The Thursday Take is our weekly column: one racing talking point, put to the test against the Proform database — 20+ years of runs, every price, every result. Including when the popular story doesn't survive contact with the numbers. This week it mostly does — and the interesting part is what sits underneath it.
The short version
- The published Ebor trends are all true. Aged 4–6, ran within 66 days, three or more runs this season, won over the trip — each one picks out winners a little better than it picks out the field, exactly as a trend should.
- And none of them is an edge. Stack all four and the strike rate climbs from 4.95% to 6.92% — about 40% better — but the A/E is 1.007 against a 0.998 baseline. You are paying a fair price for a better horse.
- “Ran within 66 days” describes 85.1% of the entire field. So a trends piece telling you ten of the last twelve winners had run that recently — 83% — is reporting the base rate, not beating it.
- The value is hiding inside that one tick-box. Horses back after 14–27 days are the biggest group in the race (39.3%) and the worst priced, at A/E 0.933. Horses back after 42–66 days are only 8.8% of runners and the best priced, at 1.097. Both ends hold in each half of the data.
- On the Ebor itself: 8 of the last 17 winners — and 6 of the last 8 — came back off a 42–66 day break. The winner is not the horse that ran a fortnight ago; it is the horse being aimed at the race.
- Build it: Field Size 14–40, Handicap = Yes, Course Country = GB, Age 4–6, Days Since Last Run 42–66.
This week's talking point: the Ebor trends
It is Ebor week at York, and on Saturday the £500,000 Sky Bet Ebor — Europe's richest Flat handicap — goes off with twenty-odd runners on the Knavesmire. Which means the trends pieces are out, and if you have read three of them you will have met the same handful of lines:
- The winner is almost always aged four to six.
- It has run within the last 66 days.
- It has had three or more runs already this season.
- It has won over the trip before.
Those are checkable claims about a knowable population, which makes them exactly the sort of thing this column exists for. So we tested all four properly — not against a dozen Ebors, but against every big-field Flat handicap in Britain since 2018.
The headline: they are all true, and none of them will make you a penny. That is not a criticism — they are sold as descriptions of a winner, not as betting systems, and as descriptions they are accurate. But underneath one of them is the widest value gap in the race, and nobody quotes it.
The cohort
- GB Flat handicaps (turf and all-weather), 14 or more declared runners, open to 3yo+, resulted, 1 January 2018 to 17 August 2026.
- 2,880 races. 45,467 runners. 2,887 winners — a 6.35% strike rate.
- 99.9% have a Betfair Starting Price, so we can ask not just "did it win?" but "were you paid properly?"
- Baseline A/E: 0.998. As a whole, this cohort is priced almost perfectly.
Two terms, in plain English.
A/E — add up what the betting market thought each horse's chance was, then compare it with how many actually won. 1.00 means the market had it exactly right. Below 1.00 means you were paying over the odds; above means the group won more than its prices implied. It is a fairness score, not a profit.
Impact value — what share of the runners does this describe, against what share of the winners it produced. 1.00 means it picks out winners no better than it picks out the field.
The four trends, tested
| Trend | Runners | % of field | % of winners | Impact value | Strike rate | A/E |
|---|---|---|---|---|---|---|
| Aged 4–6 | 26,062 | 57.4% | 60.6% | 1.06 | 6.70% | 1.006 |
| Ran within 66 days | 38,632 | 85.1% | 87.7% | 1.03 | 6.55% | 0.987 |
| 3+ runs this season | 29,801 | 65.6% | 66.8% | 1.02 | 6.46% | 0.989 |
| Won over the trip before | 24,127 | 53.1% | 52.3% | 0.98 | 6.25% | 0.975 |
Every single one lands within a whisker of its own base rate — impact values of 1.06, 1.03, 1.02 and 0.98 — and every one is priced at or a shade below par. The age trend is the pick of them, and even that is 1.006: the market has it exactly right.
Look hardest at the second row. "Ran within 66 days" describes 85.1% of every runner in the cohort. So when a trends piece tells you ten of the last twelve winners had run within 66 days — that is 83% — it is reporting the base rate almost to the decimal point.
Base rate, in plain English. If 85 runners in every 100 tick a box, then about 85 winners in every 100 will tick it too — automatically, whether or not the box means anything. Any trend that describes most of the field will look like it "found" most of the winners. Always ask what share of the runners qualified before you are impressed by the share of the winners.
And stacking them does work — as advertised
We scored all 45,467 runners on how many of the four they satisfy:
| Boxes ticked | Runners | Wins | Strike rate | A/E | ROI at Betfair SP |
|---|---|---|---|---|---|
| 0 | 1,112 | 55 | 4.95% | 1.097 | −2.65% |
| 1 | 5,276 | 299 | 5.67% | 1.070 | +0.05% |
| 2 | 13,228 | 836 | 6.32% | 1.013 | +4.14% |
| 3 | 16,326 | 1,038 | 6.36% | 0.956 | −10.49% |
| All four | 9,478 | 656 | 6.92% | 1.007 | +1.20% |
The strike rate climbs the whole way: 4.95% with none of them to 6.92% with all four, a 40% improvement. The trends genuinely do describe a better horse, and the more of them you stack the better that horse gets.
And the A/E at the top is 1.007, against a baseline of 0.998. Which is the honest verdict on the whole exercise: you are paying a fair price for a better horse. No edge, no con, nothing to expose. The trends work exactly as they say on the tin — and the market has read the same trends you have.
(The 3-box row at 0.956 is the one wrinkle. It is below par in both halves of the data, but there is no orderly story around it and we are not going to invent one.)
Now the bit nobody quotes
If "ran within 66 days" covers 85% of the field, the interesting question is not whether a horse is inside that window. It is where inside it. So we cut the layoff finer:
| Days since last run | Runners | % of field | Strike rate | A/E | 2018–22 | 2023–26 |
|---|---|---|---|---|---|---|
| 0–13 | 9,515 | 21.0% | 7.49% | 1.005 | 1.007 | 0.999 |
| 14–27 | 17,807 | 39.3% | 6.04% | 0.933 | 0.917 | 0.964 |
| 28–41 | 7,306 | 16.1% | 6.52% | 1.037 | 1.015 | 1.071 |
| 42–66 | 4,004 | 8.8% | 6.62% | 1.097 | 1.144 | 1.032 |
| 120+ | 4,835 | 10.7% | 5.46% | 1.092 | 1.119 | 1.051 |
There it is. The single tick-box "ran within 66 days" contains both the worst-priced group of runners in the race and the best.
- 14 to 27 days: A/E 0.933. The biggest single group in the cohort — 39.3% of all runners — and the worst value in it. Negative in both eras.
- 42 to 66 days: A/E 1.097. Only 8.8% of runners. Positive in both eras.
That is a spread of 0.164 in value, hiding inside a criterion that every trends piece treats as a single yes-or-no.
And the reason is not mysterious. The horse that ran a fortnight ago is the tidiest profile on the page: recent form figures, nothing to dislike, no questions to answer. Nothing to dislike is exactly the problem — everybody likes it, so everybody backs it. The horse having its first run in six or eight weeks looks slightly rusty and slightly unproven, and gets priced accordingly.
(One band we are deliberately not selling you: 67–119 days looks superb on total return, but it swings from 1.233 in the first half of the data to 0.685 in the second on fewer than 2,000 runners. That is one or two big winners, not a pattern. We would rather flag it than dress it up.)
Which brings us to Saturday
Our database holds the Ebor runner by runner back to 1997, with Betfair prices from 2009 — so seventeen runnings we can actually test. Here is how the published trends have done on the race they were written for:
| Test | Ebor winners, 2009–2025 |
|---|---|
| Aged 4–6 | 16 of 17 |
| Ran within 66 days | 14 of 17 |
| Official rating 101–108 | 11 of 17 |
| Back off 14–27 days (worst-priced band) | 4 of 17 |
| Back off 42–66 days (best-priced band) | 8 of 17 — and 6 of the last 8 |
The trends hold. Age four to six is 16 from 17. The 66-day window is 14 from 17. Nobody quoting those at you this week is wrong.
But the 42-to-66-day band is 8.8% of runners in the wider cohort and 47% of Ebor winners — and six of the last eight. The typical Ebor winner is not a horse that ran a fortnight ago. It is a horse having its first run in six to nine weeks, because the Ebor is a race horses are aimed at, and the horse being aimed at it has usually been given a holiday first.
For completeness: across 5,665 runners in 14+ runner handicaps at York since 2018 the A/E is 0.9999. The Knavesmire prices this stuff about as perfectly as a racecourse can. Which is rather the point — if there is an edge to be had there, it is not going to come from a criterion 85% of the field satisfies.
And one cliché that is simply false
| Market position | Runners | Strike rate | A/E |
|---|---|---|---|
| Favourite | 2,897 | 21.2% | 1.003 |
| 2nd favourite | 2,881 | 15.2% | 1.046 |
| 3rd favourite | 2,897 | 10.5% | 0.915 |
| 4th–6th | 8,645 | 7.8% | 0.965 |
| 7th–10th | 11,511 | 4.8% | 1.058 |
| 11th or bigger | 16,589 | 1.8% | 0.979 |
"You can't back favourites in these" is one of the most repeated lines in racing, and it is wrong. Favourites in 14+ runner GB Flat handicaps win 21.2% of the time at an A/E of 1.003 — priced exactly right. The genuinely poor cell is the third favourite, at 0.915.
Three things that looked like findings and died
We publish these too, because a column that only ever reports discoveries is not testing anything.
| Candidate | How it looked at first | What killed it |
|---|---|---|
| "Back the 4-year-olds" | A/E 1.028 on 12,485 runners, +£1,663 profit, ROI +13.3% | Cap winners at Betfair SP 50 and +£1,663 becomes −£549. Twenty-six winners out of 12,485 bets carried the whole thing. A/E 1.028 is the honest number, and it is small. |
| Class droppers (2+ classes easier) | A/E 1.295, ROI +39.9% in 2018–22 | Out of sample: A/E 1.016. The edge more than halved. Same-class and one-class moves show no order at all. |
| Top-weight bands | Top weights A/E 0.881 in 2018–22 — a tidy negative | 2023–26 flips it to 1.112. No pattern across the five weight bands in either direction. Noise. |
Why this column leads with A/E and not "profit to £1 level stakes". The 4-year-old case above is the reason, in one line: a single 80/1 winner rewrites an ROI figure and barely moves an A/E figure. When somebody shows you a back-tested profit, ask how many bets produced it — and what happens if you delete the three biggest winners. If the answer changes the conclusion, it was never a conclusion.
One more thing worth knowing about big fields
The race Saturday's Ebor belongs to is getting rarer. British Flat handicaps with 14 or more runners averaged 423.6 a year across 2015–19; across 2023–25 they averaged 281.3 — down 33.6%. And it is specifically the big one that has gone: the number of GB Flat handicaps has risen over the same period (4,206 to 4,624), and the average field size has barely moved (9.34 to 9.13).
An honesty note about 2026, because part-years lie. To 17 August this year Britain had staged 189 Flat handicaps with 14+ runners. Measured to the same date in earlier years, that is 19% below the 2015–19 average (233) but 10.5% above the 2023–25 average (171). 2026 is a modest bounce inside a much lower plateau — not the collapse year. We would rather tell you that than sell you a cliff.
Build it yourself
Every Thursday Take ends with the recipe, so you can run it — and argue with it — in the RaceMetrics System Builder.
| Filter | Setting |
|---|---|
| Race Type | Turf + A/W |
| Handicap | Yes |
| Field Size | 14 → 40 |
| Race Lower Age Limit | 3 → 20 |
| Course Country | GB |
| Age | 4 → 6 |
| Days Since Last Run | 42 → 66 |
Then open the breakdowns and put Days Since Last Run, Age, SP Rank and Class on screen. The layoff table above is one of those breakdowns, unedited.
Two things our own product cannot do yet — and we would rather say so.
1. There is no "pounds below the top-rated" filter. Official Rating Rank gives a horse's rank in the field, not how many pounds behind the top-rated it sits. The nearest existing field, Lbs Below Highest Winning OR, answers a different question. A proper one is on the build list.
2. "Placed in any of the last four" is not expressible. System filters combine with AND and the four previous runs are separate fields, so the System Builder can ask "placed in all four", not "in any one of them". Also on the list.
Not a Premium member? The System Builder answers questions like this in about ninety seconds, and flags every qualifying runner on today's and tomorrow's cards. See what's included.
FAQ
Do the Ebor Handicap trends actually work?
As descriptions, yes. Across 45,467 runners in GB Flat handicaps of 14+ runners since 2018, horses satisfying all four commonly-published trends — aged 4–6, ran within 66 days, three or more runs that season, won over the trip — win at 6.92% against 4.95% for horses satisfying none. But their A/E is 1.007, meaning the market has priced that advantage almost exactly. They describe a better horse at a fair price rather than a betting edge.
How many days since its last run should an Ebor winner have?
The published trend is "within 66 days", and 14 of the last 17 Ebor winners qualify. But the value sits in a narrower band: horses returning after 42 to 66 days return an A/E of 1.097 across the wider cohort while accounting for only 8.8% of runners, and they have produced 8 of the last 17 Ebor winners including six of the last eight. Horses back after 14 to 27 days are the worst-priced group in the race at 0.933.
What is A/E in horse racing betting?
A/E — actual over expected — compares how many horses in a group actually won with how many the betting market implied would win. 1.00 means the market priced the group perfectly, below 1.00 means you were paying over the odds, above 1.00 means the group outperformed its prices. Unlike profit or ROI, it is not distorted by one or two huge-priced winners.
Why do "10 of the last 12 winners" trends often mean nothing?
Usually because of base rates. "Ran within 66 days" describes 85.1% of all runners in a big-field Flat handicap, so roughly 85% of winners will satisfy it automatically, with no predictive content whatsoever. A trend that hits 10 from 12 — 83% — is reporting that base rate, not beating it. Always check what share of the runners qualified.
Can you back favourites in big-field handicaps?
They win far more often than the cliché suggests — 21.2% of 14+ runner GB Flat handicaps since 2018 — but at an A/E of 1.003 they are priced almost exactly right, so there is no systematic edge either way. The weakest position in the market is actually the third favourite, at 0.915.
The Thursday Take runs every Thursday morning, with a special whenever the big meetings come around. All figures: GB Flat handicaps with 14 or more declared runners, open to 3yo+, 1 January 2018 – 17 August 2026, Proform database — 2,880 races and 45,467 runners. Ebor figures: 17 runnings, 2009–2025. A/E calculated at Betfair Starting Price including all starters, non-finishers included; profit and loss net of 2% commission.
