Race Predictors… Do They Work?

The chances are you’ve seen the race predictors on the Racing Post, At The Races and Racing UK websites.

But have you ever wondered how they’re predicting a race, whether they work and what is different between them?

I certainly have, and today I’ve written up my thoughts on them.

It’s important for me to clarify that the following are my personal thoughts. They are not facts, they’re assumptions based on what I’ve seen.

My first assumption is that the race predictors are all provided by the same company or use a very similar method for analysis.

Each work in slightly a different way, here are screenshots of each of them…


The Racing Post

Screen Shot 2016-01-25 at 11.02.59


At The Races

Screen Shot 2016-01-25 at 11.06.13


Racing UK

Screen Shot 2016-01-25 at 11.09.06


As you can see, the layout is slightly different in each case but the basic principle is the same. There is an icon of a horse and the further to the right the better the horse is predicted to perform.

There are two major differences between the predictors. The Racing Post provide parameters which you can adjust, At The Races give each horse a score and Racing UK provide a prediction based on Win, Place, Going, Jockey/Trainer.

But are they effective…

I need to be honest and say that I haven’t tracked their performance for long enough to give a statistical answer. So I’m not going to. If you’d be interested in doing this and reporting on your results then please get in touch with me.

However, I would immediately never consider using Racing UK’s race predictor. And that’s for one simple reason. There is no control over it and there is no score. This means that you can’t make an assessment on how likely the horse is to win, how close the competition is or change the settings so you know what it’s doing. In terms of it’s display, I find it much harder to compare the results to the other runners due to their layout.

I like the Racing Post and At The Races predictors for different reasons. If you use either of these predictors, like anything, you should use it as a tool for part of your overall analysis. You shouldn’t use it as your only method of analysis.

Here is how I would go about using them…

I would always begin with the At The Races because this doesn’t have settings and provides a score. But most importantly because they have a data quality bar.

This data quality bar gives us an indication of how much data they’ve used in order to calculate the prediction. The higher the better, but I would be concerned about using anything less than 50%.

If one website has got low quality data for a race then you can assume the other sites do as well. That means this can be used as an immediate indicator as to whether you want to analyse the race or not.

You may want to start by only considering races that have a data quality of 75% or higher to make sure you’re only focusing on the races that contain the best information.

Once you’ve determined if you’re going to look at a race in more detail, then you need to check the…


At The Races Scores

Screen Shot 2016-01-25 at 11.10.53


The scores are important because it shows us immediately that they’re predicting Sizzler to be in the lead by quite a way at 489 and next are Paddys Motorbike at 421 and Cai Shen at 411. If we use a 10% possible variance around the scores then this means that the horses could actually have scores of…


Sizzler: 440 – 538

Paddys Motorbike: 379 – 463

Cai Shen: 370 – 452


This probably changes your opinion of how far ahead Sizzler is. You can see that if Sizzler has a bad race and either Paddys Motorbike or Cai Shen have a good race then they could beat Sizzler.

It’s always important to remember that scores and ratings are estimates of a horse’s performance. The true score could be a bit higher or lower.

If you kept going in order of highest score to lowest you will eventually see something similar to below…


Sizzler: 440 – 538

Paddys Motorbike: 379 – 463

Cai Shen: 370 – 452

Stencive: 328 – 400


Look at Stencive above and you will see that his highest likely score of 400 is below the best horse’s (Sizzler) lowest likely score of 440.

That means that even if this runner performed at his best he would be unlikely to beat the best runner in the race even if that runner was running at his worst, based on these ratings. When you see that there is no point in looking at other runners if you trust the ratings.

Once you’ve done that you will have your shortlist of runners, in this example Sizzler, Paddys Motorbike and Cai Shen, and you can move on to…


The Racing Post’s Predictor Settings

Screen Shot 2016-01-25 at 11.13.21


You will see in the bottom right a button called Adjust Parameters. Click on this and you will see…

Screen Shot 2016-01-25 at 11.15.14

Looking above you can see that on the Racing Post’s race predictor you have the option to apply different amounts of importance to different factors. By default all factors have the same weighting of 3, but we can use the race conditions to adjust these.

This race is an all-weather handicap race over 1 mile 4 furlongs with 9 runners.

We know that in these types of races the Going and Distance is going to be quite important, so I would raise those by one. The course and draw are less important so I would reduce those by 1.

Personally I don’t like using trainer form as a big indicator in performance of a horse, so I would reduce that by 2.

That leaves Ability and Recent Form to change the weighting. Recent Form is always important so I would put that and Ability up by two. These changes would leave the settings like this…

Screen Shot 2016-01-25 at 11.18.26

Once you’ve changed your settings hit the ‘Re-calculate prediction’ button to see the changes in the results.

Screen Shot 2016-01-25 at 11.18.52

Based on this there is only one runner from our competitors on the At The Races predictor and that is Paddys Motorbike who has been put likely to come second.

This would now be the horse I would focus my analysis on to see if there are any other runners that are likely to be able to seriously compete with this runner and that would determine whether a bet was placed and if so what type of bet and stake.

Using this approach can turn these race predictors into a very useful analysis tool to provide you with a shortlist of runners to focus your analysis on. I would use them in combination with each other as above, making use of each of the tools advantages and then analyse the shortlist further to determine whether you want to place a bet.

Michael Wilding

Michael started the Race Advisor in 2009 to help bettors become long-term profitable. After writing hundreds of articles I started to build software that contained my personal ratings. The Race Advisor has more factors for UK horse racing than any other site, and we pride ourselves on creating tools and strategies that are unique, and allow you to make a long-term profit without the need for tipsters. You can also check out my personal blog or my personal Instagram account.


  1. Hi Michael

    Extremely interesting article..thanks

    As I said to Eddie in a recent email, the products you offer plus the regular articles and the standard of customer service are exceptional. Well done

    Mike West (Lancashire)

  2. I have been using the Racing UK predictor for some time and I noticed that the ratings are not consistent from overnight rating and the rating on the race day. I have not been able to use the predictor with any confidence of selections.

    1. Thank you for this info Fred. That’s something most of the big sites do. They create ratings the night before and then adjust them when the early morning updates from the bookies come in. As you say this can make them
      unpredictable if you can’t always use them mid-morning on the day of racing.

  3. Hi Michael

    As you haven’t followed the Predictors in detail I thought I would do one of them for you. I chose ATR as it is the (to me) most comprehensive.

    There were 19 races today in which the Predictor found the winner 16 times. A creditable Strike Rate of 84%. Unfortunately there were 74 horses to chose from. The actual strike rate being 21%.

    At SP the winnings amounted to 44.93 points. Betting a point as stake this would have given you a loss of 10.07 points.

    This how the winners were placed in the race layout:

    Ist place: 10/11 11/10 2/1 2/1 2/1 5 wins
    2nd place: 4/5 9/4 8/1 5/2 4 wins
    3rd place: 4/1 7/1 9/4 3 wins
    Last place: 11/10 11/4 11/2 5/6 4 wins

    You did state that the Predictor should be used as an extra aid to ones own selections. In this regard I found the predictions to be helpful.

    I used the predictions from my site (there are three per race) and put them through my procedure to select the value bets only. From 57 horses selected I had 24 bets and made a profit of 20.84 points (BSP net).

    As a second stream I chose the value bets from my selections and if they were selected by the Predictor and placed a bet on them. There were 18 bets in all of which 8 were winners giving me a profit overall of 17.16 points to SP or 18.92 points to BSP net.

    Although the profit using the Predictor was less I found the exercise worthwhile as there were fewer bets and I know that Mondays are my best days of the week for profit so I expect better results using the Predictor than I would by using all my value selections on other days.

    Your suggestion of an increase of 10% and decrease of 10% is only useful after the event when you know the horses finishing position and you can establish whether it ran to the Predictors form or not. Regrettably not as a method of chosing a winner other than by using it as an additional filter for ones selections.


    1. A great analysis of today’s racing, thank you for posting 🙂

      With the 10% either way, I use this pre-race to see the overlap of runners on all factors I use. You do need to have a knowledge of the factors performance first, as you mentioned, but I find it a crucial element in our models. 10% is where I start but some are a lot lower, and a few are higher. It’s a constantly refined amount as we get more data on a factor.

      1. Hi Michael

        Sorry, forgot to mention the percentages of the winning and losing races.

        Percentages of winning races were:
        23, 40, 57, 58, 68, 71, 72, 73, 74 (3), 75 (2), 81, 87

        Losing percentages:
        70, 75, 79

        I understand your desire to refine your performance by the introduction of extra factors, etc. Through experience I have found that an increase in factors leads to an increase in effort with no increase in performance.

        I once had a system that had in excess of 90 factors used to determine my selection. As this was all contained in one program there was no increase in effort other than the time to write the system. It was no more successful than the method I now employ which has only six factors and uses mathematical probability. I regularly have a strike rate nearing 50% a month.

        I have a data base on favourites from which I can glean numerous statistics. Unfortunately I can’t tell from them when a favourite will win a race. There are some factors that just defy logic.

        1. I completely agree regarding lots of factors. I take lots of factors and put them into primary categories. Then I find those which independently indicate a significant relationship with the outcome. Then I combine factors within the same category so that there is no crossover in factors. This reduces me to a similar number as you. That’s the quick version 🙂

          1. Hi

            Having studied yesterdays results in more depth I found that I was able to turn a loss of 10.07 points into a profit of 5.35 points SP. This change came about when I re-read the article again and noted the remarks on 10%.

            Assuming the top rated horse runs below par (ie -10%) if any of the remaining horses had a +10% score below that figure they were discounted. This reduced the number of bets by 24. Within the 24 horses were two winners at 11/4 and 5/6 which only confirmed the method used was the correct one.

            Today there have been 47 selections with 10 exclusions making 37 bets in all. The Strike Rate today was 58% as there were 7 winners from the 12 races.

            The winners made a profit of 29.42 points (SP) giving a loss of 8.58 points on the day.

            The result may seem disappointing but I do believe there is merit in its usage. I have studied selections by other sites and note that they also have had poor returns as I have. Not a good day all round.

  4. Hi Michael

    Report on day 3 of the in-depth monitoring of the ATR Predictor.

    A Strike Rate of 66% from 32 races.

    31 selections excluded (included 3 winners at 4/1 8/1 and 6/1 again showing the validity of excluding)

    Total selections: 97 Total wins: 21 Total profit: 65.80 points (SP)

    Final loss: 10.20 points

    Finally I compare my selections with the Predictor. My strike rate was 23% using my value selections but I only made a loss of 3.56 points net (BSP). However, the overall strike rate my selections achieved was 60% not too dissimilar to that of the Predictor.


  5. Hi Michael

    Todays results for the ATR Predictor follow the same pattern as previous days.

    A Strike Rate of 62% from 34 races.

    35 selections excluded (included 4 winners at 6/1 9/4 4/1 and 15/8).

    Total selections: 104 Total wins: 19 Total profit: 76.49 points (SP)

    Profit (at last): 3.40 points (SP)

    After the three days of monitoring these selections I can say with some degree of certainty that the ATR Race Predictor has an impressive strike rate but it is geared towards short-priced horses which negates the Strike Rate significantly.

    There have been in excess of 200 possible bets and at the end of the day a loss was of around 17 points was made.

    In summation – the answer to the question “Race predictors – do they work?” as far as the ATR Predictor is concerned is NO. The only pattern within the selections is that in which you exclude selections which have a score of 10% plus which is lower that the top rated horses -10% score.

    Incidentally I have had a terrible day at Chelmsford. I couldn’t find a winner until the last race. Teversham won at 66/1 and Betfair paid 128.54.


  6. Michael, another interesting read and thanks again for all your information!

    On another note you have just scuppered my lifetime’s work. You might remember in my ‘Four Corners’ piece that I stated trainer form is more valuable than the sum of the other parts! You have just said: “Personally I don’t like using trainer form as a big indicator in performance of a horse, so I would reduce that by 2.” Yikes!

    Wonder what the results would be with trainer form on maximum setting?

    Love your work and thanks again…

    The Racing Horse

    1. Thanks Paul. I should clarify, that’s my personal thing because I’ve never been able to get an affinity with trainers. Eddie (and others) would argue completely the opposite. Not sure what the performance would be with a trainer form on max, one to test 🙂

Back to top button