Cheltenham 2021 – Interesting Stats and Potential Angles
For this years edition of Interesting Stats and Potential Angles I’m going to concentrate on the handicaps. I will look at how different trainers do in them, which ones seem to get improvement out of their horses for Cheltenham, and which ones perform better or worse than market expectations. I will also look at which courses form holds up well at the festival, how horses with different run styles do plus plenty more. For the purposes of this piece I will be using Proform Software.
If you want my best bets and analysis for each day of the festival, you need to subscribe to my Cheltenham 2021 Tips email list. We have had 7 winning festivals from 8 since this site was launched, for a profit of +342 points, and a remarkable ROI of+47%. That doesn’t include the many extra winners I have provided such as Sire Du Berlais at 14.5 and Champ at 6.0 last year.
Before we get into the data I’ll explain what each column means. The data is based on the last six festivals, I would rather use eight for this type of research, as it’s long enough to give a decent sample size, but not going so far back as to make the early part irrelevant. The reason I’m only using six years, is Proform recently introduced Betfair Place SP’s into the systems analyzer, and were as I used to have to export the data into excel, and import the place SP’s myself I can now do it within the software, but they only have the place SP data since 2015. The Betfair Place SP is very important when analyzing small samples, as although we are trying to predict winners, and horses who will be under bet in the win market, the increased amount of places versus winners in the data, has the effect of making that stat much more robust, and as such it will be a much better predictor of whether the market over, or undervalues a certain variable, than the historical win results. For each stat I’ve set a minimum sample of 10 to be included, as any less really is too little to tell us anything.
Runners: This is simply the amount of runners.
Winners: The amount of winners.
PRB^2: This is the decimal percentage of rivals beaten, squared, and will be a much better stat to use than either win or place strike rates, when evaluating past performances like this. Win strike rates have far too much noise in them with small samples, so place ones are much more useful, but you could argue percentage of rivals beaten is more valuable again as it includes more information. The negative for percentage of rivals beaten is it will rate the difference between 9th and 10th in a 10 runner race, the same as the difference between 1st and 2nd. PRB^2 accounts for this by squaring the decimal of the percentage of rivals beaten, and as such weights the difference between high up finishes higher. A trainer with a winner and a last placed finish from two runners will have a PRB^2 score of 50%, but a trainer with two runners who both beat 50% of their rivals, will have a PRB^2 score of just 25%, and this is a better reflection of their achievement than the percentage of rivals beaten stat which would have both on 50%.
Ex Wins: This is expected wins based on Betfair SP, so a horse with a Betfair SP of 5.0, or 4/1, would have an expected win of 0.2. A 2.0 or even money shot would be 0.5
A/E: This is actual wins divided by expected. So a stat that had 10 winners when only 8 was expected would be 1.25, and this would also be the percentage profit you would have made if you had backed each of the qualifiers to return one unit.
bpPlcs: This is the amount of Betfair place winners, which could differ slightly to official places as if say a 16 runner handicap has a non runner, the official places will become three, were as the Betfair places will still be four.
bpWax: This is the difference between Betfair expected places and actual. A positive number means you would have made a profit backing the qualifiers for a place, and this is also a very good indicator of future win profits, as it is more robust in small samples.
bpExp: This is the expected amount of places based on Betfair Place SP, so a horse at 5.0 would be 0.2, while an even money shot to place would be 0.5.
bpAE: This is the Place AE which is actual places divided by expected.
bpChi: This is a chi square statistic which is a test of significance, in this case the significance between the difference in actual places and expected places.
Cheltenham Handicap Hurdles – Trainers
The major standout with trainers in handicap hurdles is Gordon Elliott’s record. Both the Win A/E of 1.81 and Place A/E of 1.62, show the market has massively underestimated his runners the past six years. He has had 25 places when only 15 would be expected given their price, while his PRB^2 score shows his handicap hurdlers at the festival have performed exceptionally well. Not far behind him on that metric is Dan Skelton and with a Place AE of 1.17 his runners have also exceeded market expectations, and his high PRB^2 figure suggests he has had plenty of runners that just missed out on a place too. With Denise Foster taking over following Gordon Elliott’s suspension, the question is how much effect that will have, but I would expect little will change with how the yard operates, the horses that have moved to other yards are mostly running in the graded races, and you’d assume the likely sorts for the handicaps had been plotted out well before any suspension, and as such I wouldn’t treat runners from Cullentra any different to normal.
Cheltenham Handicap Hurdles – LTO Course
It’s a small sample but horses running in a handicap hurdle at Cheltenham that had their prep run at Thurles have a very high PRB^2 score of 52.6%. They only had 2 places from 13 runners, when 3.13 would have been expected, but the PRB^2 stat suggests plenty of those that were unplaced were very close to the places. Naas and Chepstow come out very well, both on PRB^2 scores and having significantly more places than expected, with Chepstow a case in point of why just using win stats is misleading with small samples. At the other end of the scale, Haydock and Newbury perform very poorly, with horses having their prep at Haydock having 0 places from 38 tries.
Cheltenham Handicap Hurdles – LTO Course when horse placed first 4
I now filtered for just horses that finished in the first four last time out, and again Thurles and Naas come out on top with the PRB^2 scores, while of the tracks with a decent sample Kempton form seems to work out very well. Once again the Haydock form just doesn’t seem to work out at Cheltenham. It’s also notable how horses who ran well at Cheltenham last time, do poorly against market expectations with only 2 places, when 5.2 would be expected. It should be noted that a PRB^2 score of 37.49 would generally have more than 2 places from 25 though, so I’d imagine there were a few close misses in that sample.
Cheltenham Handicap Hurdles – Track Direction for horses placed LTO
You would assume in big competitive fields at Cheltenham that is would be better for a horse that ran well last time, to be going in the same direction this time, but not so according to this, as horses who finished in the first four last time going right handed have a higher PRB^2 score, and also do better than market expectations with 50 places, when only 46 were expected. The win rates do slightly favour the horses who ran well left handed last time, but as I’ve already explained with a success rate of only around 5% the samples are tiny, and the place ones will be a better predictor of future win rates.
Cheltenham Handicap Hurdles – Race Type LTO
It seems horses having their prep for a Cheltenham Handicap Hurdle in either a chase or flat race, haven’t done much good, with zero wins, a very poor PRB^2 score, and between them only 3 places, when 6.25 would be expected.
Cheltenham Handicap Hurdles – Rating Limit LTO
I filtered for the rating limit for a horses last start, and it’s interesting that horses who ran in non handicaps last time, denoted with a blank above, did best. They had the highest PRB^2 score and also outperformed market expectations in both the win market with an AE of 1.23, while also having a Place AE of 1.15. Of the horses who ran in a handicap last time, there isn’t much between the different rating limits, except when we get down to 0-135 grade races, and they seemed to find the jump in class a bit much at Cheltenham.
Cheltenham Handicap Hurdles – Run Style LTO
Given the big field nature of Cheltenham Handicap Hurdles it is hardly a surprise that horses who led last time do poorly, they will often have managed to take up their preferred position at the front without too much pressure, but it’s very rare they will get the lead as easy at Cheltenham, and as such perform poorly all round.
Cheltenham Handicap Hurdles – Days Since Last Run
Horses running again within seven days have done well, I assume most have won and are lining up with a penalty. Apart from that horses that are lining up from Cheltenham having been off the track for between 57 and 150 days do very well, in both PRB^2 scores and outperforming market expectations. This makes sense really as often horses that look likely types after a promising run are put away to protect their handicap mark for the festival.
There was a few interesting nuggets when I filtered by race classification last time out. We already discovered horses having their prep in a chase don’t do well, but horses whose last start was in a grade one do exceptional well, with 10 places from 31 starts, when only 5.49 would be expected. Horses coming from a handicap do quite poorly versus the average. It’s a small sample but mares having a prep against their own sex has seen them outperform expectations, but the other significant finding is how well horses coming from a novice event do, with a good PRB^2 score, and 26 places when only 20.33 would be expected.
Cheltenham Handicap Hurdles – Horse Age
6 seems to be the optimum age performance wise based on the PRB^2 numbers, with a fall off in results from 8 year’s old and up. That said while the older age group do worse on that metric, the market seems to have over estimated the negative effect, as overall horses 8 and older have had 23 places when only 18 would be expected.
Cheltenham Handicap Hurdles – Horse Sex
We already discovered that mares having a prep against their own sex do well, and that transfers to mares on the whole, as seen by a high PRB^2 score, and also the fact they had 7 places when only 4.95 would be expected.
Cheltenham Handicap Hurdles – Jockey Claim
Experience seems to count in big field Cheltenham Handicap Hurdles, as there is a notable drop off in PRB^2 scores for jockeys with a claim, and they have just 1 win from 118 tries between them.
Cheltenham Handicap Hurdles – Owner
Not many owners fulfilled the 10 runners in the last 6 years criteria, but of those that did Gigginstown and JP McManus did very well. However the market more than accounted for JP’s runners, but underestimated the Gigginstown ones.
Cheltenham Handicap Hurdles – Jockeys
I’m not convinced of the predictive benefits of using a jockeys past results at a four day meeting, as unlike trainers when even a smallish sample can show their ability to line one up for the meeting, the jockeys good or bad stats will mostly just reflect what trainer they ride for, rather than their own ability to ride the track well, which would take a much bigger sample. The most significant result in the table above is probably how well Davy Russel did, but he will miss this years festival.
Cheltenham Handicap Chases – Trainers
On to the handicap chases now, and David Pipe, Gordon Elliott and Tom George do very well, with Tom George’s PRB^2 score and the fact he had 5 places from 19 runners a better reflection of his performance than the zero winners they produced. At the other end of the scale Colin Tizzard has a poor recent record with his handicap chasers at the meeting with just 2 places from 29 runners.
Cheltenham Handicap Chases – LTO Course
Horses having their prep for a Cheltenham handicap chase at either Fairyhouse or Gowran Park have done very well, albeit from small enough samples, while a prep run at Cheltenham seems to hold more importance for chasers. Once again, as it did for hurdles, Haydock does very poorly. I assume it must be ground related as Haydock often has very heavy ground during the winter, and horses who do well there, find the conditions at Cheltenham in March much different.
Cheltenham Handicap Chases – Course LTO for horses who placed LTO.
When I filtered just for horses who were in the first four last time, Fairyhouse falls out of the list due to not having the required sample of 10, Gowran does very well with 5 places from 13 runners, while those horses that had a good run at Cheltenham last time also do well. Kempton form, as it did for hurdlers seems to stand up well, which is surprising as the characteristics of the tracks are very different, so perhaps the reason is the races at Kempton are competitive, and as such horses who do well there are better handicapped than horses who have gone up in the weights for winning weak races elsewhere.
Cheltenham Handicap Chases – Track Direction for horse who placed LTO
Horses who placed last time out on a left handed track do slightly better than those who ran right handed, but only just.
Cheltenham Handicap Chases – Race Type LTO
When we looked at handicap hurdlers we found that horses who had their prep in a chase did poorly, but it’s the opposite for handicap chases, as horses who prepped in a hurdle do exceptionally well, with 11 places when only 4.55 was expected. It’s easy to find logic to back up the stat too, as horses who go back over hurdles having been chasing generally do so because they’ve been running poorly and going back over hurdles is an attempt to find some form rather than any long term plan. Contrast that to horses who run in a Cheltenham handicap chase having had their last start over hurdles, and the profile of those horses tends to be ones who look well handicapped over fences, and connections give them a prep over hurdles so as not to ruin their chase mark.
Cheltenham Handicap Chases – Rating Limit LTO
Horses who ran in a non handicap again do quite well with a PRB^2 score of 35.31, but unlike in the handicap hurdles, the market has these type nearly spot on, with the amount of places almost matching expected. Overall horses in 0-145 grade or lower don’t perform as well as those coming from higher grade handicaps, but the market seems to account for this, as as a group they have a similar amount of places to what was expected.
Cheltenham Handicap Chases – Run Style LTO
This is much the same as for the hurdlers, horses who led last time do poorly in handicap chases at the festival, and again you would assume the reason is the competitive nature of the races mean they find they have to work much harder to take up their preferred position, and contrary to that hold up horses who have been giving away cheap ground all season at the start, suddenly get a good pace to aim at, without having to sacrifice easy ground at the start.
Cheltenham Handicap Chases – Days Since Last Run
There really isn’t much of note for days since last run for the chasers. Those coming off the back of a 150 day or more lay off do best, but the sample is quite small, and any positive to take from it should be tempered by the poor performance of horses in the 91-150 range.
Horses coming from a grade 1 and running in a Cheltenham handicap hurdle did remarkable well, but those doing the same before running in a handicap chase do poorly. Is the difference merely variance or can it be explained? I’d say it’s a bit of both variance and maybe how the handicapper rates such performances, maybe he underestimates the grade 1 hurdlers, and not the chasers, maybe its the market that does the over/under estimation. It should be noted that like the handicap hurdlers horses who had their last start in a novice event do very well.
Cheltenham Handicap Chases – Horse Age
Performance wise the optimal ages seem to be be 6 to 8, with a noticeable drop off from 9 upwards, although the market does seem to account for most of this, and you would have only made a slight loss backing all 9+ year old horses for a place the past 6 festivals.
Cheltenham Handicap Chases – Horse Sex
Mares did well in Cheltenham handicap hurdles, but the same cannot be said for the handicap chases, albeit from a sample of just 12 having tried over the past 6 festivals.
Cheltenham Handicap Chases – Jockey Claim
Three pound claimers do well in the handicap chases, having done poorly in handicap hurdles at the meeting. I can’t think of many logical reasons for the difference so it’s possible just variance, but the average handicap hurdle at Cheltenham has a field size of 23.6 versus 21.5 for the handicap chases, so maybe that helps the inexperienced but useful up and coming jockeys. It could also be due in part to the only amateur handicap being the Kim Muir which is a chase, although the Martin Pipe is over hurdles for conditionals. No amateurs are allowed ride at this years festival. Once again five and seven pound claimers do poorly.
Cheltenham Handicap Chases – Jockeys
As I said with the hurdlers I wouldn’t place too much significance on the jockeys stats, but Davy Russell again did very well, while Jeremiah McGrath has had 6 places from just 10 rides in handicap chases at the meeting.
Cheltenham Handicap Chases – Owners
JP and Gigginstown are again a close match based on PRB^2 scores, but unlike for the handicap hurdlers, the market seems to have had both operations spot on with a similar amount of places as expected for both.
I hope you found something useful in this piece, it certainly took long enough to gather the data and put it together. Using PRB^2 and place actual versus expected, makes the smallish samples much more useful for analysis, but they are still too small for any cast iron conclusions. Like all stats though they work best when you can back up the data with a logical reason as to why it may be so, and I hope my analysis has helped a bit there too.
If you want my best bets for each day of the 2021 Cheltenham Festival, then you need to sign up to receive my Analysis and Tips package. We have had 7 winning years from 8, with a exceptional profit of +342 points, and an ROI of +47%, and that doesn't include extra advice like doing the reverse forecast on Farclas and Mr Adjudicator in the 2018 Triumph Hurdle, when even just a €10 reverse forecast paid €727, or advice to back Espoir D'Allen on the exchanges in 2019, when he won at a Betfair SP of 25/1, while last year extra winners included Sire Du Berlais at 14.5 and Champ at 6.0.