Rating changes
Question: I generally do the form soon after acceptances to see if I can nab an early price or two. Do you guys ever tweak your ratings after acceptances? If so, when are the ratings' locked in '?
TRB Response: In rare cases, a rating could change after the first file production. If horses are on a quick backup from their last start, we need the verified times and individual sectionals to do our meeting reviews properly and finalise the WPRs for each race. They can take 2-4 days to come in after a meeting. Metros come faster than country meetings.
Once we have those, the review is done, and final ratings are determined.
Even if a horse backs up in 7 days, our normal timing will still have the final rating in the first file produced for 95%+ of these horses. However, if there is a slight delay to sectionals for whatever reason or a horse is racing on a Monday or Tuesday, and it's 6-7 days after their last run... and those files are first run on a Thursday, then it's impossible to have the final rating in the first cut as that file run is only 2-3 days after the race. They will always be up to date in the morning re-run, which is done the day before each meeting.
Question: A horse called Cornforth is declared in races 2 and 5 at Rockhampton for Friday. The WPR figure given for its last start win on the 29th of May shows up as 85 in race 2 and yet 83.8 in race 5. How can a horse have two different ratings for the same performance, and which one is correct?
TRB Response: When we do our WFA Performance Ratings, the figure assigned to each horse is normalised to WFA, i.e. as if they carried a weight equal to WFA for their age/sex/distance/time of year. That means performances and horses can be directly compared to understand the relative quality between horses on comparable terms.
At the weights, a 4YO Gelding with 54kg a horse might run a 99 rating which wins the race, but when normalised to WFA, the horse would end up with something like a 95 rating.
When that horse comes to race again, his last start is a 95 at WFA, but then we must consider the weight he is allocated for today's race. Weight below WFA will increase his potential performance (at the weights), while a weight above WFA will reduce it from that number.
This is how the numbers are shown in the form of GTX, the performance the horse could achieve today (based on the last start, 2nd last start or whatever) at the weight it will carry today.
In the case of Cornforth, he ran an 83.1 rating last start, normalised to WFA. In race 2, he has 56kg, which elevates his potential performance at that weight to 85.0. In race 5 he has 57.5kg, which elevates his performance from 83.1 to 83.8 at the weight to be carried.
So, in essence, both ratings are correct... it's just a case of different races/conditions.
Question: Why are the historical WPRs different for the same horse for the SAME race when you step back in time?
TRB Response: The numbers you see on the screen against each run represent the horses' WPR normalised to WFA, then adjusted for the weight to be carried in today's race. So they will change depending on the weight the horse has in the upcoming race.
For example, if a horse has a last start rating of 100 normalised to WFA (which sits in our main database) and then comes into a race carrying WFA of, say 59kg, then the LS figure in GTX will show 100. If it went into a race with 54kg or 5kg below WFA, the figure would show something like 104, reflecting the benefit of weight below WFA. The figures against each run in GTX show the horse's forecast performance "at the weights" in the upcoming race. This means you can directly compare the figures of individual horses as you see them on the screen without doing any further adjustments.
The ratings adjustment for an upcoming race is more of an art form rather than specific science with fixed rules.
The idea is that you are trying to forecast the performance the horse is likely to run to today based on all of its past performances and other factors like the suitability of the race, fitness, track, going, distance, pace, likely position in run etc.
If the automatic rating in GTX looks okay, then just leave it. You don't need to split hairs... it doesn't add any value to your betting decisions. In my experience, the best betting decisions are the clearest when you do a simple review of each race/ratings. If you have to start splitting between 0.5 points here and there based on assumptions and uncertainty, then it's a good sign there's no clear betting opportunity.
When forecasting performance, you should consider the concept of regression to the mean. Each horse has a given level of talent, and over time, their form is made up of runs that are peaks beyond their general level of talent and runs well below.
A horse that has typically rated 88-89 in its good runs and suddenly rates 93 shouldn't necessarily be expected to rate 93 next start, especially if it's well exposed in terms of the number of career starts and runs this preparation. The horse is more likely to regress next start, perhaps not all the way to 88-89, but still lower than 93. Think of it like a cricketer that might have an average of 40.00, then comes out one inning and scores 100. If you are rating the expected performance of that cricketer next time, you intuitively wouldn't expect him to get 100 again.
The same applies if the horse suddenly rates 75. There are many reasons why horses rate lower, and you can typically expect them to progress more towards their median level of form next start. If a horse has 88-89 figures and then suddenly rates 75, I would perhaps forecast it to rate 86-87 next start.
When it comes to younger, lightly raced horses, then you have to allow for the prospects of improvement. A sudden spike can be trusted a little more because the horse is likely still improving.
As I said, there are no fixed rules, it really is an art form that you can develop over time. The ultimate test is to ask yourself "in the context of this horse's history and recent form, keeping in mind regression to the mean, does the figure I've allocated reflect a level of performance that is reasonable to expect from this horse?" The purpose of adjusting ratings is simply to help identify good betting prospects so always keep that in mind. Don't get caught up in the mechanics of tinkering with ratings and forget the actual purpose. If a relatively simple process you follow doesn't highlight any clear betting opportunities then it's typically best to move on to another race. There are plenty of opportunities that present themselves without spending a large amount of time on individual races.
TRB Comment:
WFA Performance Ratings are initially calculated and normalised to WFA. They're the figures we quote in reports and commentary. That's because we are looking post-race and comparing horses on an even playing field. The actual WFA Performance Rating for each of a horse's past runs is the figure the horse ran as if it carried WFA in the race. So an 89 rating for a 5YO male horse over 1400m is the rating achieved assuming 59kg.
When they are entered in a future race though, they are assigned a weight to be carried. At this point, the rating and weight carried become relevant to forecasting performance and comparing horses within the race. If that horse then goes into a race and is set to carry only 54kg, then using that previous run, his actual potential rating is 89 + the allowance for carrying 5kg under WFA. So in this case GTX would show 93 against the last start. (the horse normalised WFA number 89, adjusted for the weight to be carried in today's race.) If the horse was to carry 60kg which is +1kg on WFA then his 89 rating would be reduced in GTX to show 88.2.... which is the level of performance he could be expected to achieve at the weight to be carried today.
The bottom line is that you don't need to worry about the weight carried when you view the ratings in GTX. The ratings shown on the screen already take into account the weight to be carried today. In a WFA race, there is no adjustment, so the numbers you see on the screen are equal to the normalised rating.
Question: I am wanting to get more from the WPRs and have been tweaking them to make them more predictive in today's race.
However, a couple of questions if I may:
1. In adjusting the WPR for today's race, will the new WPR follow through to the horse's next future race? That is, is it like Wintune where fine-tuning a rating will hopefully lead to better future prediction.
2. Why are the historical WPRs different for the same horse for the SAME race when you step back in time? For example, in today's race, a horse has a set of prior ratings, but go back to the last run and the lead-up ratings (for the same dates) are different.
Are you able to share your technique(s) for adjusting the ratings? The User Guide you prepared gives a good overview without being too specific.
To give you some background my approach is like this:
Looking at the WPR Rating Graph, then adjust the Base Rating up or down to reflect where I think it should be, based on:
- current price (i.e. if the rated price is too low then reduce the rating)
- recent past performance ( win/close to win/same class = most likely no adjustment but reduce rating when opposite is true)
Overall trying to keep the rating to a mean trend line rather than sudden ups or downs. Your comments would be appreciated.
TRB Response: There is no Win-Tuning for WPRs, it defeats the purpose of relying on our process of developing the ratings in the first place. If users wish to develop their own set of customised ratings then others like Trakform, Trakline or Formline are available to tinker with.
When using GTX for an upcoming race, any changes to the base rating of a horse are affecting your forecast assessment for today's race, not the merit of past runs.
Question: Could you confirm that an apprentice jockeys allowance has been accounted for in the horse's ratings? If it's been accounted for, how is the rating modified by the allowance amount, e.g. 86 becomes 89 if the apprentice claims 3kg? or what is the method?
TRB Response: The WPR ratings a user sees on the screen in the horse's form, including the base rating, is already adjusted for weight to be carried, including any apprentice claims. If a horse has no jockey and the late file adds an apprentice jockey, the ratings will change to reflect the claim.