The Thursday Take is our weekly column: one racing talking point, put to the test against the Proform database — every British and Irish runner, every result. This week, with the Charlie Hall Chase on Saturday and the November meeting at Cheltenham a fortnight away, the line you will read in every preview: "won this race last year".
The short version
- On the Flat, it means nothing. Last year's winner back for the same race wins as often as its Betfair price says (A/E 0.989).
- Over jumps, it costs you. Last year's winner back returns 0.927, a 15.7% loss at Betfair SP. Horses that were placed last year are the only returners above par (1.031).
- Not in the big races. Class 1 and 2 returners are priced right (1.009). The shortfall is in ordinary handicaps, with the horse a few pounds higher than when it won.
- The test that matters: a jumps handicapper 1-10lb above the mark it won off a year ago wins 18% less often than its price says when it comes back to the same course and trip (0.822, under par in 7 of 8 years). Running anywhere else, the same kind of horse is priced fine.
- The lesson: the market is not overrating the horse. It is overrating the memory.
This week's talking point
Every big-race preview has a line for it. "Won this last year." "Back to defend his crown." "Loves it round here." It is the most comfortable reason there is to back a horse, because you have already watched it win. On Saturday it will be said about horses at Wetherby and Down Royal, and in a fortnight about every returning handicapper at Cheltenham. So does last year's winner win again as often as its price says?
How to read the numbers. A/E (actual over expected) is winners divided by the winners the Betfair SPs said to expect. 1.00 means the prices were exactly right; 0.82 means 18% fewer winners than the prices implied. Every runner counts, fallers and pulled-up horses included. "Back for the same race" means the horse ran at the same course, over the same trip, in the same type of race and with the same handicap status 11 to 13 months earlier. British and Irish racing, 2019 to 26 October 2026.
Last year's winner, last year's placed horses
| Back for the same race, after | Runs | Wins | A/E | Return at Betfair SP |
|---|---|---|---|---|
| Flat: winning it last year | 6,776 | 871 | 0.989 | −3.1% |
| Flat: placed 2nd or 3rd | 9,475 | 1,004 | 0.964 | −3.1% |
| Flat: anything else | 19,001 | 1,527 | 0.950 | −4.8% |
| Jumps: winning it last year | 2,803 | 420 | 0.927 | −15.7% |
| Jumps: placed 2nd or 3rd | 3,855 | 508 | 1.031 | −5.2% |
| Jumps: anything else | 5,653 | 449 | 0.940 | −3.0% |
Every runner in the period gives an A/E of 1.000 on the Flat and 1.000 over jumps, so almost every group of returners sits a little under par: punters like a horse they have seen at the track before. On the Flat the effect is small and it does not depend on what happened last year. Over jumps, last year's winner is the one the market overrates most, and last year's placed horse is the only returner priced generously. On its own the 0.927 is about one and a half standard errors from 1.00, the range chance alone covers, so we went looking for where it sits.
Where the shortfall hides
| Last year's jumps winner, back | Runs | A/E | Return |
|---|---|---|---|
| Class 1 and 2 races | 934 | 1.009 | −13.8% |
| Class 3 and below | 1,869 | 0.875 | −16.6% |
| Non-handicaps | 487 | 1.021 | −6.9% |
| Handicaps, same mark or lower | 663 | 0.933 | −12.6% |
| Handicaps, 1-5lb higher | 726 | 0.828 | −34.2% |
| Handicaps, 6-10lb higher | 551 | 0.842 | −16.5% |
| Handicaps, 11lb or more higher | 376 | 0.972 | +4.2% |
| Starting at under 2/1 | 483 | 1.000 | −2.8% |
| Starting at 5/1 or bigger | 1,577 | 0.799 | −23.9% |
The big races are priced right: the Class 1 and 2 returner, and the short-priced one, win exactly as often as they should. The gap is in the ordinary handicaps, with a horse a few pounds above the mark it won off, sent off at 5/1 or more. That is the horse the preview writer cannot resist, and the one the handicapper has not yet caught.
The real test: is it the course, or the weights?
There is an obvious objection. A horse a few pounds higher than when it last won is meant to find it harder: that is what the handicapper is for. Maybe the market is simply too slow to price a rise in the weights, wherever the horse runs. So we took every jumps handicapper whose last handicap win came 11 to 13 months ago, running 1 to 10lb above that winning mark, and split them by where they turned up.
| Jumps handicapper, 1-10lb above the mark it won off a year ago | Runs | Wins | A/E | Return |
|---|---|---|---|---|
| Back at the same course and trip | 1,328 | 147 | 0.822 | −25.9% |
| Running anywhere else | 5,880 | 729 | 1.084 | +2.6% |
| Every other jumps handicap runner | 194,600 | 20,582 | 0.998 | −1.1% |
Same kind of horse, same rise in the weights. Running somewhere new, it is priced about right — if anything a touch long. Running back at "its" track and trip, it wins 18% less often than its price says, and it has been under par in 7 of the last 8 years (0.72, 0.93, 0.79, 0.83, 0.69, 0.84, 1.13, 0.57). The rise in the weights is not what the market gets wrong. The memory is.
On the Flat the same test shows nothing: 0.962 back at the same course, 0.955 elsewhere.
How sure are we? The same-course figure is about 2.4 standard errors under 1.00 and the most consistent thing in this piece, but it is a contrast we chose after looking, with two cuts tried, not a test we wrote down in advance. The "anywhere else" figure of 1.084 is one cut at about two standard errors: do not read it as an edge. The honest reading is that a returning handicap winner is a reason to look twice at the price, not a system to lay.
The two races of the next fortnight
The Charlie Hall Chase (Wetherby, Saturday). Defending champions are 0 from 10 since 2005: Ollie Magern and Bravemansgame second, Cue Card third at 4/5 in 2016, three that did not finish. The last horse to win it two years running was See More Business in 2000. This year that run will not be tested: last year's winner, Djelo, is not among the four-day entries. And a fair warning: the Charlie Hall is a Grade 2 non-handicap, exactly the kind of race where returners are priced right as a group. Ten runs is colour, not evidence.
The November Gold Cup (Cheltenham, 14 November). Defending champions are 0 from 5 since 2004, the best of them Johns Spirit, second at 12/1 in 2014. This one is the shape the test above describes: a handicap winner coming back a year later, a few pounds higher. If last year's winner, Panic Attack, lines up, that is the horse to look at hardest against its price.
Build it yourself
The System Builder gets close to this with five rules. It cannot say "the course of its last win", so it uses course and distance wins instead, which also counts wins from years ago.
- Race Type is one of Chase, Hurdle, NHF, and Handicap = Yes.
- Days Since Last Win between 340 and 390.
- OR Above Best Winning OR between 1 and 10 (today's mark, 1-10lb above the best mark it has ever won off).
- Then run it twice: Course & Distance Wins of 1 or more, and again with Course & Distance Wins = 0.
| System Builder, 2019 to 27 October 2026 | Runs | Wins | A/E | Return |
|---|---|---|---|---|
| With a course and distance win | 1,238 | 141 | 0.92 | −3.8% |
| Without one | 3,220 | 350 | 1.02 | −3.9% |
The Builder's version shows the same gap in A/E, and the course-and-distance winners are under par in 6 of the 8 years in its Year breakdown. But be clear what it does not show: the return at Betfair SP is the same, −3.8% either way, because a handful of big-priced winners swing the profit in a set this size. A/E is the steadier measure, which is why we lead on it, and it is also why this is a lens for your own selections rather than a system to bet blind.
The take
"Won this last year" is the most comfortable line in racing, and on the Flat and in the big jumps races it does no harm: the market prices those returners right. But in an ordinary jumps handicap, a horse back for the race it won, a few pounds higher, wins about 18% less often than its price says, year after year, while the same kind of horse running somewhere new is priced fine. The handicapper's rise is not the problem. The memory is. If you fancy last year's winner, ask whether you fancy the horse or the replay.
Last week we looked at whether the rising-star jockey is worth following. Next week: horses that move from Britain to Ireland. Is there a second race in them?
Questions we get asked about this
Do horses that won a race last year win it again?
Less often than their prices say over jumps. Last year's jumps winner back for the same race returned an A/E of 0.927 at Betfair SP (2,803 runs, 2019 to October 2026), a 15.7% loss. On the Flat it was 0.989, close to fair.
Is it worth backing last year's winner in a handicap?
Over jumps, usually not. Back at the same course and trip, 1 to 10lb above the mark it won off, it returned 0.822 (1,328 runs, −25.9% at Betfair SP), under par in 7 of 8 years. The same kind of horse running elsewhere was priced about right.
Are big-race defending champions overbet?
Not as a group. Returners in Class 1 and 2 races won as often as their prices said (A/E 1.009). Individual races can have long losing runs, such as the Charlie Hall Chase (0 from 10 since 2005), but ten runs is too few to read anything into.
Is a course winner a good bet?
The market already pays for it. Course form is real, but punters like it so much that the price overshoots, most clearly in ordinary jumps handicaps.
Can I test this in the System Builder?
Yes. Race Type (Chase, Hurdle, NHF), Handicap = Yes, Days Since Last Win 340-390 and OR Above Best Winning OR 1-10, then compare Course & Distance Wins 1 or more against 0: A/E 0.92 against 1.02 since 2019.
