Revisiting All-Weather Draw Biases
It has been a while since I looked at all-weather draw biases so with plenty more racing on those surfaces to come this winter it seemed a good time to revisit the stats, writes Dave Renham.
Introduction
When I think about draw biases in turf racing, they tend to be far less predictable or consistent than draw biases on the all-weather (AW). There are a few reasons why this tends to be the case:
1. Some turf courses are increasing the number of meetings and they move rails to alter where the horses run. This can nullify or even change a draw bias.
2. The positioning of the stalls can change at some courses altering the track location that the horses are running from each draw. For example, on a straight track a horse could be drawn 1 when the stalls are positioned far side and be next to the far rail, but if the stalls on are the stands’ side the horse drawn 1 could be out in the middle of the course. Hence, if we hypothetically assume that the ground next to the far rail is quicker than the ground down the centre, then a horse drawn 1 could have a big advantage if the stalls are placed far side. Conversely if the stalls are stands' side, then the horse drawn 1 does not have this advantage. Moreover, running against a rail is generally more of an advantage than running in the middle of the track.
3. Turf courses use watering which potentially can change the going on certain parts of a track: that may then have an impact on any bias.
As far as the AW is concerned, these three reasons are not much of an issue. At AW courses the rails are fixed so moving them is not an option, and the position of the stalls doesn’t change either. Also, the going tends to be very similar on all-weather – the surface they race on is clearly the same each time, and whether it rides ‘standard’ or slower than ‘standard’ depends on the specific course surface, any additional harrowing undertaken and possibly the weather conditions.
The table below shows the percentage of races at each course in the past three years that has ridden either ‘standard’ or ‘standard to slow’. Those are the only two going descriptions we have had since the start of 2022.
As can be seen most of the courses primarily race on ‘standard’ going. At Chelmsford, Lingfield and Wolverhampton they rarely race on slower going. Kempton has raced on ‘standard to slow’ at every meeting over the past three years (indeed every meeting since 11th July 2018 - when does that become the new 'standard'? - Ed.), while Southwell sees slower ground about one meeting in five.
Newcastle is the course where there is the most even split with just under two-thirds of meetings having raced on ‘standard’, the other third on ‘standard to slow’. It should be noted that between the months of June and September around 60% of the Newcastle meetings have been on ‘standard to slow’, so one could logically surmise that either drier or warmer weather has the most impact in terms of the going there. All the course data points to the fact that how an AW course rides is rarely affected by significant changes in going.
In addition to all of the above, four of the six UK AW tracks see all races run around a bend. This can help in terms of consistency when it comes to the draw. Running around a bend means horses running closest to the rail having to run a shorter distance, thus on round courses lower draws tend to have an edge because they are drawn closer to the inside.
This introduction has made the hypothetical case for more consistent draw bias on the artifical surfaces, so let us see if this has actually been the case. The draw data I have collated goes back to the start of 2017 with the focus on 8+ runner handicaps, which are the best races to analyse when it comes to the draw. I will be looking at four course/distance combinations in detail, arguably the four strongest AW draw biases, starting with the minimum trip at Chelmsford.
Chelmsford 5f
The 5f distance at Chelmsford should favour lower draws because the bend they run starts little more than a furlong into the race and then sweeps round for two further furlongs until they turn into the straight. The racecourse map is shown below:
Let us look at the draw splits for the whole-time frame in terms of win percentages within each third of the draw:
As expected, lower draws have an edge over middle drawn runners, who in turn enjoy an advantage over high draw runners. If there was no bias all these figures should be close to the 33.3% mark. The lower the draw the better with horses drawn 1 winning 31 races (18.3% of all races), horses drawn 2 winning 24 (14.2%). Combined, the two lowest draws have won 55 races, compared with the two highest drawn horses in each race who have collectively won 22 between them.
This suggests that horses drawn in the lowest two stalls are 2.5 times more likely to win than those drawn in the highest two stalls. This disparity increases when the number of runners increases, peaking at 4.3 times more likely when the field size is 11 or 12. It should also be noted that horses drawn 1 have made a profit to SP and BSP, as have those drawn 2. The profits are very small to SP, but with 169 runs for each stall position this suggests there is some value backing these two draws.
In terms of Percentage of Rivals Beaten (PRB) there is positive correlation with the win data with low on 0.54, middle 0.52 and high down at 0.45.
If we now look at the record of the lowest third of the draw by individual year we might expect to see big fluctuations, due mainly to small sample sizes and standard variances. To counter that, I have grouped the annual data in a slightly different way using a method I first saw in Nick Mordin’s excellent book, Winning Without Thinking. He looked at data in batches / groups of years, which is a good way to compare things more effectively due to more reliable sample sizes. You can also see patterns changing more easily – if indeed they do change. He used five-year batches; I’m going to look at four-year batches. Below is a graph comparing the win percentage for the lowest third of the draw using the batch/group method:
The bias has remained fairly consistent with all four-year batches showing 40%+ figures. The graph suggests that there may be a slight strengthening of the bias over the last eight years but, regardless of whether that is true, it is clear I hope that this bias is likely to continue for the foreseeable future.
Before moving on, there seems to be a draw bias ‘cut off point’ at stall 7. Specifically, horses drawn 7 or lower have won 144 races from 1183 runners (SR 12.2%). Those drawn 8 or higher have won 25 races from 424 runners (SR 5.9%). In addition, over this track and trip, horses drawn 11 or 12 (the maximum field size is 12) are 1 win from 50 (SR 2%).
My betting strategy as regards the draw over 5f at Chelmsford is to primarily focus on horses drawn 1 and 2, but I am prepared to look at horses drawn as high as 7. Those drawn 8 or higher will get a line through them unless they have several strong factors to bring to the table.
Chelmsford 6f
Remaining at Chelmsford we move up a furlong to six. With the bend being an additional furlong away from the start, one would expect that the low draw bias might be less potent than it is over 5f. Let’s look at the win percentages by third of draw first:
As expected the edge for low draws is less strong and, also, high draws have fared much better at this trip in terms of wins compared with 5f. However, in Percentage of Rivals Beaten (PRB) terms the figures are more akin to the 5f cohort, especially looking at the low third and the high third. The PRB for low draws stands at 0.55 (0.54 over 5f), middle lies on 0.49 (0.52 over 5f) and high at 0.46 (0.45 over 5f). These figures suggest low’s advantage over high is similar to that of the minimum distance. The win & placed figures are also very similar when comparing the draw thirds over the two distances. Essentially, I would say that from a win perspective, the bias is less strong over six furlongs while in terms of getting placed it is similar.
It should also be noted that the maximum field size over 6f is 14, two more than over five furlongs. It is rare that 13 or 14 runners come to post here but, for the record, no horse drawn 13 or 14 won in the eight-year period from 38 runners to try.
Looking through the value lens, however, using A/E indices, higher draws have scored best. Their figure of 0.92 is well above the low figure of 0.81 and middle one of 0.82. Perhaps bookmakers and punters perceive the bias to be stronger than it is at this distance.
My betting strategy as regards the draw over 6f at Chelmsford is to only rule out horses drawn 13 or 14. Hence for most races that means I would not be put off by any draw position. Indeed, as the A/E indices showed there is probably value looking higher than lower. [In fact, run style is arguably much more material at this course and distance - see this article (from late 2021), partially replicated below.
Kempton 6f
Traditionally this range at Kempton has produced the strongest and most consistent draw bias on the AW. For years low draws have held sway. There is a similarly strong bias over 5f, but qualifying races are rare with only three in the past three years (that trio were won by horses drawn 1, 1 and 2). Let us look at the win percentage splits for each third of the draw going back to 2017:
Low draws have won 46.5% of all races which is the biggest percentage for any all-weather course and distance combination using the 8+ runner handicaps condition. Additionally, the number of races during this time frame was 400 so a strong sample size. Backing horses drawn 3 or 4 would have yielded a profit to BSP of 10p in the £. In terms of Percentage of Rivals Beaten (PRB) these figure correlate positively showing that clear edge to lower drawn runners:
The low to high comparison of 0.57 to 0.42 further illustrates the strength of this bias. From a punting perspective low draws offer the best value as well, with an A/E index of 0.93 compared with 0.88 for middle draws and a lowly 0.68 for high.
Earlier I discussed a Nick Mordin idea of grouping data in yearly batches to give more accurate comparisons over different time frames / years. As I did with Chelmsford over 5f earlier, here are the four-year win percentages over 6f at Kempton for the lowest third of the draw using this grouping method:
As can be seen this bias has remained consistently strong over the whole of the eight-year time frame. I also used this idea for the PRBs, across all three sections of the draw, and we see excellent consistency once more.
Before moving on, it should be noted once more that the low draw bias increases as field size increases. The maximum field size is 12 and looking solely at races contested by 11 or 12 runners, the lowest third of the draw won more than half (52.2%) of the races with a PRB of 0.59; the highest third won just a sixth of the time with a PRB of 0.41.
Wolverhampton 5f
The fourth and final course/distance combination takes us to Wolverhampton and their minimum trip. Let's start with a look at the draw splits based on win percentage by third of the draw:
The implication is again of a decent low draw bias, with a repeat of the pattern seen above in terms of the lower the better. The PRB figures correlate positively with the win percentage splits as the graph below illustrates:
All things considered one would much rather be drawn low than high.
Let us now compare 2017-2020 with 2021-2024 rather than the 4-yearly groupings I used earlier. I am doing this for two reasons: one, to mix the article up a bit; and two, to clearly highlight the more recent struggles of higher draws. This would have been shown using the earlier Mordin method, too, just so readers know I’m not moving the goalposts in any way:
The chart illustrates how higher draws have diminished nearly 8% in terms of win success across the two four-yearly time frames. This is quite a steep decline but it is not easy to understand why. One reason may be because the average price of the highest three drawn runners has been slightly higher in the 2021-2024 period compared with 2017-2020. However, the average price difference is negligible really and it could be a 'self-fulfilling prophecy' due to this bias being subsumed in the market, so perhaps something else is going on. Maybe it is down to statistical variance? It is impossible to say.
My final graph this week shows the win success rate of each individual stall / draw position at Wolverhampton over 5f.
This presents a clear indication that stall 4 is probably the cut-off point for top bias. In fact, if you had backed stalls 1, 2, 3 and 4 blind over the eight-year review period you would have ended up with a profit to BSP of £102.79 to £1 level stakes. This equates to a 7p in the £ return.
Summary
To conclude, the draw biases shown have been highly consistent especially in terms of the strongest stall positions which, in each case, has been the lowest drawn third.
Wolverhampton 5f has been less consistent with its high draw data, but overall, the draw biases at these four track/range combos have virtually remained the same throughout the eight-year period. In addition, there is no reason to suggest this might change any time soon (although markets are likely to eventually further cotton on and adjust prices accordingly).
Finally, and the most important thing from a punter’s perspective, low draws continue to offer the best value at each of Chelmsford 5f, Kempton 6f and Wolverhampton 5f, while fielding against the bias with higher draws over 6f at Chelmsford - especially with runners capable of a forward run style - looks the way to go.
Until next week.
- DR
Leave a Reply
Want to join the discussion?Feel free to contribute!