• Hi Guest, The forum will be moving hosts on 26 July and as such will be closed from Midday until the move has completed.
    As we will be with new hosts it may take a while before DNS get updated so it could take while before you can get back on the forum.
    I think it will take at least 4 hours but could easily be 48!
    Ark Royal
  • There seems to be a problem with some alerts not being emailed to members. I have told the hosts and they are investigating.
  • 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
  • Sorry for the ongoing issues that you may have been experiencing whilst using the forum lately

    It really is frustrating when the forum slows down or Server Error 500 pops up.

    Apparently the hosts acknowledge there is a problem.
    Thank you for using our services and sorry for the experienced delay!
    Unfortunately, these errors are due to a higher server load. Our senior department knows about the issue and they are working towards a permanent resolution of the issue, however, I'd advise you to consider using our new cPanel cloud solutions: https://www.tsohost.com/web-hosting


    I will have to investigate what the differences are with what We have know compared to the alternative service they want us to migrate to.
    Keep safe.
    AR

Standard Times

After an educational few days reading and posting a few times the speed figures thread I've decided to have a go at producing some standard times for AW tracks. The way I've got these times is this:

1. I have found load of old HRB results downloaded but based this particular set on the 2019 times. I have most of the 2020 times too so its as simple as loading some more and my eventual idea was to perhaps take the previous season and adjust as the current season is ongoing.
2. The finishing time of each horse is converted to what it would likely have achieved carrying 9st and if it had an OR of 100.
3. Only horses aged 4 and over are included (as such it will include 4yo performances in 3yo+ races).
4. If the number of lengths beaten exceeds the number of furlongs in the race, the performance is discounted, so a horse beaten 7 lengths in a 5 furlong race is ignored, and horse beaten 20 lengths in a 12 furlong race are also ignored.
5. The 20th percentile of all runs is settled on - there are arguments for 10th, 20th, even the 50th percentile and I'm open to adjusting.
6. For each track, 10,000 sample random runs have been generated at each distance. Where the R-squared (which is always incredibly close to 1 either way) between the distances is better for observed races I've stuck with the observed races, where it appears more advantageous to use the random sample runs I've gone with those but typically, its only making differences of around +/- 0.1 seconds per furlong either way maximum.

1619208980509.png

I don't subscribe to RP, so it would be laborious to go through the race times and add their standards for comparison though from the ones I've seen they come quite close to around the 50th-65th percentile. Here's what happens to Wolverhampton when I run it from there. Lb/length barely actually changes, in truth, even when the longer 2 mile and a bit distance is 3 seconds slower.

1619209482609.png

I'm attaching my final times sheet. I can easily recalculate to different percentiles as per the above, and can answer any questions and provide tweaked calculations to suit. I'd have happily attached the actual working spreadsheet itself, but its 12MB and too big despite my efforts to reduce it.

I'd appreciate any views on these attempts and for seasoned speed compilers, will be interested in maybe how they compare with your own standards or way of working. I have become convinced that the investment in producing accurate standard times is well worth it and can offer something else over and above the average punter's armoury. The next step will be getting figures to work off the back of them. AW first, then flat, then NH. No quadrilateral exchanges or reliance on power trendlines at this stage.
 

Attachments

  • AW Standard Times 2019.xlsx
    11.8 KB · Views: 10
  • 1619209370643.png
    1619209370643.png
    31.8 KB · Views: 8
I don't subscribe to RP, so it would be laborious to go through the race times and add their standards for comparison though from the ones I've seen they come quite close to around the 50th-65th percentile.
When I was using SPSS and looking at the RP standard times I found the closest percentile to Dave Edwards RP standard time was the 15th percentile, one of the main problems when dealing with racecourse that have the polytrack surfaces is the speed of the surface, for instance, Kempton is always publishing the official going allowance as standard to slow -0.25s/f, not once have I found it to be so.

The subject of going allowances has been extensively covered on this forum by both davejb davejb and O Outlander who understand the complexities of the subject.


Mike.
 
Last edited:
TheBluesBrother TheBluesBrother thank you for that link. I had wondered already about dismissing certain outliers and that was in my mind with the dismissal of certain runs where lengths beat > furlongs run. From your examples in that thread, it looks as if you only remove a selection of the fastest times before going to the 15th percentile.

I will try that method later and see what comes of it.
 
This is a sheet for Wolverhampton's 5 furlongs. What I did was put in the top 25 times in my list, then when the z-score was 1.5 or greater deleted it, working from top to bottom until each z-score in the 25 times was less than 1.5. This left the 17th to 24th fastest times untouched, and when plotting the intercept it gave a z-score time of 58.79. My 20th percentile time from the charts above was 58.98. Will do it now for 6 furlongs as well.

1619251919562.png
 
AustinDillon75 AustinDillon75

Looking at the examples you put up, in a sample size of 25 you should only have one or two outliers, not a shedful.

Mike.

This was the sheet when first produced. In your Deaville example in the other thread you got rid of the red times but then chose to delete another one "because it didn't look correct on the graph". I kept deleting times until z-scores either side of -1.5/1.5 stopped appearing.

So if I just stick to the original sheet and omit the outlier times (times 1-3) and take the 15th percentile of those remaining it produces a standard time for Wolverhampton 5f of 58.41.

Does this maybe look a bit more like it should?


1619253431176.png
 

Attachments

  • 1619252938264.png
    1619252938264.png
    45.7 KB · Views: 2
5 furlongs at Kempton and Dundalk. Surprisingly few times to go off at Kempton, hence it seems to take the slower times as the outliers.

1619253733273.png

1619253779886.png
 
AustinDillon75 AustinDillon75

A major point to consider regarding Dundalk, just recently the Polytrack racing surface was relaid, I found that this quicken up the race times and an adjustment in the future will no doubt be required.

I was just comparing davejb davejb standards against the RP, looking at your standard time for the 15th percentile of 57.47s it looks good to me, at the end of the day all it comes down to is opinions, the real work begins learning about going allowances which on the AW can range from -0.25s/f (std to soft) up to +0.50s/f (fast), so that's a range of 75lbs per furlong.

dundalk.png

Mike.
 
AustinDillon75 AustinDillon75

The most powerful statistics program I have installed on my system is SPSS.

I just loaded Dave Edwards RP standard times into SPSS and ran a regression curve estimation using the quadratic function which plots a straight line through his data points returning an adjusted set of standard times.

Quadratic.PNG

Visually they look pretty good to me.

Mike.
 
Yes, that's a lot better, start off with a list of 35 of your best times remove the outliers, then readjust your sheet to suit.

Mike.
I've done rightly or wrongly my exercise of generating 10,000 random runs for the above distnaces. When I have run 10,000 horses at Dundalk and done it a few times the result never varies by more than .01 seconds either way at 5f. The right hand of these two is the fastest ten runs by 10,000 horses so little wonder the Z-scores are so high. The fastest 93 runs are thrown out but so too are the slowest 75%.

1619259274293.png

When the small sample size of 21 is enlarged to 10,000 for Kempton it has the effect of adding about .23 seconds to the standard at the same distance once the outliers are disregarded. In this case, the top 633 and bottom 657 runs become outliers so the 15th percentile is taken from the middle band of 8,170 runs.

1619259405010.png

So I am guessing that over time, as the sample size increases, it may come to pass that runners could be more fairly treated by the adjusted standard time of 57.86 (ironically that is the standard time I arrived at in the charts posted above). Clearly there were enough runners in the Dundalk example for the standard time to be accurate with great confidence but that is around 0.3s faster than the time in the chart above.

What I'm looking at is a suitable way of dealing with smaller samples. Would add that my trial and (mostly) error are for all 2019 runs so any more recent track alterations will not be taken into account for the purpose of this exercise but will bear in mind when adding later data.
 
Last edited:
AustinDillon75 AustinDillon75

The most powerful statistics program I have installed on my system is SPSS.

I just loaded Dave Edwards RP standard times into SPSS and ran a regression curve estimation using the quadratic function which plots a straight line through his data points returning an adjusted set of standard times.

View attachment 99263

Visually they look pretty good to me.

Mike.
I'm supposing that SPSS must be a great tool but sadly, I am only versed with Excel.
 
Look what happens here with Southwell. This is over 1m4f.

When you take the best 25 actual times achieved on the track/distance, you would only remove the 149.59 time and the 15th percentile of those remaining is 151.14.

Over 10,000 random races, it removes the top 539 runs (the fastest theoretical 25 times are on the right with z-scores all below minus 5) and only the next 2,256 are within the z-score range (between -1.5/1.5), and it amazingly results in a 15th percentile standard time of 151.15 (0.01s slower than that observed).

Clearly, there must already be enough runs at Southwell over that distance for the observed times to be used pretty confidently as a suitable standard time.

1619260645099.png
 
Over 10,000 random races, it removes the top 539 runs (the fastest theoretical 25 times are on the right with z-scores all below minus 5) and only the next 2,256 are within the z-score range (between -1.5/1.5), and it amazingly results in a 15th percentile standard time of 151.15 (0.01s slower than that observed).
Don't waste your time with Southwell AW, they are ripping up the surface and replacing it with Tapeta.

The standard time of 151.14s/151.15s for the trip looks pretty good, the RP has the standard times at 154.0s?

Mike.
 
Don't waste your time with Southwell AW, they are ripping up the surface and replacing it with Tapeta.

The standard time of 151.14s/151.15s for the trip looks pretty good, the RP has the standard times at 154.0s?

Mike.
Again noted with Southwell but its fine, just experimenting now and with the help on outliers and z-scores the times are falling into a bit more of a line. When Southwell goes to tapeta we'll all be back at the same drawing board with that course obviously.
 
With the benefit of all the above advice and some further trial and error, including some corrections to clear errors in the spreadsheet, I've now got the following standard times to work with. Would be interested to know if those who know better feel these are somewhere close to how they should be looking. They are actually a fair bit different to the ones I posted yesterday but are based on removing outliers and taking a slightly lower percentile (actually, the 6th).

1619281663722.png
 

Outlander

Gelding
With the benefit of all the above advice and some further trial and error, including some corrections to clear errors in the spreadsheet, I've now got the following standard times to work with. Would be interested to know if those who know better feel these are somewhere close to how they should be looking. They are actually a fair bit different to the ones I posted yesterday but are based on removing outliers and taking a slightly lower percentile (actually, the 6th).

View attachment 99279
Interesting work AustinDillon75 AustinDillon75 , very impressive
ive been sick as a dog for a week now but will definitely be contributing when I recover
 
Would be interested to know if those who know better feel these are somewhere close to how they should be looking.
I am very impressed, I would add that your standard times are more accurate than the RP's, you putting up the Laytown beach standards did make me smile.

When dealing with the Irish standard times you will meet a different set of criteria, notably the fictitious official race distances.

The day when the Irish authority gets around to actually measuring the racecourse distances there will be a lot of red faces, Lorcan Wyer had a private firm to measure Galway's race distances back in 2015 and to his horror, he found that the 2m1f(C) distance was in fact 2m2f.

It was because of this, I put together my universal linear pars table.

Mike.
 

Attachments

  • LinearPars.xls
    43 KB · Views: 15
I am very impressed, I would add that your standard times are more accurate than the RP's, you putting up the Laytown beach standards did make me smile.

When dealing with the Irish standard times you will meet a different set of criteria, notably the fictitious official race distances.

The day when the Irish authority gets around to actually measuring the racecourse distances there will be a lot of red faces, Lorcan Wyer had a private firm to measure Galway's race distances back in 2015 and to his horror, he found that the 2m1f(C) distance was in fact 2m2f.

It was because of this, I put together my universal linear pars table.

Mike.
Interesting work AustinDillon75 AustinDillon75 , very impressive
ive been sick as a dog for a week now but will definitely be contributing when I recover
Its these type of replies from people who know their game that really encourage me to keep working. I think its now about keeping it simple from here and for now I think I'll just stick to AW times until I get more confident in making use of them. If good standard times really are the foundation of good figures and from there better selections the time will be well worth it.

I've imposed a 6 month sabbatical on all betting to try and find a different route to success, too much money just being frittered away without thinking.

Hope you are feeling better soon O Outlander, I had my first covid jab last week and it knocked me out for 24 hours. I had it at Aintree Racecourse, and had a completely unapprehended walk of the course which was beautiful.
 
Top