• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a 20% discount on Inform Racing.
    Simply enter the coupon code ukbettingform when subscribing here.
    We have a lot of members who are existing users of Inform Racing so help is always available if needed.
    Best Wishes
    AR
  • Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

Approaching speed figures

O Outlander are you using a going allowance per race as opposed to the whole card?

Edit I think what you've said is that this one race ended up informing your going allowance, meaning a very fast track against which all other races looked like a slow plod.

It seems justifable to use two going allowances for that meeting.
Yes what you have wrote in your edit, is the situation with my sheets
 
Here's what we've been talking about in practice. The first four home have an insane rating, decent horses as they are it is implausible that the best 4 performances (5 of the top 20) over the course of nine years all came in one race.

1621010687704.png

So I've looked at a formula which will identify outlier races and where that occurs, it will re-assess the going rating for that race and it will apply to the race in question. Its quite complicated, but it doesn't actually identify a large number of such races, and it has the reverse effect of upgrading some performances too.

So following rating adjustments, this is the top 20 for Dundalk, spot the differences now

1621010855812.png

And here is the original list with a new rating down the right hand side. It's pleasing that my -33 is actually just 1 different from that TheBluesBrother TheBluesBrother identified.

1621010970922.png

Another race that looked wrong was this one below, it seems to have over-rated quite a number of runners so the performances of these are adjusted down to the tune of 22lb.

1621014848090.png

This race also read as an outlier, but in the opposite direction, the actual figure not flattering the performances at all, and it has upgraded the performances from this race. So in reverse order, the race was won by Queen of the Sand and its rating of 38 was considered to more realistically merit a 54. The remaining races at that meeting received the normal going allowance.

1621014741120.png
In this race, a three runner one, it did not figure in the going allowances for the meeting and it meant this race received its own upgrade of 40lb. On my figures, Panama Hat had previous run to 73 on 19/08/2014, so its not a ridiculous proposition.

1621015156018.png

So the races that are isolated are those where most horses have made no contribution to the actual going allowance for the meeting, as their Z-Scores sit outside -1.5/+1.5. It only amounts to around 300 horses of the 30 odd thousand. Some might say its criminal to not take the face value rating but if there's a way of getting rid of Primo Oumo (down to 65) from the top of the list and most other ratings remain intact it might not be a bad thing.

This is bearing in mind I am experimenting - still.
 

Attachments

  • 1621010955139.png
    1621010955139.png
    26 KB · Views: 3
  • 1621014653997.png
    1621014653997.png
    25 KB · Views: 4
Last edited:
The other point to all this is that Panama Hat's performance is not actually likely to have had anything to do with the going.

It would probably have been a tactical slowly run race with a quick finish, leading to a relatively modest time.

It will be interesting to take this sheet back to Lingfield and look at the 2015 AW Championship performance of Tryster. It won't be much of a shock if that race ends up in an outlier category, as it was run in a desperately slow time but I remember he ran the last 2 furlongs in something like 21 seconds that day, and its probable most horses bare time was actually well below their capabilities. I'm going to guess I end up adding a fair few lbs to his rating for that one.

It can also be noted that in the revised list, that doughty old camapaigner and class act Gordon Lord Byron makes an appearance. He actually earned a 37 on base rating but with the upgrading I mentioned above, his performance actually earned a 78. His time was 1.73 seconds slower than the winning time of the previous race, but I ended up give the winner of that previous race, Sophie's Echo, a 49.

The two horses GLB beat home in that win were Clancy Avenue (up from a 5 to a 46) and Battleroftheboyne (his 18 goes up to 59). Again, a three runner race so surely one where it may not have been truly run. No reason why a marker can't go against those figures. I'd rather put the marker against the adjusted figure than the unadjusted one because the adjusted figure just seems more plausible to me.
 
Last edited:
Here's what we've been talking about in practice. The first four home have an insane rating, decent horses as they are it is implausible that the best 4 performances (5 of the top 20) over the course of nine years all came in one race.

View attachment 100449

So I've looked at a formula which will identify outlier races and where that occurs, it will re-assess the going rating for that race and it will apply to the race in question. Its quite complicated, but it doesn't actually identify a large number of such races, and it has the reverse effect of upgrading some performances too.

So following rating adjustments, this is the top 20 for Dundalk, spot the differences now

View attachment 100450

And here is the original list with a new rating down the right hand side. It's pleasing that my -33 is actually just 1 different from that TheBluesBrother TheBluesBrother identified.

View attachment 100452

Another race that looked wrong was this one below, it seems to have over-rated quite a number of runners so the performances of these are adjusted down to the tune of 22lb.

View attachment 100455

This race also read as an outlier, but in the opposite direction, the actual figure not flattering the performances at all, and it has upgraded the performances from this race. So in reverse order, the race was won by Queen of the Sand and its rating of 38 was considered to more realistically merit a 54. The remaining races at that meeting received the normal going allowance.

View attachment 100454
In this race, a three runner one, it did not figure in the going allowances for the meeting and it meant this race received its own upgrade of 40lb. On my figures, Panama Hat had previous run to 73 on 19/08/2014, so its not a ridiculous proposition.

View attachment 100456

So the races that are isolated are those where most horses have made no contribution to the actual going allowance for the meeting, as their Z-Scores sit outside -1.5/+1.5. It only amounts to around 300 horses of the 30 odd thousand. Some might say its criminal to not take the face value rating but if there's a way of getting rid of Primo Oumo (down to 65) from the top of the list and most other ratings remain intact it might not be a bad thing.

This is bearing in mind I am experimenting - still.
AustinDillon75 AustinDillon75 This sounds excellent solution to fast outliers that can't be real for whatever reason( time or distance errors, wind or going anomalies) and reducing rating to a more sensible level. I don't understand doing it the other way, if you are creating speed rating why upgrade a slow performance ?. if you do this you are moving into form rating territory .
 
AustinDillon75 AustinDillon75 This sounds excellent solution to fast outliers that can't be real for whatever reason( time or distance errors, wind or going anomalies) and reducing rating to a more sensible level. I don't understand doing it the other way, if you are creating speed rating why upgrade a slow performance ?. if you do this you are moving into form rating territory .
Yes all understood. That said if its possible to give horses downgrades or upgrades why not have a column that gives the rating on the ground but put something next door to it that offers an indication that it might have been in a false run contest, or on the Primo Oumo side of the coin there's some kind of bizarre lack of logic to the performances. You always read of Timeform offering sectional upgrades to certain performances, so they are giving ratings that advance on the actual final time.

How much of an issue would it be to say Gordon Lord Byron managed a speed figure of 37, but the maths implies in a truly run race he might have been good enough for a 78? It's a fascinating point of debate.

An initial look at the Tryster race I talk about suggests he got a baseline figure of 25 but the upgrade moves him onto a figure of 68. In his previous track victory he achieved an unadjusted figure of 83 so again we aren't generating numbers from a bingo machine - they are still realistic appraisals.
 
Last edited:
At the risk of number blindness and general tedium this is the Lingfield update following amendments and calibration with the Dundalk sheet. There shouldn't be that much surprise in the best horses being 10lb or so higher as they are generally graded level horses running Class 2 races in the AW Championships. It accounts for why there isn't anything for 2020 in here, the best performance being (one for O Outlander here) Watchable, who produced a 76 on 29 Feb.

1621021413169.png

The 14th different set of plausible ratings for Lingfield. I think I need to get on with 2021 soon enough.
 

Attachments

  • 1621021258212.png
    1621021258212.png
    74.5 KB · Views: 1
Yes all understood. That said if its possible to give horses downgrades or upgrades why not have a column that gives the rating on the ground but put something next door to it that offers an indication that it might have been in a false run contest, or on the Primo Oumo side of the coin there's some kind of bizarre lack of logic to the performances. You always read of Timeform offering sectional upgrades to certain performances, so they are giving ratings that advance on the actual final time.

How much of an issue would it be to say Gordon Lord Byron managed a speed figure of 37, but the maths implies in a truly run race he might have been good enough for a 78? It's a fascinating point of debate.

An initial look at the Tryster race I talk about suggests he got a baseline figure of 25 but the upgrade moves him onto a figure of 68. In his previous track victory he achieved an unadjusted figure of 83.
AustinDillon75 AustinDillon75
I think your whole approach shows a lot of knowledge and skill, no doubt your ratings will prove useful to you.
Your thread has made me want to get my own figures up to date again so has been inspiring, some of you ideas I’ve found very interesting.
Its a thread that interests me because it’s someone coming at it fresh trying learning as you go along and documenting your findings, making corrections and trying to improve continuously.
Speed ratings can be addictive , it’s great when you have your own set of ratings for a card to base your opinions on. Without my ratings I lose more, with them I‘m treading water or making modest profits at best.
I prefer having the numbers there as a guide, I feel that I’m guessing without them.
 
AustinDillon75 AustinDillon75
I think your whole approach shows a lot of knowledge and skill, no doubt your ratings will prove useful to you.
Your thread has made me want to get my own figures up to date again so has been inspiring, some of you ideas I’ve found very interesting.
Its a thread that interests me because it’s someone coming at it fresh trying learning as you go along and documenting your findings, making corrections and trying to improve continuously.
Speed ratings can be addictive , it’s great when you have your own set of ratings for a card to base your opinions on. Without my ratings I lose more, with them I‘m treading water or making modest profits at best.
I prefer having the numbers there as a guide, I feel that I’m guessing without them.
They certainly are addictive O Outlander. Thankfully my wife isn't badgering me about time spent at the kitchen table - yet. I am honestly just grabbing a few minutes here and there and having a play.

As you do one course you transfer it to another course and something is slightly a miss. The key is knowing why that might be the case. It feels like the ratings are more realistic when you adjust distances at each track then calibrate each track as a whole. It makes sense to me but may not to anyone else.

Until I have figures to hand I am doing no betting at all. I don't care if it takes months, years to get right. And above all, I find it fun so thats what matters most.
 
Fair play to you AustinDillon75 AustinDillon75, but why not try and get your wife interested as well - it's certainly worked for me (even though I have had to fork out a few times when she's won the tipping competition)
dave58 dave58 the reason the marriage works is actually because we don't share any interests. The only thing we do together is watch Coronation Street. She isn't much a sports fan and I don't really care much for netflix or sky atlantic. She's happy enough to come racing at Aintree and enjoy the social occasions.

I'd love my kids to take a keen interest, I really would. They are only 8 and 5 so there's time for all that. They are good with numbers and that'll help.

My daughter's first race meeting was aged 18 months, we went to a midweek meeting at Beverley whilst on a caravan break in Withernsea.

I remember the bouncy castle. I also remember getting no winners.
 
I'm still looking at this outlier business and to be honest, still not totally satisfied. Its after looking at the meeting on 12 April 2013, where the first three runner race has thrown everything else out of kilter. Its also a formula issue as well.

I'm also of the mind that I need an OR for all horses in a meeting adjusted for age, and instead of converting RPRs, I am going to try the algorithm I've used to convert HRB ratings to an OR. If there is an OR, I will continue with it, but it will fill gaps for unrated horses. From there, the going allowance will be per race, but converted to the meeting average, though if any race allowance is an outlier compared to the meeting, it will remain in place.

This isn't binning the solid work done so far, its there to fall back on if this doesn't work. I know what I'm talking about at least if I am expecting no-one else to. I can share the general idea behind the HRB algorithm later, suffice to say the last race vs OR does fit a polynomial line perfectly (subject to setting an OR range of 45-120), whereas jockey/trainer/today ratings seem to fit the linear curve perfectly, in my settings at least. Once the formulas are ready they can just simply be plugged in.

We are still in development so its fine. What's surprised me is that given all the calculations, my spreadsheets have not ground to a complete halt.
 
Thanks, I'll run it my way via Excel and see what comes up
I used the Excel file I had from the Z-score post.

The mean is 56.88
SD x 3 (0.529) = 1.587
Upper limit 56.88 + 1.587 = 58.467
Lower limit = 56.88 - 1.587 = 55.293

No Outliers within range.

SPSS
SPSS.PNG

z-score.PNG

Mike.
 

Attachments

  • ZScoreOutliers.xlsx
    17 KB · Views: 11
So all the formulas have been modified and corrected, as we mentioned, I've pulled out algorithms to create ORs rather than zeros for all horses and adjusted them for age.

As O Outlander mentioned its debatable as to whether upgrade a time but I can have the best of both worlds, this sheet is sorted to "RTG", which is the actual time produced if there was no outlier on the day. So Primo Uomo delivered a 91 on 3 March 2017, with three other horses figuring in this list, but I have now created a different going allowance for that race using outliers that has it fast to the tune of 0.25 seconds per furlong. The result is that it will downgrade that performance to a 58 (see ADJ3) which I think most would say is a more realistic appraisal of his run. The remainder of that day's racing was a track slow by 0.12 seconds per furlong with no other abnormal ratings.

There is an interesting occurrence on 16 November 2020, where the overall going was track slower than 0.08 seconds per furlong to standard, but the formulas suggest the race won by Saltonstall was slower than would have been expected, despite being 2 seconds quicker than the much lower quality race prior to it. The going for that race was 0.23 seconds per furlong below standard and justified adding 12lb to the speed figure for horses in that race, so after rounding we can either stick with Saltonstall producing a 56, or upgrading him to a 69. Romanised ran to 65 and this could have been worth a 77, if we are minded.

Two races later there's a completely different scenario where the winner St Georges Head is flattered by even a poor figure of 39. The suggestion is that that race would have a going 0.11 faster than standard, so that results in a downgrading for him to 28.

Take my word on this now but St Georges Head prior to this race ran the following non-adjusted times:
16 October 2020 - 28
30 October 2020 - 28
6 November 2020 - 28

So his 39 on paper could be argued not to have really been any better than those three previous runs. He managed an unadjusted 42 the race after in running 2nd but flopped the week after with a rating of 2.

So this table ranks all performances prior to adjustment - the top 20 again

1621091289771.png

And this one everything once we add a few adjustments. Don't kill me for upgrading three horses in the ADJ2 column. It does again set up Gordon Lord Byron as a decent horse much better than the 41 he actually managed, a thought vindicated by the passing of time.

1621092433183.png

I think I can work with these tables. I know some might baulk at my approach but generally the figures are looking and feeling really realistic to me. The race won by Urban Beat doesn't look odd to me, it seems like a very good race won by a good horse with other good horses in tow. Everything in it is 1 for 30 since but it could be a combination of several switching to turf and not being as effective, and the likelihood that several have reached their ceiling.
 
Back
Top