The phrase “shrewd trainer” is used so loosely in racing that it can mean almost anything. A winner after a price contraction is enough for the label to appear. A well-backed loser is forgotten. The market move is remembered, but the trainer’s normal strike rate, the timing of the support and the price available when anyone could act are rarely considered.

Market Form’s database allows us to test the idea rather than repeat it. This analysis covers 82,534 matched trainer and runner records from 27 December 2025 to 12 September 2026. It compares each trainer’s ordinary strike rate with the results achieved when their runners shortened by defined amounts in implied probability.

The central finding is clear. Support is informative, but it is not automatically valuable. Some yards show a repeatable rise in win probability while prices still leave something for the bettor. Others produce far more winners when backed, yet the market adjusts so efficiently that following the move still loses money. The shrewdest signal is therefore not movement alone. It is movement interpreted against the trainer’s own history.

How the analysis works

Prices were converted into implied probability and the change was measured in probability points. A move from 5.00 to 3.00 is a rise from 20.0% to 33.3%, or 13.3 probability points. This gives comparable meaning to moves at different prices and avoids treating every numerical odds change as equal.

Four cumulative support bands were tested: 5pp+, 10pp+, 15pp+ and 20pp+. A runner moving by 20pp is included in every lower band as well, exactly as on the Stable Intent page. The two most useful windows for bettors are Open to 10am and Open to 1pm because the signal exists before the race. Open to SP and 1pm to SP describe what the completed market did, but cannot be treated as advance betting systems.

Profit and loss uses one point level stakes at the recorded decimal price. Strike-rate comparisons use each trainer’s own valid Market Form sample as the baseline. Clear naming duplicates were consolidated, including Andrew Balding with A M Balding and K R Burke with K. R. Burke.

The market-wide result

Across all trainers, stronger support produced progressively more winners. That did not make the lower bands profitable. Open to 10am runners supported by at least 5pp won 25.9% of the time but lost 933.2 points at SP. At 10pp+, the strike rate rose to 33.6%, yet the group still lost 159.9 points. Only the extreme 20pp+ group moved into a small overall profit, returning 4.8 points from 358 bets.

This is the first warning against a simplistic “back every steamer” approach. The market is good at identifying horses with improved chances, but the shortened price usually absorbs that information. Trainer context is what begins to separate useful intent from market-wide noise.

Andrew Balding shows the cleanest early pattern

Andrew Balding produced one of the strongest combinations of scale, conversion and profitability in the sample. His stable baseline was 20.0%, with 161 winners from 805 runners. When a runner had gained at least 5pp by 10am, the strike rate doubled to 40.5%: 53 winners from 131. Those bets returned 26.40 points at SP.

At 10pp+ the record strengthened again to 33 winners from 67, a 49.3% strike rate and 13.76 points profit. The result was not dependent on one freak-priced winner. The largest winning return in that group was five points, accounting for roughly 36% of the net profit rather than creating all of it.

This is what a convincing trainer-support profile looks like. The sample is meaningful, the lift over baseline is large, the result persists across adjacent thresholds and the return is not built around a single outlier. It does not guarantee the next supported runner will win. It does show that a major early move for this yard has carried substantially more information than the stable’s ordinary runner profile.

Ed Walker combines conversion with price

Ed Walker’s baseline was 16.2%, with 72 winners from 445 runners. His 10pp+ Open to 10am group won 19 of 39 races, a 48.7% strike rate. The group made 16.88 points at SP, an ROI of 43.3%, and 15.26 points at the recorded 1pm prices.

The broader 5pp+ group also remained profitable, winning 30 of 87 and returning 12.77 points at SP. That matters because it shows the result is not confined to one narrowly selected cut-off. Walker was one of only a small group of trainers who showed a positive strike-rate lift and positive SP returns across both the 5pp and 10pp bands in both early market windows.

James Owen demonstrates why the threshold matters

James Owen’s figures are a useful lesson in selectivity. His baseline was 16.9%, based on 140 winners from 827 runners. A 5pp move by 10am lifted the strike rate to 32.1%, but backing all 134 qualifiers still lost 12.57 points at SP. The information improved the chance of winning without creating value at that threshold.

At 10pp+, the picture changed. Twenty-eight of 59 runners won, a 47.5% strike rate, for a profit of 10.90 points at SP. The 15pp+ group won 12 of 26 and the 20pp+ group won six of nine, although nine bets is too small to support a firm conclusion on its own.

The practical point is not that bigger must always be better. It is that this yard’s data contained a clear dividing line. Modest support was meaningful but heavily priced; stronger early support was materially more selective.

K R Burke and the difference between winners and value

After consolidating the K R Burke and K. R. Burke records, the stable baseline was 19.6%, with 111 winners from 565 runners. The 5pp+ Open to 10am group won 35 of 82 and returned 26.51 points at SP. That is a powerful rise to a 42.7% strike rate.

The 10pp+ early group still won 16 of 34, but finished almost exactly level at SP. By 1pm, the 10pp+ group won 22 of 54 and returned 9.63 points. The apparent 20pp+ record was eight wins from nine and 17.36 points profit, but that is a tiny sample. In addition, one 15.00 SP winner generated 14 points and more than accounted for the net profit in the wider 10pp group. The conversion signal is strong; the profit claim requires more restraint.

Kevin Ryan and Jane Chapple Hyam are strong developing cases

Kevin Ryan recorded a low 9.8% baseline, with 38 winners from 389 runners. When support reached 5pp by 10am, 16 of 41 won and the group returned 9.22 points at SP. By 1pm, the 10pp+ group produced 13 winners from 26, returning 8.64 points. The lift is exceptional, but 26 bets remains a developing rather than definitive sample.

Jane Chapple-Hyam’s baseline was 14.5%, with 25 winners from 172 runners. Her 10pp+ Open to 1pm group won 13 of 22 and returned 23 points at SP. The largest winner contributed five points, so the result was not manufactured by one double-figure success. Even so, 22 bets calls for continued monitoring. The numbers justify attention, not certainty.

Roger Varian shows how stronger bands can sharpen the read

Roger Varian began from a higher baseline of 22.8%. His 5pp+ early groups raised the strike rate but lost money at SP. The more concentrated Open to 1pm bands were different: 21 of 43 won at 10pp+, 13 of 19 at 15pp+ and eight of 11 at 20pp+. The 15pp group returned 4.17 points at SP and 5.25 points at the recorded 1pm prices.

This profile illustrates why a single trainer headline can mislead. “Varian runners are significant when backed” is too broad. The data suggests that the more substantial moves carried the clearest evidence, while ordinary support was not enough to overcome the price.

Willie Mullins is the clearest warning about short prices

Willie Mullins supplied perhaps the best example of the difference between being right and getting value. His baseline was already 19.5%, from 168 winners among 863 runners. At 10pp+ by 10am, 17 of 38 won, a 44.7% strike rate. Despite that impressive conversion rate, the group lost 5.88 points at SP and 7.67 points at the recorded 1pm prices.

At 5pp+ by 10am, 33 of 95 won, yet the loss reached 27.84 points. Every tested early band from 5pp through 20pp showed an SP loss. The market was correctly identifying likely winners, but it was also charging fully for the information.

Mullins also had a winner on 61 of 74 days when at least two runners from the yard shortened from Open to SP, an 82.4% day-level figure. That sounds extraordinary, but it must be read correctly: the test counts any stable winner on a day with two or more supported runners, not necessarily a winner from the supported subset, and the support definition uses SP. It describes powerful stable-day activity; it is not a pre-race betting rule.

Tony Carroll and Ian Williams show the danger of assuming a smooth progression

Tony Carroll’s baseline was 12.5%. His 10pp+ Open to 10am runners won 13 of 32 and returned 5.61 points at SP. Yet the 15pp+ subset won only three of 11 and lost 3.97 points. Ian Williams showed a similar break: his 10pp+ early runners won 12 of 34 and returned 12.95 points, but the 15pp+ subset produced only one winner from 13.

These results are reminders that cumulative bands do not always form a neat staircase. Smaller samples become volatile, race composition changes and an extreme move can occur for reasons that do not improve the horse’s chance in a linear way. A trainer profile should be read as a distribution, not reduced to “the bigger the move, the better.”

A comparison of the leading early profiles

What the evidence says about shrewd trainers

A shrewd stable cannot be identified from one backed winner. The strongest cases in this study shared several features: a meaningful sample, a large improvement over the yard’s own baseline, evidence across more than one adjacent threshold and a return that was not wholly dependent on one outsider.

Andrew Balding and Ed Walker supplied the most rounded early-market profiles. James Owen’s record showed a meaningful threshold at 10pp. K R Burke’s runners converted support strongly, although the price and one large winner complicate the profitability story. Kevin Ryan and Jane Chapple-Hyam produced eye-catching results that deserve further data. Roger Varian’s strongest evidence appeared only at the higher bands. Willie Mullins showed that even a very accurate market signal can be a poor blind bet.

That is ultimately what Market Form is built to reveal. Odds movement is only the visible event. Its meaning depends on who trains the horse, how unusual the move is for that yard, when the support arrived, what happened under comparable conditions before and whether the available price still compensated for the risk.

The market often knows. The more important question is whether it has already made you pay for knowing too.

Methodology and limitations

• The sample contains 82,534 matched Market Form and race-result records from 27 December 2025 to 12 September 2026.
• Support is measured as the change in implied probability, expressed in percentage points rather than the percentage change in odds.
• Bands are cumulative. Results at 20pp+ are also present in the 5pp+, 10pp+ and 15pp+ samples.
• Level-stakes returns use recorded decimal prices and do not include commission, staking variation, liquidity or an allowance for prices becoming unavailable.
• Open to SP and 1pm to SP are descriptive closing-market studies. The full move is not known early enough to act on as stated.
• Small samples can produce extreme strike rates and returns. Figures should be monitored as the database grows rather than treated as permanent trainer characteristics.
• The analysis is descriptive and does not prove that trainer behaviour caused the price move or subsequent result.