Pace & draw

Horse racing draw bias explained

At some tracks the starting stall is worth several lengths; at most it is worth very little. The difficulty is telling the two apart, because the most common way of presenting draw statistics — win percentage by stall — has a bias of its own built in.

Published 21 September 2026 · Last reviewed 21 September 2026 · Written and maintained by The Racing Formula · How our numbers work

Check draw stats for every UK & Irish course →

What the draw is

In flat races started from stalls, each horse is allocated a starting stall by a random draw. That stall number is the draw. Over jumps there are no stalls, so the draw does not apply.

Which physical position a "low" or "high" number gives you depends on the course and on where the stalls are placed that day. A low draw at one track is not the same place as a low draw at another, which is why draw statistics are always course-specific.

Draw bias is when horses starting from one part of the stalls win more often than their numbers alone would explain, at a particular course and distance.

When the draw matters

Most of the time, over most distances, the draw makes little difference. It starts to matter when one or more of these apply:

Short races. In a five- or six-furlong sprint there is little time to recover from a poor position.
Early bends. On a tight, turning track, a horse drawn wide can cover noticeably more ground, or have to drop in behind the field. Chester is the best-known example in Britain.
Big fields. The more runners, the further the widest horse is from the inside rail.
Uneven ground. On a straight course, one side of the track can ride faster than the other, especially on soft ground or after watering.
Where the pace is. Horses drawn near the early leaders can get a tow into the race; one on the wrong side can end up racing alone. Draw and pace should be read together — see the pace maps guide.

Because the draw is made at random, draw statistics avoid a problem that affects most racing statistics: better horses do not get better draws. That makes them worth doing properly.

Why raw stall win percentages mislead

The usual way to present draw bias is to group the stalls — low, middle and high — and show each group's win percentage. It is a sensible start, and it has three traps built in.

1. Small samples

Some course and distance combinations are run only a handful of times a year. A group with 40 runs and 6 winners shows a 15% win rate that could easily be 8% or 22% with a few different results. Always check the number of runs before the percentage.

2. Field size is baked into the groups

A runner in a 5-horse race has a 1-in-5 chance at random; a runner in a 20-horse race has a 1-in-20 chance. So which field sizes end up in each draw group changes the win rate you should expect from that group before any draw effect at all.

If groups are fixed stall numbers — stalls 1–4, 5–8, 9 and above — the high group only exists in big fields, where every runner's chance is small. High draws then look worse than they are.

If groups are thirds of each field — the approach our own tables use — the split is uneven in small fields (in a 4-runner race, two of the four runners fall in the top third and one in the bottom), so the high group ends up with more runners from small fields and should win slightly more often at random. Across UK and Irish sprints since 2015, if the draw made no difference at all, you would expect the low third to win 10.0% of the time and the high third 10.6%. Any raw comparison starts with that tilt.

3. Conditions change

Rail movements, re-sited starts, watering policy, changes to the track and the going can all shift a bias — or remove it. A table that pools ten years of results tells you what has usually happened, not what will happen on today's ground.

A fairer comparison: actual against expected

The fix for the field-size problem is to compare each group with what it should have won at random. For every runner, its expected share of a win is 1 ÷ the number of runners in its race. Add those up for a draw group and you have its expected winners. Divide actual winners by expected winners and you get an actual-to-expected ratio: 1.00 means the group won exactly what chance predicts; 1.20 means 20% more; 0.80 means 20% fewer.

Worked example: Chester, 5–7 furlongs

704 races. Grouping every runner by which third of the field its stall fell in:

Group	Runs	Wins	Win %	Expected	Actual ÷ exp.
Low	1,512	234	15.5%	11.1%	1.39
Middle	2,129	241	11.3%	11.5%	0.99
High	2,502	229	9.2%	11.7%	0.79

Chester's famous low-draw edge is real, and slightly bigger than the raw figures suggest: low draws won 39% more races than chance predicts, high draws 21% fewer. On this many races, that is far outside what luck produces.

Worked example: Thirsk, 5–7 furlongs

882 races, same method:

Group	Runs	Wins	Win %	Expected	Actual ÷ exp.
Low	2,388	181	7.6%	9.0%	0.84
Middle	3,227	307	9.5%	9.4%	1.01
High	3,833	398	10.4%	9.5%	1.10

Here the tilt runs the other way and is smaller: high draws 10% above expectation, low draws 16% below. It is large enough to take into account, and small enough that a few seasons of different results could shrink it — the kind of bias to weigh, not to build a view around.

How these figures were calculated. UK and Irish flat and all-weather races with a recorded draw, 1 January 2015 to 20 September 2026 (824,565 runners in 86,346 races across all courses and distances). Sprints are races under a mile. A runner's draw group is its stall divided by the number of runners: up to 0.33 is the low third, up to 0.67 the middle third, anything above the high third — the same rule as the course tables on this site. Expected wins are the sum of 1 ÷ field size for each runner in the group. Dead heats count as wins for each horse involved.

How to read any draw statistic

Check the sample first. Runs and wins, not just the percentage. A few hundred runs in a group is a starting point; a few dozen is an anecdote.
Compare against expectation, not against each other. If the groups are built differently, their raw win rates are not directly comparable.
Match the trip. Pooling five and seven furlongs can hide opposite effects at different starts.
Think about today. Ground, rail position and where the pace is drawn can strengthen or cancel a historic bias.
Keep it in proportion. Even a strong bias moves a horse's chance by a few percentage points. It is a reason to upgrade or downgrade, rarely a reason on its own.

Common mistakes

Assuming every sprint has a bias. Most course and distance combinations show little or none.
Trusting a percentage without its sample. 3 winners from 12 runs is 25% — and almost meaningless.
Ignoring field size. A bias measured mostly in big fields may not apply to a six-runner race, where every horse can get to the rail.
Forgetting the going. Soft ground can create a bias for one side on a straight course that good ground does not.

How TRF shows draw information

Every flat and all-weather racecourse page has draw tables for four distance bands — 5–7f, 8–9f, 10–13f and 14f and over — showing runs, wins and win percentage for the low, middle and high thirds of the field, with the date the tables were compiled. They are refreshed weekly from the full results history.

Those tables show raw win percentages, so they carry the small field-size tilt described above, and they are not split by going. Use the runs column to judge how much weight a figure can bear.

On the racecards, a draw-bias banner appears on a race only when the best draw group for that course and distance has won at least 14% of the time and at least 5 percentage points more than the worst. Track bias brings the draw, going and favourite figures for every course together.

Put it into practice

Every flat and all-weather course page shows historic win rates for low, middle and high draws across four distance bands, with the number of runs behind each figure.