Nine years ago I wrote a blog post looking at All Weather racing statistics. The summary was…
- 12 runners or less
- Horses that Lead the pace
- Who last raced between 22 and 56 days ago
The results from this produced an average of +54 units profit per year at SP odds.
Pretty good.
The question now is, will these results still hold up?
And if not, what can we do to bring them back to profit.
Which is what I’m investigating today.
Do The All Weather Racing Statistics Hold Up?
In nine years a huge amount can change, especially with betting. Which makes our first job to determine whether the trends from all weather racing nine years ago, still hold up today.
Since 1st January 2018 the results for these bets has been:
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -1117.29 | 5426 | 816 | 15% | -21% | 15% |
Definitely not profitable to SP Odds!
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -317.17 | 5426 | 816 | 15% | -6% | 15% |
At Betfair SP we can see that it’s still not profitable, but there’s a significant improvement in ROI. These selections Betfair odds must be significantly higher to take the ROI from -21% to -6% (after commission). That’s very telling.
But the bottom line is… you don’t want to be betting on these selections any more.
Let’s Find Out What We Should Be Betting on
Now that we know that we shouldn’t be betting on these selections, we can focus on how to make them perform in such a way that maybe we can adapt them to make them profitable again.
After all, to Betfair SP they’re only losing -5%, which isn’t a huge amount to recover. They’re three selections a day (on average), but let’s begin by looking at what happens if we only use two rules at a time.
All the results going forwards will be to Betfair SP.
Less Than 13 Runners & Raced Between 22 and 56 Days Ago
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -1730.62 | 43607 | 5625 | 13% | -4% | 39% |
Leaders & Raced Between 22 and 56 Days Ago
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -809.24 | 6703 | 909 | 14% | -12% | 20% |
Leaders & Less Than 13 Runners
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -664.43 | 6704 | 998 | 15% | -10% | 19% |
In terms of ROI the first example is clearly the best, however in terms of race strike rate the first is the best.
I’m also going to take a quick look at the strongest leader in the race only, the other two factors are too broad to offer any insight individually.
Leaders Only
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -785.57 | 18352 | 2515 | 14% | -4% | 14% |
The strike rate is low, but the ROI is the same as our first example.
As always, let’s look at this logically. What happens if we look at Leaders in races that are predicted to be run Fast.
Leaders In Fast Races
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -284.19 | 4203 | 566 | 13% | -7% | 13% |
The strike rate and ROI have both dropped, which indicates that these horses being leaders is already being accounted for within their odds.
What happens if we do the same but limit the distance to a maximum of 5 furlongs?
Leaders In Fast Races Up To 5 Furlongs
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 58.28 | 206 | 26 | 13% | 28% | 13% |
Immediately we’ve made a profit over the eighteen month sample. The strike rate has stayed the same, but the ROI has jumped to 28%, which is excellent.
However, the selections have dropped significantly to around one every three days. With a strike rate of just 13% this could mean very long periods of time during downswings, making it less than ideal.
But it’s beginning to give us a direction for us to look at. Just increasing the distance to include six furlong races shows us that in the last eighteen months you’d have made…
Leaders In Fast Races Up To 6 Furlongs
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 212.16 | 785 | 117 | 15% | 27% | 15% |
A small increase in strike rate and a small decrease in ROI, but importantly we’ve increased the number of selections by over 300%.
Following the same train of thought, what happens if we consider the four horses predicted to have the fastest pace in these races…
Top 4 Leaders In Fast Races Up To 6 Furlongs
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 133.22 | 3719 | 459 | 12% | 4% | 24% |
A decrease in profit, but a huge improvement in race strike rate. We’re now winning 24% of all races that we bet in, and are making a positive return, whilst finding an average of 6 selections a day.
Taking these and betting on favourites, second favourites, third favourites and fourth favourites only produces the following.
Favourites
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 17.54 | 451 | 139 | 31% | 4% | 31% |
2nd Favourites
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 49.87 | 474 | 106 | 22% | 11% | 23% |
3rd Favourites
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| -85.71 | 476 | 56 | 12% | -18% | 12% |
4th Favourites
| Profit | Bets | Wins | Strike Rate | ROI | Race SR |
| 27.99 | 462 | 58 | 13% | 6% | 13% |
For some reason the third favourites have made a loss, and a hefty loss at that, but this could be statistical variance as the sample is not very large.
We can also see that the first and second favourites have made a profit with good strike rates.
What You Can Do With This Information
This data tells us that the market has already taken into account the horses expected to have the fastest pace in the race. Not only has it taken account of them, but it’s pushed the odds down far too far so that betting on them makes a significant loss.
However, the market is not giving enough importance to the predicted pace of those fastest horses when they’re racing in short races of six furlongs or less, which is why betting them in these races makes a small profit.
Unsurprisingly, focusing your betting on the top four in the market significantly improves the strike rate, but drops the number of selections in half.
With 24% of winners in six furlong and less All Weather races coming from the four horses predicted to have the fastest pace, this gives you a very strong starting point. Especially knowing that these horses make a small 4% ROI to Betfair SP.
By shortlisting these horses, you can now look to confirm whether they’re in form, have performed previously on All Weather and if they’re suited to the distance.
Just these three pieces of information will enable you to remove the horses which are unlikely to win, but if you’re a member of the Race Advisor Pro Members Club, you can also consider the DSLGR and average competitive speed rating (ACSPCLTD) then you’ll be well on your way to adding another profitable strategy to your portfolio.
Are you a fan of All Weather racing, or do you prefer flat turf and jumps?
Do you already use pace in your betting, or will you start to include it now?
Please let me know by leaving me a comment below.
Great Leighs!?! Methinks this article needs bringing up to date.
Hi Jack, thank you for the comment. This article was originally published in 2010. I have made a note to do a re-write of it 🙂
Whoa there soldier
Due to reopen if all goes to plan, currently used by many stables for training
AFAIK to be renamed Chelmsford
You mention back-fitting of your analysis of AW data above.
I know that it is said that many systems are built in this way and are likely not to work.
What I do not understand is what the alternative is, assuming that one is going to review historic data in some way to see if an idea might lead to making a profit.
A great question Ralph, and you’re right there is in fact no other way to do it. By definition all forms of system building are a form of backfitting. A better phrase is probably overfitting. You need to make sure you aren’t overfitting the data. This means that when you build your system you aren’t creating rules based specifically on the dataset you are using, because if you do then it won’t work when you make it live. There are a few ways to do this. One way is to make sure that you’re rules make sense from a racing perspective rather than just applying them purely because of the results. The best way is an approach called bootstrapping. This is where we take a sample of data and build the first rule on it. We then take a new random sample of data and build our second rule on it etc… each of these random samples should choose horses not races (unless you’re building a selection process to pick multiple runners in a race) which keeps them as individual events. Then when your system is built you test it on completely unseen data.
Now that can be quite a job, a simpler way is to always check your system against unseen data after you’ve finished making it to confirm that the results are similar to the ones you achieved while building the system.
Michael, thats the best explanation of system building ive ever come across, thankyou, ive created many like a lot of us presumably have but have never considered using unseen data to verify, ive simply chipped away with rules to create what looks like a profitable angle/system only to see many yo yo,
Thank you George. Unseen data to verify is a very important step 🙂
Any chance of this one being updated?
Hi Kenny, it’s on my list to do in the near future.
Was there any reason you didn’t look at 7 furlong races as well, as that would make it all sprint races?
For me seven furlongs is pushing the length of a sprint race and you start getting some stamina coming into play in them, dependent on track. However they are still considered sprints in the UK. In some countries they’d be considered positively long distance 😀
The back-fitting comment is a good one. I too can’t see how you can avoid referring to previous races to sharpen any betting approach. For me one of Michael’s comments elsewhere provides a key point of difference and that is about keeping any approach simple. To me the worst cases of back-fitting are where people keep adding rules until they generate a system that makes a profit. At this stage you are reducing your sample size to the point where replication is unlikely. Keeping it simple I think avoids the some of the worst aspects of what we think of as back-fitting.
hi Michael, what is the difference between strike rate and race strike rate?
The strike rate is the percentage of winners from all selections. The race strike rate is only relevant if you have more than one selection in a race, but if you do then this is the percentage of races that you found the winner in.