Luck of the Draw: Analysing Stall Draw Bias at Newmarket (Part 1 – The Rowley Mile Course)

Introduction

Stall draw bias in horse racing is an important factor that often fails to cross the punters’ mind when trying to pick a winner. In simple terms, stall draw bias refers to a potential advantage or disadvantage a horse is given by springing from their elected gate number. The lowest stall number is usually allocated to the closest stall to the inside running rail (known as the ‘inside draw’), with the highest stall number being that of the stall furthest from the inside rail (known as the ‘wide draw’). Stall draws have differing degrees of impact at each racecourse around the UK. This article will focus on the impact of the draw at Newmarket, which provides an interesting place for analysis due to its two courses (July and Rowley Mile) and the wide variety of race distances that are hosted throughout the Spring and Summer months. The goal of this article is to analyse whether or not horses gain an advantage (or disadvantage) based on which stall number they are given.

How Stall Draw Bias Can Impact Race Outcomes:

1. Track and Course Layout: On courses and race distances that include bends, an inside draw can be favourable as it allows the horse to cover less ground, as they are closer to the inside running rail. On the other hand, a wide draw is often seen as a disadvantage due to the fact horses have to travel a greater distance to get around the bend. For straight courses (often with shorter race distances) factors such as pace and rail positioning affect stall bias rather than track curvature. Newmarket’s Rowley Mile course has historically shown trends where certain parts of the track are more favourable, as we will explore later on.
2. Race Distance: The assumption is that short race distances (known as ‘sprints’, up to 6f in distance) display a more pronounced bias as there is less time to recover from a poor starting position. Again an inside draw is often advantageous in these races. For longer races (7f+), the bias is less prominent as there is more time and ground to recover from a poor position out of the stalls. However, stall bias can still affect a horses’ chances if they fail to settle into a good early rhythm.
3. The Going (Ground Conditions): More extreme ground (soft or heavy) can exaggerate stall bias if the track drains unevenly, leading to better ground on one part of the track, giving an advantage to horses drawn on that side.
4. Pace and Running Style: A horse’s natural pace can either enhance or neutralise a stall draw bias. For example, if a race lacks early pace, an inside draw can be advantageous for a front-runner who can dictate the pace from the lead. If there is a strong pace bias towards one side of the track, held-up horses may benefit from a stronger tow into the race.
5. Field Size: Stall draw bias is often more prominent in larger fields, due to crowding out effects preventing some horses from getting a clear run as they are boxed in.

Rowley Mile Track Layout at Newmarket:

The Rowley Mile course is a straight track for races 1m 2f or less, with races longer than 1m 4f involving a wide bend onto the home straight. Despite this bend suggesting a wide draw may compromise chances, due to the home straight being so long it somewhat dilutes the stall draw bias.

Map of the Rowley Mile Course

Methodology for Data Analysis

This article will call upon historical race results from a 10-year period (2014-2024, 10 or 11 races) in order to deliver an answer as to the degree of stall draw bias at Newmarket. For the data analysis, variables will include the likes of going conditions, odds, race distance, field size, and of course stall number. Data will be drawn from three of the biggest races (prize money wise) on each course, ensuring that each race covers a different race distance to provide a wider perception of the findings. Statistical software package STATA will be used to manipulate and visualise the data.

Analysis

The 2000 Guineas

The 2000 Guineas is a Group 1 Flat race held on the Rowley Mile every year during the first weekend of May. The race is run over 1 mile by 3 year old colts and fillies, and nowadays the race is worth over £500,000.

The Findings: The Short Version

Key trends of last 11 races:

• Stall 3 has had the highest number of winners since 2014 (3 from 11)
• Stall 3 also has the best percentage of top 3 and top 5 places, at 55% and 72% (6/11 and 8/11 respectively)
• When Stall 19 has been used (2 of last 11) the horse has finished 2^nd both times
• Stalls 2, 5, 7, 8, 10, 11, 12, 13, 14, 18 and 19 have not had a winner
• Stalls 12, 13 and 18 have not had top 3 finish
• Stall 18 has not had a top 5 finish in the 2 races of the last 11 that it was used
• Stalls 1 and 3 have had multiple winners
• Of Stalls that have been present for at least 80% of the runs in the past decade (9 from 11), only 1 winner has been drawn from Stalls 7-13, and 6 of the 11 winners from these races were drawn in Stalls 1-3
• When Stalls 16 or higher are used, they have a 25% win rate and a 66% place in the top 3
• Average field size is 14

Graphical Representation

Delving Deeper

Results of Linear Regression Model: Stall number has a statistical significance of 0.06, which means that it is (only just) statistically insignificant at a 5% significance level. However it would be significant at a 10% significance level, suggesting evidence of a weak stall draw bias. The coefficient is -0.38, implying that as stall number increases, chances of winning decreases – lower stall numbers are slightly favourable.

Stall bias by ground type is statistically insignificant at both 5% and 10% significant levels. However, the interest here is in the difference between the coefficients. For going that is good to soft of slower, the coefficient presented is 0.2, and for good to firm or faster is only 0.03. This implies that stall draw bias is more prevalent when the ground is softer, which makes sense if there is uneven drainage on the track. By comparing R² values here we see that for good to soft or softer ground the stall draw can explain 4% of variations in finishing positions, but for faster ground this is reduced to just 0.1%.

Results of Binary Logit Model: Generally stall draw bias is statistically insignificant in explaining top 3 and 5 places. But by comparing pseudo R² values here, it does account for 15% of variations in top 3 finishes and for top 5 finishes this lowers to 7%.

Conclusion

For the 2000 Guineas Stakes, there is evidence of some patterns in relation to stall draw affecting the outcome. The number of winning stalls is certainly not evenly spread across all the stalls, with the findings showing greater success for lower stall numbers. However, in large fields, higher stall numbers are often advantageous. Generally speaking, middle stalls tend to have less success (possibly due to crowding out/lack of a clear run). Whilst statistically stall draw bias is insignificant in explaining results, there is still evidence to suggest a weak stall bias is present in this race.

King Charles II Stakes

The King Charles II Stakes is a listed flat horse race held on the Rowley Mile track and is run over a distance of 7 furlongs by 3 year olds, and nowadays takes place over the Guineas Festival at the start of May, but was previously staged a couple of weeks later.

The Findings: The Short Version

Key trends of last 10 races:

• Stall 2 has been the most favourable draw, with 4 wins and 7 top 3 places.
• Stalls 1 & 2 have accounted for 60% of winners
• Stall 3 has not had a single winner in the 8 times it has been used, with just half of the runners finishing in the top 3
• Stall 4, 5 & 6 have seen reasonable degrees of success, with at least one winner
• When stalls 7 or higher have been used, they have generally been unfavourable
• Average field size of 6

Graphical Representation

Delving Deeper

Result of Linear Regression Model: No statistical significance of stall number affecting finishing position. Little point running for ground types as over the past 10 year in this race, the ground has always been in and around good (good to firm – good to soft).

Results of Binary Logit Model: Much more interestingly by running this model (accounting for odds, field size and ground type) stall number returns a p-value of 0.045, below the 5% significance level, therefore implying that stall number is statistically significant in explaining the winning horse. The negative coefficient of -0.46 suggests that as stall number increases, chances of winning decrease. The Pseudo R² value of 0.2475 tells us that this model can be used to explain 25% of variations in winning outcomes, which in the context of horse racing is actually pretty strong.

Conclusion

For the King Charles II Stakes, there is little evidence to suggest a strong stall draw bias. Whilst there is some evidence of stall number affecting the chances of winning, generally speaking there is no link between stall draw and finishing position. There has been a pattern over the past 10 years suggesting an inside draw (stalls 1 & 2) is favourable, and you do not want to be in the middle (stall 3). However statistically speaking, stall draw bias is not a prominent factor, with odds playing a much bigger role here with the favourite winning 7 out of the last 10 races. It is perhaps unsurprising that stall draw bias is not prevalent here due to the generally small field sizes (reducing crowding out affects) and the fact this race is a sprint along the straight of the Rowley Mile Course, meaning no turn in the track to cause issues.

Cesarewitch Handicap

The Cesarewitch Handicap is a flat handicap race run at Newmarket’s Rowley Mile over 2m2f. It is open to horses aged 3 and older, and nowadays is worth around £200,000 in total prize money.

The Findings: The Short Version

Key trends of last 10 races:

• The only stall to have more than one winner is stall 20
• Stalls 3, 6, 9, 13, 15, 19, 23 and 24 have had one winner each
• No winners from stalls 25 or higher
• The middle stalls (13 to 24) have had the greatest success, with 7 of the last 10 winners
• Stall 5 presents the greatest number of top 3 finishes, with 3 from 8 (37.5%)
• Stalls 3, 5, 15, 20 and 28 are the only to have top 3 finishes in more frequent than 20%
• Stall 18 has the highest number of top 6 finishes with 5/10 (50%), whilst stall 19 has the greatest top 6 strike rate of 57.1% (4 from 7)
• Stalls 3, 4, 8, 18 and 19 are the only to have a top 6 strike rate of greater than or equal to 40%
• Only stalls 4, 8 and 25 have an average finishing position of 12^th or better, whereas stalls 1, 12, 14, 28, 29, 31 and 35 all boast average finishing positions of 20^th or higher
• Average field size is 30

Graphical Representation

Delving Deeper

Linear regression models suggest there is no statistical significance of stall number affecting finishing positions. Similarly, a binary logistic regression model (logit) suggest no statistical significance of stall number affecting the likelihood of the horse winning.

Conclusion

The lack of evidence to support stall draw bias here is perhaps unsurprising given the distance of this race, giving time to recover from what might otherwise be considered a poor stall position. Interestingly there doesn’t appear to be any major crowding out affect given the large field sizes, but again this may be counteracted by the longer race distance.

Final Conclusion

The impact of stall draw bias at Newmarket’s Rowley Mile Course varies considerably depending on factors such as race distance and field size. For sprint races (7 furlongs or shorter) and smaller field sizes there is little evidence to suggest a stall number effects finishing position, but rather that it does have some effect on the outright winner only. For middle race distances (around 1 mile) and medium field sizes, there is stronger effect of the stall number on finishing positions to the point where it can be said that it is statistically significant. For longer distances (say 1 mile 4 furlongs+) there is little evidence to suggest stall number has an impact on finishing position or the outright likelihood of winning. This is likely due there being enough time to recover from a poor starting position.

There are several key trends regarding stall number for each of the example races from the past 10 years that should definitely be considered, as you will have read earlier, with some stall numbers performing considerably better than others.

I believe there is enough evidence to suggest that stall draw can have an impact on the race outcome, with some stalls being more favourable and others damaging chances.