet slip). A bet slip bar appears at the foot of the screen and auto-builds the accumulator as you add legs (Double, Treble, N-Fold acca), showing the combined price and "£10 pays £X". Add an Each way if you want it, type a stake and tap Place Bet. Tap Show all selections to see every leg and remove any with the ×, or tap a selected price again on the card to drop it. One selection in the bar places a single; two or more place the acca. It settles leg-by-leg from our results: all legs must win (a non-runner leg drops to odds 1.00), and each-way accas pay on placed legs. For the full permutation bets (Trixie, Patent, Lucky 15 / 31 / 63) use the Bet calculator below. You can cancel a still-pending bet entered by mistake, but a settled bet cannot be removed, so the record stays honest. To carry your bets and bankroll to another device, sign in from Your account (the person icon, top right): once signed in with a passkey (Face ID / Touch ID) or an emailed code, they follow you automatically on every device.

👤 Your account (the person icon, top right): an optional, free sign-in that saves and syncs your stable, slip and virtual bankroll across your devices. Signing in is passwordless: tap Continue with a passkey (Face ID / Touch ID) or have a 6-digit code emailed to you. The rest of the site stays exactly as it is whether you sign in or not, and there is no password, no card and no tipping list: we store only your email and your choices. Once signed in, the same panel gives you one-tap access to your Stable, Wallet, open slip and odds format, and a Sign out. The slip is always play money, never real cash, and Stable alerts tell you when a horse runs, not what to back.

🧮 Bet calculator: work out the return on any multiple: Double, Treble, accumulators, Trixie, Patent, Lucky 15 / 31 / 63. Enter prices as fractions, decimals or "evs", toggle each leg Win / Plc / Lose, add each-way and lucky bonuses. Prefill from today's picks loads the model's selections. Add to slip saves the whole multiple to your tracker; when its legs have run, Settle suggests the return by resolving each leg against our results (adjust it if your bookmaker's Rule 4 or terms differ).

Angles live in two places, on purpose. Today's Angles (the Angles tab) is the simple, act-now surface: tap an angle to see today's qualifiers and back them. Build your Angle (this tab) is the research tool: build a custom angle over 1.9M runs, see its honest record, then send it to Today's Angles or save it.

Today's Angles. A grid of angle tiles: our ready-made angles (mid-price value, outsiders, favourites, sprinters, big-field handicaps, top-rated), popular punter angles (last-time-out winners, in-form trainers and jockeys, course specialists, droppers in class), and your own saved angles. Tap a tile to see today's qualifiers with their tier and price: tap any price to back a single, or build the qualifiers into an acca (capped at a sensible size) through the shared bet slip. Each of our filter angles also shows its honest forward record.

The honest record. Measured on our live, pre-race scores settled at Betfair SP (the leak-free forward record, the only place an accuracy claim means anything), never the full back-history. It shows the sample, win rate, A/E vs BSP (below 1.00 means backing the angle blind loses to the market: no edge), level-stakes ROI, rivals-beaten (PRB) and a calibration line. This is information and calibration, not tips, and not a betting edge: strike rate is never the scoreboard, because you can back favourites, pick the most winners, and still lose at the prices.

Build your Angle. Build from the full filters (course, class, score, price, going, distance, field size, race type, age, rating, horse / trainer / jockey). You get the honest forward record plus the all-history context: the same filters over the full 1.9M-run archive settled at starting price, with the matching runs, Group by (Trainer, Jockey, Sire, Course, Class, Going) and export. The history is colour on how a profile has behaved, not our forward edge. Then send it to Today's Angles to see today's qualifiers and back them, or save it to your account (sign in to keep it, it syncs across your devices and appears on Today's Angles). The page address updates as you query, so you can bookmark or share any angle.

Ready-made report tabs (🔥 Hot trainers, 🏇 Hot jockeys, 🧬 Top sires, 📍 Course specialists) are time-bound: 14 days by default, up to a year. Short windows show who's hot now; long windows show who beats the market (A·E over 1.00). Click any row to jump to that trainer or jockey's declared runners today, ranked by our score with ＋ Bet and ☆ right there.

Performance is one honest scoreboard, in observation mode: nothing is staked and we never advise a bet. We track every runner we rate and every value call we publish, then show the results in full, win or lose. It opens with a plain-English "what the numbers mean" key (Win%, Placed, A/E, ROI, BSP) so you can read it cold. The sample is young, so treat it as a work in progress.

At the very top of the Performance tab is the calibration strip: it groups every runner we rate into its Verdict tier (Standout, Top rated, Shortlist, Watch, Pass) and shows how often each tier has actually won and placed, over a rolling sample of the last several days. It is the honest "do our ratings mean anything?" read: higher tiers win more often (Standouts and Top-rated horses win roughly one in five, down in steady steps to about one in fourteen for Pass). It measures accuracy, not profit: a high win rate is not a betting edge, because short-priced favourites win often but pay little. It refreshes daily.

Just below it, the "Our top-rated runner, in every race" card takes our single highest-rated runner in every settled race, win or lose. Because we rate on our own figures rather than the price, our top pick is often not the favourite, so it wins fewer races outright than the favourite does (shown side by side for comparison), while still placing in a good share of them. That is deliberately honest: it shows we are not just following the market, and it makes the mission's core point, that backing either our pick or the favourite blind still loses at the prices, because the favourite's higher strike rate is paid for in short odds. Winners are not profit. Forward and leak-free, settled at Betfair SP, refreshed daily.

Each tier also carries a "vs the market" line, how that tier has done against the Betfair exchange price (A/E above 1.00 = ahead of the market so far). This sits under a Learning mode in progress badge on purpose: it is built from only a few days of data, far too little to prove an edge, and our wider research has found no reliable edge against Betfair SP. We show it openly as the sample grows rather than hide it, but treat anything above the market as "not yet disproven", not a confirmed edge.

Below the tiers, "How our published value calls do against the exchange" is a single summary line: how many calls have settled, the strike rate, the A/E versus Betfair SP and the level-stakes ROI after commission. Judged at Betfair SP (which has no edge by design, the exchange price already reflects all of this) the honest read is no reliable edge, so we show the real numbers rather than a sales pitch.

The banner at the top of the dashboard is the model's forward record: the live, running results of every selection the current model has made since it went live, settled against the real result. It shows the start date, how many picks have settled, the strike rate, the A/E (above 1.00 is beating the market) and the level-stakes ROI. It is in observation mode (tracked, not staked) while the sample builds, so treat early numbers as indicative: a few weeks of racing is not yet statistically meaningful, and the backtest expectation is an A/E around 1.10. Prices are shown at SP until exchange BSP capture is restored.

The model's logic is built on data we own outright: 1.9 million settled UK & Irish results, 2010 to 2026. On top of that we use the official handicap ratings and realised Betfair SP on every runner, and build our own par-adjusted speed figures in-house from race times and beaten lengths, alongside a performance rating and a speed rating for each runner from our data provider where available.

• Official plus home-grown data: official handicap ratings and realised Betfair SP so we track true edge, plus our own par-adjusted speed figures, a provider performance and speed rating per runner where available, and run-style reads built from 1.9M results. Every feed is monitored daily and self-heals the day it breaks.
• Home-grown class rating: a skill rating learned purely from real results; rewards proven class, marks down horses running above their level.
• Field-size honesty: a top-rated horse wins far less often in a big field than a small one (the market prices this; our raw score did not), so our rating is marked down in large fields, heavily in 16-plus runner races. It keeps a score of 70 meaning the same win chance whether it is a 5-runner or a 20-runner race.
• Honest win-chance: the score-to-win-chance table is fitted on all 1.9M results, so the Edge figure on every pick is properly grounded.
• Stakes saved for races worth backing: Class 5-and-below handicaps have been long-term losers, so those runners are shown as research only.
• A NAP that earns it: the headline NAP clears the same value test as every pick; no short-priced no-value favourites.

Why the rules look the way they do. The rating runs on a calibrated scale re-based in September 2026 where 50 is an average runner: Standout (65+), Top Rated (60+), Shortlist (54 to 59) and Watch (50 to 53). The number is the model's honest read of a runner's strength, spread so the genuinely strong stand out while the pack sits near average; the within-race order is unchanged. The old VALUE and COURSE DEBUTANT badges were retired after testing on their own population (sub-70 runners): neither beat the field, so a low-scoring horse now reaches the board only on an angle the numbers support. Sit-outs and the market-backed-favourite rule exist because chasing weak races and opposing well-backed last-time-out winners both lose money across tens of thousands of settled bets. The two ROI figures are both correct: they just answer "did it pick winners?" versus "did the advised staking pay?".

The principle behind all of it: back value, not opinion, using data we own outright.