Methodology & Track Record

How we test strategies before trusting them, what the testing found, and the live results for every strategy currently running.

How we validate a strategy

Every strategy shown below started as a hypothesis tested against historical results, not a rule we simply assumed would work. A large part of that testing is spent trying to find reasons a promising-looking result might be wrong, before ever trusting it.

One concrete example: when we checked whether our model could improve on the market's own prices, we wrote the pass mark down first, fitted the combination on earlier races, and judged it only on later races it had never seen (36,386 races to fit, 16,439 to test). It did not pass. That result is why this site presents the model as an honest view of each race rather than a source of winning bets.

Why we don't just keep looking until something looks good

Test enough different rules against the same historical data and some will look profitable purely by chance, not because they reflect a genuine pattern. We treat a promising backtest as a starting hypothesis, not a conclusion. Before we rely on it, it gets tracked live -- on real races it was never tested against -- and its results are reported here exactly as they happen, whether they confirm the backtest or not.

What our testing found

In October 2026 we tested the model as hard as we could against the betting market, writing down what would count as a pass before each test. We checked it against the price at the off and the early price; whether the market later moves towards its view; a fresh model adding form to the starting price over 16,000 races it had never seen; each-way betting at SP; and laying at Betfair SP. None showed an edge over the market.

So we describe the model honestly: its chances are well calibrated (a 10% chance wins about 10% of the time) and broadly in line with the market. It is a useful second opinion on a race, not a betting system. Where the model and the market differ we show it as "Model higher" or "Model lower", which is a difference of opinion, not a sign of value.

Live track record

These are the actual, ongoing results for every strategy currently being tracked live. Nothing here is hand-picked or edited after the fact: results are added automatically as each bet settles, apart from the dated corrections listed under Corrections below, which applied one rule to every affected bet.

A strategy being profitable so far does not prove it has a lasting edge, and a well-calibrated model does not mean it can beat market prices. Read the results below with both in mind.
Strategy Settled W / P / L Staked Profit ROI
Value Betting (EV > 0) 496 120 / 25 / 351 496.00 -9.22 -1.9%
Disagreement Strategy (nh_all) 151 33 / 0 / 118 151.00 +0.92 +0.6%

When the prices are taken. Since 18 October 2026 the full daily run happens at 07:30 and every pick is logged at the prices available then. Before that date the daily run happened at 11:10, so results before 18 October are not directly comparable with those after it. A lighter job at 11:10 only removes withdrawn horses; it does not change any price or prediction. Prices shown as “now” on the racecards and Track record are live and are for information only. Bets are still settled on the price logged at 07:30 (or the starting price, if higher), never the “now” price.

Gap in the Disagreement Strategy (nh_all) record. The strategy was paused from 15 October 2026 and no new bets were logged between 15 and 19 October. It has since been switched back on. Bets logged before the pause are unchanged.

Read these numbers with the sample size in mind. A strategy with a few dozen settled bets can look very different a few hundred bets from now, in either direction. These figures are reported as-is, at whatever size they currently stand -- not a claim that the current rate will continue. Past results do not guarantee future performance, on this page or anywhere else.

See every individual bet behind these numbers →

Retired strategies

Flat Handicap Disagreement (EV > 0) backed the model's top-rated horse in a Flat handicap when it disagreed with the market's favourite and the bet had positive expected value. It looked strong on historical races, so we ran it live, logging every pick before the race and settling it automatically. We stopped adding new picks on 15 October 2026 because the early results showed no edge over the market. It is not part of the live track record above; its full record, every bet included, stays on its own page.

See the full record for this strategy →

Corrections

31 October 2026 – each-way terms. Each-way bets in handicaps with 12 or more runners had been settled at 1/5 of the odds for a place instead of the standard 1/4 (and the few 16+ runner non-handicaps to four places instead of three). Every affected settled bet has been re-settled at the standard terms (Value Betting: 13 bets, a net change of +1.51 points). From this date the strategies also use the correct terms when working out whether an each-way bet is value, so slightly more each-way bets in big handicaps can qualify than before.

31 October 2026 – short fields. An each-way bet whose race ends up with fewer than 5 runners after non-runners is now settled as a win bet on the whole stake, as most bookmakers do. It used to be voided. Past bets voided this way have been re-settled the same way.

New to the racecards? Reading a racecard explains the model's figures, the stats line under each horse and the extra sections under each race.