Draw Bias in Horse Racing: UK & Irish Flat Courses by Distance

Home › Racecourses › Draw Bias

By The Guv'nor · Last updated October 2026

Draw bias is when horses starting from one end of the stalls win more than their share at a course and trip. Across 9,187 UK and Irish turf Flat races with eight or more runners (March 2024 to October 2026), runners in the lowest third of the stalls won 9.5% of the time and those in the highest third 8.5%. At Beverley's 5f it was 13.0% against 6.2%.

Look, plenty of draw bias pages are opinions, or tables that haven't been touched in years. I'm essentially a statistics/maths geek, so I pulled the published results for every turf Flat race in Britain and Ireland since March 2024 and counted it myself: 103,194 runners, 53 courses, every stall. Below is the master table, which biases are real and which are noise, and whether the prices already allow for them.

What draw bias is in horse racing, and why it happens

Almost every Flat race starts from stalls, and each horse is given a numbered stall when the field is declared. Stall 1 is "low" and the highest number is "high". Draw bias is when one end of those stalls wins more than its fair share at a course and distance. Track bias is the wider idea: a strip of ground, a rail or a running style that's favoured on the day, whatever the stall numbers say.

It happens for a handful of reasons, and they stack up:

• Bends. On a round course a horse drawn wide either burns energy getting across or races wide round the turn. The sooner the bend comes after the start, the bigger the effect.
• Cambers and slopes. Where the ground falls away from one rail, runners can drift down the slope and lose ground.
• The ground. Watering, drainage and rail movements can make one side of a straight course quicker than the other, and that can change from meeting to meeting.
• Pace. Big-field sprints often split into two groups, and the group with more early speed can come out on top whatever the stalls say. Plenty of "draw bias" is really pace bias in disguise.

One thing that trips people up: stall 1 isn't always on the same side. On the straight courses in my data, race comments put low-drawn runners on the far side and high-drawn runners on the stands' side at most tracks. At York's 6f, 77 of the 83 side mentions for low-drawn runners said far side. Windsor and Newmarket's July Course are the odd ones out: at Windsor's 6f, 50 of 54 mentions for low-drawn runners said stands' side, and on the July Course 62 of 63 did. Always check where the stalls are on the day; my guide to UK and Irish racecourses has each track's direction and shape.

The master table: every course and trip with 40 or more races

One row per course and trip, built from 9,187 turf Flat races with eight or more runners between 18 March 2024 and 4 October 2026. Trips run from 5 furlongs (about 1,000 metres) up. How to read it:

• Win % low / mid / high splits each field into thirds by stall, after taking out non-runners.
• Winners' average draw puts the winner's stall on a 0 to 100 scale (0 = lowest stall, 100 = highest). With no bias you'd expect 50. Below 50 means low-drawn winners, above 50 means high.
• Verdict is a plain statistical test of that average. "Strong" means a gap that big would turn up by chance less than 1 time in 100, "likely" less than 1 in 20, and "lean, unproven" is a hint you shouldn't bet on by itself.

A health warning, because I'd rather be honest than exciting. With 82 rows, you'd expect about four to reach "likely" by pure luck. 18 did, so there's real signal here, but some of those 18 will be flukes. Three rows hit "strong", against roughly one you'd expect by chance.

Great Britain

Ireland

The courses with the strongest biases

Rank the master table by the strength of the evidence (how unlikely each gap is to be luck) and these are the course and trip combinations at the top. I've added Goodwood's 5f even though it only has 32 qualifying races, because the gap is too big to leave out. Treat it as a small sample.

Low dominates. Pooled across every race, low-drawn runners won slightly more than their share, and most of the standout rows are low. Plenty fit the bend explanation: Catterick's sprints both start on or join the round course, and the 1m 2f trips at Newbury and York are round-course races.

Then look at Goodwood's 5f: the high third had 131 runners in those 32 races and two of them won. Crazy, right? But 32 races is 32 races. I'd want another season before calling it gospel.

When the bias flips: going, field size and pace

Going

Pooled across all 9,187 races, the low edge was stronger on soft ground. On good or faster (6,221 races) the winners' average draw was 49.1. On good to soft or softer, including Irish yielding (2,959 races), it was 47.0, a gap bigger than chance would explain. Some tracks swing a lot more:

Pontefract's 6f is the eye-catcher: nothing on fast ground, then a big low edge once it softens. Catterick's 5f goes the other way. Every soft-ground sample here is small (15 to 37 races), so read these as "watch for it", not rules. The going terms are explained in horse racing going explained.

Field size

The pooled low edge showed up in 8 to 11 runner races (5,885 races, average 48.6) and 12 to 15 runner races (2,396 races, average 47.7), but not in fields of 16 or more (906 races, average 49.1, within chance). That doesn't mean the draw stops mattering in big fields. Big-field biases are course-specific and point both ways, so they cancel out when pooled. Use the course row, not a general rule.

Pace

Using the race comments, I flagged every runner described as leading or disputing the lead early. Those early leaders won 17.3% of the time (1,460 of 8,455), against 8.9% for the average runner. Low-drawn runners got to the front more often: 9.5% of them led early, against 6.9% of high-drawn runners. At Chester's 6f, 13 of 133 low-drawn runners led early and only 2 of 160 high-drawn ones did. That's draw and pace working together, so always ask where the speed is drawn. The comment-based flag is approximate, so treat these as ballpark figures.

Is the bias priced in? Returns by stall group

This is the question that matters for your pocket. A draw bias only makes money if the price hasn't already allowed for it. So I checked returns at starting price (SP) to 1pt level stakes, and A/E (actual winners divided by the winners the SP implied; 1.00 means the price was right). Then I ran a fair test: I took the course and trip combinations that showed a "likely" bias in 2024 and 2025 (30 or more races each) and checked the favoured side in 2026, a season they hadn't seen.

The market already knows most of this. Low-drawn runners won more often overall, but their A/E was only 1.03, so the prices were nearly spot on. The 10 biased combinations looked brilliant with hindsight (A/E 1.23), but in the season they hadn't seen, the edge shrank to an A/E of 1.04 and the favoured side still lost 15.3% of stakes at SP. That 2026 test is only 193 races, so the true figure could sit some way either side.

Now the last row. In those same races, everyone outside the favoured third lost 41.2% of stakes. Their A/E was 0.98, so the market priced them about right too; they lost more mainly because they were bigger prices, and big prices lose more at SP (the bookies' margin plus the long-shot effect I've covered in value betting in horse racing). So the draw earns its keep as one reason to cross a horse off, not as a system on its own.

How I use the draw in my own betting

The question I ask isn't "is this horse well drawn?" It's "has the price already allowed for the draw?" The table above says it usually has. If a horse on the golden side is a short price because everyone's read the same stat, there's nothing left in it for me. If the draw is a positive and the horse is still a decent price, now we're talking.

That's why the draw matters most to me with bigger-priced horses. I've always said I'd rather back outsiders and each-way value than odds-on shots, because even when they place we make more profit than an odds-on shot winning. A good draw can be what gets a big price into the places, and a bad one is a fine reason to leave a horse alone.

But it's one input, never the whole bet. I don't back a horse just because of its stall, I don't up my stakes because "the draw looks huge today", and if I'm not sure the bias is live on the day, I skip the bet. Strict points staking, same as always. Understanding that losers are a thing is the key to being a winner longer term, and a perfect draw doesn't change that. It's one line on the checklist in how to pick a winner.

Course-by-course deep dives

The courses people ask about most. Same data and filters as the master table.

Windsor draw bias

Windsor's 6f 12y is the busiest sprint trip in the whole dataset: 132 races. The high third won 10.4% against 7.6% for the low third, with a winners' average draw of 55. That's a lean, not proof. Remember Windsor's low stalls race on the stands' side, so "high" here means the far side. The 5f 21y showed nothing (52, 50 races), nor did the 1m 31y (47, 79 races). The 1m 2f leaned low (44, 52 races).

Thirsk draw bias

Thirsk's 6f is one of only two GB rows where high came out "likely": high third 11.5% v low 6.2%, average 57 over 84 races. High is the stands' side here: all 62 side mentions for high-drawn 6f runners said so. The 7f leaned high (57, 59 races, and 60 on good or faster). The 5f showed nothing at all (53, 55 races).

Beverley draw bias

Beverley's 5f is the strongest GB bias with a decent sample: 100 races, low third 13.0% v high third 6.2%, average 40. It held on good or faster (41, 77 races) and on softer ground (39, 23 races, small). Backing the high third at SP lost 61.0% of stakes. But the low third's A/E was just 1.08, so the market mostly prices it. The 7f 96y (51, 53 races) and 1m 100y (55, 39 races) showed no clear bias.

Newmarket draw bias

Newmarket has two courses, and results sites file them both under "Newmarket", so I split by month: June to August as the July Course, the rest as the Rowley Mile. The race comments back that up, because the two courses number the stalls from opposite sides. At the 6f, low-drawn runners on the "July" dates were described as racing stands' side 62 times and far side once; on the "Rowley" dates it was far side 32 times and stands' side twice. So a low draw at Newmarket means a different rail depending on the course.

Either way, no trip passed the "likely" line. Rowley 6f, 7f and 1m averaged 54, 47 and 52; July Course 6f and 7f averaged 52 and 44. Check the stalls position on the day.

Doncaster draw bias

Doncaster's straight 6f 2y (50, 81 races) and 7f 6y (48, 114 races) are about as fair as it gets. The 5f 3y leaned high (59, 39 races, small). The straight mile, the Lincoln trip, averaged 42 over 32 races, and 36 in the 21 run on softer ground. Small numbers again, but worth knowing in March. My Flat season preview covers the Lincoln itself.

Goodwood draw bias

Goodwood is low all the way through the sprints. The 5f averaged 34 (32 races, small) and the 7f 39 (40 races, low "likely": 13.6% v 3.9%). The 6f leaned low too (43, 42 races). Low is the far side on the straight sprint course: 26 of the 27 side mentions for low-drawn 5f runners said far side. The 1m showed nothing (50, 36 races), and the 1m 1f 197y leaned high (58, 40 races).

Epsom draw bias

Every Epsom sample is small. The 7f 3y averaged 46 (32 races), the 1m 113y 63 (26 races), the 1m 2f 17y 48 (20 races) and the 1m 4f 6y 42 (28 races). The 1m 113y looks high on paper, but with 26 races and 82 rows tested I wouldn't trust it yet. The 5f had fewer than 15 qualifying races.

Chester, York, Ascot, the Curragh and the all-weather

These get their own pages because there's more to say, and those pages go back further than 2024, so their figures won't match this table exactly. Chester's 6f 17y leaned low here (40, 44 races), and harder on soft: see my Chester draw bias guide. York's 5f, 6f and 1m 2f 56y all came out low "likely", which isn't what every source says, so I've dug into it in York draw bias. Ascot's 5f favoured high (61, 42 races): Ascot draw bias covers the straight and round courses. The Curragh's 6f and 7f leaned high on good or faster: Curragh draw bias. And all-weather tracks are left out of this page entirely; they're in my all-weather racing guide.

Method, data and update schedule

• Source: Sporting Life's published results, which give each runner's stall, finishing position and SP.
• Period: 18 March 2024 to 4 October 2026. That's the 2024 and 2025 turf seasons in full and the 2026 season up to 4 October.
• Races: turf Flat races in Great Britain and Ireland with 8 or more runners. All-weather and jumps races are excluded (Laytown's beach meeting is listed as turf, so its six qualifying races are in). 9,187 races and 103,194 runners made it through, from 53 courses. I left out 21 dead heats and any race without a full set of stall numbers.
• Draw: non-runners are removed and the stalls re-ranked, so stall 3 in a race where stalls 1 and 2 didn't run counts as the lowest.
• Thirds and averages: each field is split into low, middle and high thirds by that rank. The winners' average draw is tested against 50 with a standard z-test: "likely" at the 5% level, "strong" at 1%.
• Prices: returns are at industry SP to 1pt level stakes. A/E uses SP probabilities scaled so each race adds up to 100%. These are SP figures, not Betfair SP (BSP).
• Course and trip: the trip is the published distance, so a rail movement that changes the yardage can split a trip into two rows. Newmarket is split by month as a stand-in for the two courses.
• Sides and pace: the side-of-track and early-pace counts come from searching the published race comments for phrases like "far side", "stands' side", "made all" and "led", so they're approximate.
• Updates: I'll refresh the tables every February to March, before the turf season starts.

Members of the Winners Club also get my Draw Biases In Flat Racing special, which stays members-only.

Draw bias FAQs

Is a low or a high draw better in horse racing?

Neither, as a rule. Pooled across 9,187 UK and Irish turf races, the low third won 9.5% and the high third 8.5%, a small edge that the prices mostly allowed for. What matters is the specific course, trip and going, so use the master table rather than a general rule.

Which racecourse has the biggest draw bias?

In this data, Beverley's 5f is the strongest bias in Britain with a solid sample (100 races, winners' average draw 40). Goodwood's 5f and Catterick's 5f 212y showed bigger gaps on smaller samples. In Ireland, Gowran Park's 1m stood out (69 races, average 37).

Does the draw matter in jump racing?

No. Jumps races start from a tape, not stalls, so there's no draw. Where a horse races still matters, but that's track position, not draw bias.

What is the draw bias for 5f races?

Across 1,164 races at 5f to just under 5.5f, the low third won 10.0% and the high third 8.9%, with a winners' average draw of 48. That's close enough to 50 that chance could explain it. The big 5f biases are course-specific, with Beverley, Goodwood and Catterick favouring low and Ascot favouring high in this data.

More in this guide

• UK and Irish racecourses: every track by direction, shape and draw
• Chester draw bias by trip and going
• York draw bias at 5f, 6f and 7f
• Ascot draw bias on the straight and round courses
• All-weather racing and the draw at every AW track
• How to read horse racing form