Every betting system, tested
Every strategy in this library has been run over real British races and settled the same honest way: industry Starting Price, fallers and pulled-up horses counted as the losers they are. Each card shows who backs the idea, what it is trying to exploit, and what it actually does: search for the ones you have heard about, or browse by kind and follow your curiosity. The full working sits one click in.
24 systems tested · 27,676 races · every result published
Win or lose, the record is the record. Under test this week: Four-fold on favourites (−30.8p per £1 staked).
New to odds? What a price really means. Or read the AI Blog.
Staking systems (2)
Backing & laying strategies (16)
Bet types (7)
Meeting-specific (2)
Return and strike are measured on real races, settled at industry Starting Price. Effort, popularity and swings are our editorial scores from one to five (swings is computed from the £200-a-week money path where one exists). Every card opens the full experiment with the sample, the error bars and the method.
The Silicon Tipster League →
ChatGPT, Gemini, Grok, Claude and DeepSeek tip every race, settled at SP: one test blind, one with the market shown. Watch five AIs prove themselves live, from scratch.
The kinds of system we test, and how
Five families, five different theories about where an edge might hide. Open any one for what it is trying to exploit and how we put it to the test.
Staking systems
Change the stake, not the horse. Martingale, Fibonacci and friends aim to exploit sequencing: the idea that losing runs must end, so escalating stakes claw everything back.
How we test it: We simulate the actual bankroll on real results, bet by bet, because averages hide the thing that matters: the bust.
Backing & laying strategies
Fixed selection rules, back the favourite, the outsider, the top-rated. Each one bets that some corner of the form book is underpriced by the crowd.
How we test it: Every qualifying race in the record, £1 at Starting Price, fallers and pulled-up horses settled as losses. No cherry-picked seasons.
Market mechanics
Quirks of how prices form: the draw, running styles, how form moves a price. These hunt structural edges rather than opinions about horses.
How we test it: We measure whether the pattern is real first, then the only question that pays: does the price already know?
Bet types
The same horses, different wrappers, each-way, doubles, accumulators, the Lucky 15. The wrapper changes how often you collect and what the bookmaker's margin does to you.
How we test it: Settled from real single-leg results; multiples are computed as analytic expected value, because stacking legs multiplies the margin on every one.
Meeting-specific
Favourite habits pointed at one meeting: the festivals, Royal Ascot. The theory is that the deepest, most-studied races make form more reliable.
How we test it: Same settlement as everything else, on a smaller sample, so we publish the error bars and say so when the noise outweighs the signal.
Everything settles the same way, and the full conventions are on How We Test.
The big questions, answered in full
Not strategies to follow: the comparative reads that make sense of the whole board. Start here if you want the wide view before picking a card above.
More comparative write-ups land on the AI Blog.
Questions about the library
What counts as a betting system?
Any fixed rule that decides what to back or how to stake, applied to every qualifying race with no judgement calls. The library covers selection rules (back the favourite, the top-rated), staking plans (Martingale, Fibonacci), bet types (each-way, accumulators, the Lucky 15) and market-mechanics ideas like the draw.
Where do the return and strike figures come from?
Real British races, settled at industry Starting Price with fallers and pulled-up horses counted as losses and joint favourites split, 27,676 races and counting. The method is published in full on our How We Test page, and every card opens the sample and error bars behind its numbers.
How do you score effort, popularity and swings?
Those three are our editorial opinion on a one-to-five scale: how hard a system is to stick to in real life, how widely it is actually played, and how violent its bankroll ride is. Swings is computed from the £200-a-week money path where we track one. Return and strike are never editorial, they are measured.
Where can I see what these systems cost in real money?
The What Betting Costs page runs the same £200 a week through every system and shows the leak per week, month and year, with the full ranked table and the week-by-week chart. It is the money view of this library.
What would £200 a week on these actually cost? →
The same weekly stake through every system: the leak per week, month and year, the full ranked table, and how to shrink it.
This is research, not tips. The numbers describe what each system has done over a large, honest sample; they are not a signal to stake. The model is a calibrated read on a race and does not beat the market. 18+, please bet responsibly.