Analyzing specific races with the help of ten year trends is a common practice, but they’re not articles I’d recommend paying much heed to, as they provide little, if any predictive value. The problem being the sample size is just too small, and most of the time their authors compound that fact, by only using winners to form their conclusions. A far better way is to group similar races, like say decent two mile handicap hurdles together, and analyze them, rather than picking a specific race, as it’s unlikely you’ll find anything other than noise in the data. With that in mind I’ve decided to take a look at Cheltenham 2016 from a broader perspective to see if I can find any useful stats, or potential punting angles we can use. For this experiment I will be using Proform Software, which to me knowledge is easily the best software out there for this type of analysis.
Lets start with trainers. For all the stats I’m using the last eight festivals. That’s big enough to give us a decent sample, but not too far back as to be irrelevant, so I’ll use the 2008 festival as the starting point. I’m including data from all races run at the Cheltenham Festival since then, and I’ve filtered to only include any trainer with at least 8 places.
Most of the fields are self explanatory. The EX Wins is expected wins based on the Betfair Win SP, and the A/E is actual winners divided by Expected. A figure over one indicates the horses out performed market expectations, and is a much better guide than using level stakes profit and loss. The Chi Scr field is the Chi Square statistic based on the actual versus expected winners, and is a test of the significance of the result. Anything over 2.0 would be reasonably significant. Proform doesn’t yet have Betfair Place SPs included which is a pity, as unless you have a really big sample size, then a good A/E figure to place sp is a better predictor of future profit for both wins and places. You can export the data to excel though, and I do this to come up with my own place expected figures based on the win price. You can actually import historical Betfair Place SPs into Proform too, but getting results takes longer this way, and my method is accurate enough for these purposes.
A few things to note from the results with this years festival in mind. Henry De Bromhead and Gordon Elliot both have decent A/E figures, and good place strike rates, indicating they are selective with the horses they send to the meeting. Gordon Elliott’s 1.39 A/E means if you backed all his horses over the past eight festivals to return a set amount, your ROI would be 39%. Not bad at all, but as I already indicated the Place A/E is more robust, and on this measure he does very well too. His place A/E is 1.53, with 19 places when only 12.36 would be expected. Henry also does quite well on this measure with a place A/E of 1.34. Given the sample sizes Gordon’s is much more significant though. Willie Mullins Win A/E is a healthy 1.25, but a note of caution is, that based on my calculations he’d be just break even backing all his horses to place. A trainer who does poorly is Paul Nicholls. His Win A/E is 0.86, while his Place A/E is just 0.82, and given the big sample that’s pretty significant, and suggests his horses are over-bet at the meeting.
The next screenshot is when I filter using the same criteria, except this time I only include hurdle races. This provides two very interesting points. Firstly Paul Nicholls has a Win A/E of 1.63, and a fairly significant Place A/E of 1.27. On their own they’d be of interest, but when you consider his overall record is poor, that suggests he must have a terrible record with his chasers, but we’ll come to that next. Once again Willie Mullins Win A/E is decent, and the Chi Square statistic indicates that’s quite significant too. Again though his Place A/E is around break even. The other significant finding is with David Pipe, he does terrible with his hurdlers, just 2 winners from 102 runners, when 6.22 would be expected given their prices, and his Place A/E of just 0.57 means it’s not just a case of him getting unlucky. I’d think twice about supported his hurdlers next week.
Onto the chasers now and Gordon Elliot backs up a Win A/E of 1.4 with 8 places in total, when only 4.8 would be expected. Henry De Bromhead also does well on both figures. Based on the Chi Square stats there is two very significant findings based on winners. Jonjo O’Neill has had 9 chase winners when only 4.24 would be expected, but his Place A/E of 1.11 casts a doubt on how likely that is to continue, as he may have just had an unsustainable amount of wins versus horses hitting the frame. Paul Nicholls though, as expected from the earlier figures has an atrocious record over fences at the meeting. He has had just 6 winners, when 14.35 would be expected, but even more damming is his place stats of just 20 placed, when 42.5 would have been expected given the horses prices. The chances of that happening by chance alone is almost zero, and he’s not a trainer I’d recommend backing with his chasers next week.
You’re getting into sample size issues when you start filtering for trainers in just handicaps, and while I did it, I didn’t find anything extra of note. The screenshot above is for handicap hurdles, and includes only horses who had their last start in the UK. The large majority of these horses will thus be UK trained, and it shows a negative Win A/E of 0.89.
I then did the same, except this time for horses who had their last start in Ireland, and it shows a profitable A/E of 1.27, backed up by a much higher placed strike rate of 22.6%. This is interesting as Irish horses tend to have higher ratings in the UK, and while many Irish trainers have over the years complained about their treatment in the UK, they clearly don’t have much of a case, and the Irish trained hurdlers are my first port of call when looking at the handicap hurdles at the festival.
I’ve now done the same for handicap chasers, and the screenshot above shows the returns for horse who had their last start in the UK. The Win A/E shows a slight profit, but the place one is fractionally under par.
On first glance horses who had their last start in Ireland before taking in a festival handicap chase, haven’t done so well, with just 7 winners, when 9.28 would be expected, but note the amount of total places, 43. This indicates a lot more than expected hit the frame without winning, and for future prediction purposes, I’d expect the Irish win strike rate to improve.
I read an article last week that was encouraging punters not to forget about the Willie Mullins trained horses not ridden by Ruby Walsh, and I know I’ve backed a few of them myself in the past,so above is Willie Mullins overall stats for the last eight years, and below is his stats when Ruby was riding.
First thing to note is Ruby Walsh’s rides for Willie Mullins certainly haven’t been over-bet historically, with a very decent Win A/E of 1.38. From the two screenshots we can work out the Win A/E for Willie horses that Ruby didn’t ride. Ruby has ridden 20 winners, which leaves 11 he didn’t ride, the overall expected winners is 24.9, Ruby’s is 14.5, which means the expected winners for horse not ridden by Ruby is 10.4. That gives us a Win A/E of 11/10.4=1.057, which while a slight profit, wouldn’t make you want to specifically target those horses.
The two screenshots above are for handicap hurdles, and are split according to weight carried. The first one is for horses who carried more than eleven stone, and shows a slight loss with a Win A/E of 0.94, while a fractional profit would have been made backing all the horses carrying eleven stone or less. The higher weighted horses do better strike rate wise, but the market seems to slightly overestimate that effect.
I’ve now done the same for handicap chases, with the first screenshot for horses carrying more than eleven stone. They again have a fractionally higher place strike rate, but a lower win one, and the market does seem to overestimate the higher weighted horses, with a Win A/E of 0.82 for them, while the lower ones are 1.15. The sample size isn’t enough for any significant findings, but it is worth bearing in mind when studying the handicaps next week.
The above shot is when I filtered for course direction last time out, and it’s mildly interesting, with horses who raced left handed last time, doing better with both strike rate, and with A/E figures. This makes sense as if a horse is used to going right handed it might find it hard to get into a rhythm in a fast run race left handed. I can’t filter for it in Proform, but it is likely these results would be more pronounced if I could filter for say horses who had their last two or even three races right handed, and I would take it as a negative for horses who had.
I don’t think jockey/track stats have much predictive ability, but I ran them all the same. The only really significant finding was Davy Russell had 13 winners, when only 6 would have been expected, but his place A/E of just 1.17 puts a bit of a dampener on it as an angle for the future. I hope you found something useful, and please check back next week for my tips for each day of the festival.