Hong Kong Speed Figures

[Outlander]

O Outlander - I have a couple of formulas that give a rating for a part sectional eg 800m on the same scale as your final timefigures - your correct in that it is connected with the final time. Give me a day or so to dig them out and i'll do some workable examples. A sectional timefigure is just a compressed timefigure if you think about it, and if their scaled the same it shows if the pace to 800 was slower or faster compared to the final time. Btw this is my own work and has nothing to do with Simon Rowlands technique, you can also produce ratings for any sectional juncture with my formulas on the same scale as your timefigures. eg start to 400. Basically they are "pace ratings" scaled to timefigures.

Thanks for this ARAZI91 , I’m carefully going through your previous post and appreciate any of your wisdom on the subject, I’m not the brightest spark but trying to piece it together. I’m using the formula you mentioned previously to create ratings for overall timefigures and for the sections (it’s the most sold way I can think of to express like for like) , I’m very confident I have high quality standard times for overall and each section, currently working solely on the last 800 to see if I can make sense of anything before I go further with other sections.
I really looking at that run by Golden Sixty 21/11/21 as the most extreme test case in my database. I have ST 1600 standard time of 93.57 GS ran 94.55 (slow 0.98), but ran unprecedented (in comparison to overall time) 43.2 for the last 800m when I have a ST 1600(800-F) standard time as 46.00 so (fast 2.8 seconds) , I think it’s almost unbelievable, the product of a top class horse and a slow overall time .
Using the formula you gave me they rate approx 79 for the overall distance and 164 for the 800-F section as I mentioned earlier. GS peak form ability would be 113-114 on the scale .
My latest thoughts are a weighted average( (79x1.5)+164)/2.5 would give a very realistic 113.assuming neutral going(not going to worry about going allowances until I can make the sections sensibly line up with overall timefigures)
Very very simplistic, and not going to have chance on my PC tomorrow to see how that brings out all the other runners in my database.
Be very interested in your formulae if you can dig them out and assuming I can understand.
Thanks for your help and effort put into these replies, always appreciated.

 

[ARAZI91]

Thanks for this ARAZI91 , I’m carefully going through your previous post and appreciate any of your wisdom on the subject, I’m not the brightest spark but trying to piece it together. I’m using the formula you mentioned before to create ratings for the sections, I’m very confident I have high quality standard times for overall and each section, currently working solely on the last 800 to see if I can make sense of anything before I go further with other sections.
I really looking at that run by Golden Sixty 21/11/21 as the most extreme test case in my database. I have ST 1600 standard time of 93.57 GS ran 94.55 (slow 0.98), but ran unprecedented (in comparison to overall time) 43.2 for the last 800m when I have a ST 1600(800-F) standard time as 46.00 so (fast 2.8 seconds) , I think it’s almost unbelievable, the product of a top class horse and a slow overall time .
Using the formula you gave me they rate approx 79 for the overall distance and 164 for the 800-F section as I mentioned earlier. GS peak form ability would be 113-114 on the scale .
My latest thoughts are a weighted average( (79x1.5)+164)/2.5 would give a very realistic 113.assuming neutral going(not going to worry about going allowances until I can make the sections sensibly line up with overall timefigures)
Very very simplistic, and not going to have chance on my PC tomorrow to see how that brings out all the other runners in my database.
Be very interested in your formulae if you can dig them out and assuming I can understand.
Thanks for your help and effort put into these replies, always appreciated.

O Outlander - as my Going allowances are in "pounds" and i use individual Class ratings as "pars" for what the horse achieved on the day -Golden Sixty's "Class" rating for that day was 131 -if we assume a neutral going allowance and using your standard times(both final and sectional time) the horse would have achieved a final timefigure on my scale for HK (0-135) using one of my formulas of 113 , with a proportional 800m sectional rating of 111. Using my own standards of 1600 -93.36 and 800mtr par - 46.36 , again assuming a neutral going allowance, Golden Sixty achieved a final timefigure of 111 and a proportional sectional rating of 119 - this reflects the faster early pace to 800 as theoretically if he had kept up that 43.20 800mtr pace he would have come home in 86.40!!! But as we all know horses "decelerate" the further the race goes so NO HORSE would ever achieve that. Sectional pars should be proportional to the standard time in terms of "relationships" - this is what matters most in standard times -the relationship between each standard to each distance. Your 800 mtr sectional par proportion is 0.48952 or 48.952% to your final standard time, my 800mtr sectional par proportion is 0.503375 or 50.3375% to my final standard time - that is not to say that for every "half" distance par to a final time should be around 0.50 or 50% just in this case that is what the relationship is. At 800mtrs as distance increases these relationships should get larger unless there is a very good reason (usually track topography). With 400mtr pars the relationship distribution decreases as your dealing with a longer % of time when starting at 1000 mtrs - here's my standard's for Sha Tin Turf , along with the 400 and 800 pars and their resultant relationships as well as time per 200mtrs. To get the 400mtr pars you just multiply the standard time by the relationship. To get the 800mtr pars you do (1-Relationship)*(Standard Time) - By the way Golden Sixty's figures above are not weight adjusted. The 800m pars also account for the distribution of Anaerobic / Aerobic energy which equalises as a threshold just under 6.5f on average - that's why the 800m pars get a little faster(per 200mtrs) at 1000 and 1200mtrs then start to get slower compared to the 400mtr par at around 1400mtrs plus. Sprinters mostly run 80-90% using Anaerobic energy whilst a 7f horse on average will use around 50/50 - the switch and % used changes from Anaerobic to Aerobic energy used as distances get longer.

Sha Tin Turf Course Standard Times, 400 mtr sectional pars, 800 mtr sectional pars, Relationships, Time Per 200 mtrs and equivalent Pearson's Correlation Coefficients

 

[Outlander]

O Outlander - as my Going allowances are in "pounds" and i use individual Class ratings as "pars" for what the horse achieved on the day -Golden Sixty's "Class" rating for that day was 131 -if we assume a neutral going allowance and using your standard times(both final and sectional time) the horse would have achieved a final timefigure on my scale for HK (0-135) using one of my formulas of 113 , with a proportional 800m sectional rating of 111. Using my own standards of 1600 -93.36 and 800mtr par - 46.36 , again assuming a neutral going allowance, Golden Sixty achieved a final timefigure of 111 and a proportional sectional rating of 119 - this reflects the faster early pace to 800 as theoretically if he had kept up that 43.20 800mtr pace he would have come home in 86.40!!! But as we all know horses "decelerate" the further the race goes so NO HORSE would ever achieve that. Sectional pars should be proportional to the standard time in terms of "relationships" - this is what matters most in standard times -the relationship between each standard to each distance. Your 800 mtr sectional par proportion is 0.48952 or 48.952% to your final standard time, my 800mtr sectional par proportion is 0.503375 or 50.3375% to my final standard time - that is not to say that for every "half" distance par to a final time should be around 0.50 or 50% just in this case that is what the relationship is. At 800mtrs as distance increases these relationships should get larger unless there is a very good reason (usually track topography). With 400mtr pars the relationship distribution decreases as your dealing with a longer % of time when starting at 1000 mtrs - here's my standard's for Sha Tin Turf , along with the 400 and 800 pars and their resultant relationships as well as time per 200mtrs. To get the 400mtr pars you just multiply the standard time by the relationship. To get the 800mtr pars you do (1-Relationship)*(Standard Time) - By the way Golden Sixty's figures above are not weight adjusted. The 800m pars also account for the distribution of Anaerobic / Aerobic energy distribution which equalises as a threshold just under 6.5f on average - that's why the 800m pars get a little faster(per 200mtrs) at 1000 and 1200mtrs then start to get slower compared to the 400mtr par at around 1400mtrs plus. Sprinters mostly run 80-90% using Anaerobic energy whilst a 7f horse on average will use around 50/50 - the switch and % used changes from Anaerobic to Aerobic energy used as distances get longer.

Sha Tin Turf Course Standard Times, 400 mtr sectional pars, 800 mtr sectional pars, Relationships, Time Per 200 mtrs and equivalent Pearson's Correlation Coefficients

View attachment 134513

Thanks ARAZI91 again superb reply as usual , clearly spent a lot of time and effort and very helpful.
We obviously working from different standard times for sectionals.
I calculated my own from the data but worryingly for me the ‘relationships’ seem way off from yours above .
still taking the ST 1600 I had close to 6000 individual runs in my database, i thinned down slightly by culling any performance by any horse not within 10lb form rating of the horses peak form rating and also culled a few runners who had never shown the ability to rate high enough to win a race in HK.
Had 4244 runs left in my database and of these the average relationship was 0.4917, this is the figure I used 1600 standard 93.57 * 0.4917 (rounded)= 46.00. I have 22.93 for the 400m-F sectional
The range is 0.4569 (Golden Sixty 21/11/21) to 0.5196
The 0.503375 is on the 91st percentile in my data.
might have to review my methods properly when I get time at my PC

my section pars don’t look as uniform as yours but derived using the same procedure from my data, so hard for me to ignore what that is telling me
if the horse with the most outstanding final 800m compared to overall time probably in HK history has been downgraded from 113 to 111 then not really what I was hoping for was hoping 800m-F could highlight horses who are better than their overall timefigures , unexposed horse not G60, we know how good he is now just using his data to look for clues to cracking to code.
Must admit had a few looks at sectional and couldn’t find any added value so might call it a day and just stick with the overall timefigures unless my weighted average idea producing anything usable .
Thanks ARAZI91 as usual great post.

 

[AustinDillon75]

Just had time to squeeze some ratings for Wednesday before heading off. I was wanting to see the jockey bookings and bear in mind on these ratings a jockey change before hand can affect some of the ratings quite significantly. Zac Purton has a very good set of rides even for him.

Anyway, attached.

 

[ARAZI91]

O Outlander - Btw the Standard times on the HKJC site are pretty mince - the Standards are not too bad but when they start to give individual standards for each "Class" they start going down the wrong road. One standard is all you need as long as you know what that standard represents. If your using either Official Class Ratings or Private Handicap ratings as "ability pars" then one standard will suffice.
One last little handy metric that is suited to Hong Kong and that is "weight bias" especially as they publish declared body weights

Using the same two races and a race split half technique by Fin.Pos we can check for any bias within Carried Weight, Body Weight and Total Weight.

Clearly weight had an effect on this race with positive deltas on Carried weight, Body Weight and Total Weight (carried &body weight summed together)


Weight did not have as much of an effect in this race - Positive deltas in Carried Weight ,slight minus in Body Weight and a Positive Delta in Total Weight but as these are averages it's only Carried Weight that is causing that.

Something else to consider.

 

[ARAZI91]

O Outlander -this is how i calculate my timefigures for UK racing - I don't touch Ireland due to reasons well documented. I have most of this stuff automated in R so only need to manually work on the Going Allowance (which is in pounds) - exported to Excel the sheet is too long to get in one image but follows from left to right - Little metrics like weight spreads and how much fillies are in the race help (fillies run on average 5lb per mile slower than males)







I don't use timefigures in the conventional sense of a Rating system , ie Top rated, 2nd Top Rated etc - i use time analysis to pinpoint races that run faster or close to their ability - here the two races highlighted in black were of interest. They are also an input into one of numerous "Power Ratings" i construct. The figures are all weight adjusted to 140lbs as are the Ability Ratings (AB-Rat) or WFA equivalent.

I use my own WFA scale



I also account for rail movement's by adjusting the standard very similar to the way davejb does
I then use 5 inputs including Raw pounds+/- Standard Time, Lengths Per sec / XLengths Per Sec , the AB-Rating and a couple of other things and run it through a multiple linear regression - basically i want to see what the R2 and std error is - i may make manual adjustments to both the AB-Ratings and the timefigures based on this.

When splitting a Going Allowance say from Round track to Straight Track - i use "proportion" as i have all of the "run-in"'s measured using Google Earth - Those races on the Round Track will still have to run up the Straight Track anyway so find "proportion" is more accurate and it's easy to calculate because the Going Allowance is in pounds anyway. eg if the Going Allowance is say 72 on the straight track and say 55 on the round track over an example distance of 12f on a track where the "run in" coming off the round track is 3.85f then its just a matter of calculating the following (12f-3.85f)=8.15f so the proportional going allowance will be ((8.15*55)+(3.85*72))/12 = a "proportional" going allowance of 60 (rounded)

With Rail Movements eg 35 yds added to an original distance of 10f where the original standard was say 124.55 , i just convert the yards to furlongs (35/220)=0.16(rounded) - Divide the original standard 124.55/10f =12.46(rounded) and multiply that by the new distance of 10.16f ....so 12.46*10.16 gives a rail adjusted standard of 126.59.

 

[The Hare]

Just a quick update of results for the last 3 meetings. Realised I also had the wrong date on meeting before 16/5 should have been 17/5. All season up to date. Still catching upwoth my ratings and things. Will try and get a card pasted for Wednesday.

 

[Outlander]

Wow ARAZI91 must be a very professional set up you have, makes my stuff look amateurish.
I would never use anyone else’s standard time calculations such as RP or HJKC, agree with you regarding different standards for different classes being a nonsense.
my Sha tin standards are

1000. 55.56 or alt. 56.02
1200. 68.58
1400. 81.01
1600. 93.57
1800. 106.22
2000. 120.28
2200. 133.24
2400. 146.55

so looks like mainly 1200, 1400 and 1600 most notable differences
thanks for sharing

 

[ARAZI91]

O Outlander -this is how i calculate my timefigures for UK racing - I don't touch Ireland due to reasons well documented. I have most of this stuff automated in R so only need to manually work on the Going Allowance (which is in pounds) - exported to Excel the sheet is too long to get in one image but follows from left to right - Little metrics like weight spreads and how much fillies are in the race help (fillies run on average 5lb per mile slower than males)

View attachment 134522

View attachment 134523

View attachment 134524

I don't use timefigures in the conventional sense of a Rating system , ie Top rated, 2nd Top Rated etc - i use time analysis to pinpoint races that run faster or close to their ability - here the two races highlighted in black were of interest. They are also an input into one of numerous "Power Ratings" i construct. The figures are all weight adjusted to 140lbs as are the Ability Ratings (AB-Rat) or WFA equivalent.

I use my own WFA scale

View attachment 134528

I also account for rail movement's by adjusting the standard very similar to the way davejb does
I then use 5 inputs including Raw pounds+/- Standard Time, Lengths Per sec / XLengths Per Sec , the AB-Rating and a couple of other things and run it through a multiple linear regression - basically i want to see what the R2 and std error is - i may make manual adjustments to both the AB-Ratings and the timefigures based on this.

When splitting a Going Allowance say from Round track to Straight Track - i use "proportion" as i have all of the "run-in"'s measured using Google Earth - Those races on the Round Track will still have to run up the Straight Track anyway so find "proportion" is more accurate and it's easy to calculate because the Going Allowance is in pounds anyway. eg if the Going Allowance is say 72 on the straight track and say 55 on the round track over an example distance of 12f on a track where the "run in" coming off the round track is 3.85f then its just a matter of calculating the following (12f-3.85f)=8.15f so the proportional going allowance will be ((8.15*55)+(3.85*72))/12 = a "proportional" going allowance of 60 (rounded)

With Rail Movements eg 35 yds added to an original distance of 10f where the original standard was say 124.55 , i just convert the yards to furlongs (35/220)=0.16(rounded) - Divide the original standard 124.55/10f =12.46(rounded) and multiply that by the new distance of 10.16f ....so 12.46*10.16 gives a rail adjusted standard of 126.59.

O Outlander - Forgot about compiling standard times - basically since starting to compile timefigures i originally used the last 5 years of data - i only do the Flat and AW and add to that original 5 years of data. So when starting a course i only use Hcap races (pace is generally more genuine and you have the added bonus of having an OR next to each time) - i adjust all raw times to 140lbs and also Class adjust them to a figure of 90 using a constant. After that i run the courses batch of times through some software that omits obvious outliers. I then use Std.Deviation*1.5 around the mean of those times. From that i then use "percentiles" based on distance. It's true that there are more truly run races at 5f than there is at say 12f so the percentiles reflect that decreasing all the time as distance increases. I do a distance at a time starting at the least and working upwards until i have rough standards at all distances at a course. If the sample is low i may go back and take individual times of close up 2nd placed horses and even close up third placed horses. i then look at that set of "rough" standards and the relationships between distances - i may make some manual adjustments here to reflect the relationships. I then run them against the distance range using linear regression looking at the R2 and the std error - again from this i may make some manual adjustments based on the linear regression outputs. After that the last step is to sort from fastest to slowest within each year and give each year a weighting. Originally when i started with the 5 years of data, i used weights of 1.00,0.80,0.60,0.40 & 0.20 for each year from recent to oldest. Again i ran them through some software (R) to achieve a final recency weighted time for each distance. 2nd Final check was again to look at the relationships within each distance - again i may make some small manual adjustments here. Then back to linear regression where through iteration i'm looking to get an R2 of at least 0.95+ . When adding a new years worth of data i repeat the above process just on that years data and compare them to the original standards, particularly looking at the relationships between the times - here i just use correlation (Last year v Original Standards) - usually the correlation is on the high side and i just integrate them into the original data but changing the weights so they are adjusted for recency - a quick linear regression check to make sure the R2 is above 0.95+ and that's about it. Because my standards are adjusted to 10 stone(140lbs) and a BHA 90 rated horse whilst using "percentiles" they are a lot looser than say the RP's but it does not really matter where you pitch your standards - you could adjust them to a 75 horse for example and 126lbs if desired - what matters the most in standards is the time relationships between distances. What matters the most in the whole timefigure process is the all encompassing Going Allowance -you could have the most accurate standards on the planet but your figures would be all over the place if you cannot "best-fit" the GA - the good thing about using "ability pars" such as Official Ratings or Private Handicap Ratings and having the going in pounds is that when you normalise for weight , maturity etc with "ability" the Total lbs column is what the Going would have to be for that horse to equal it's ability rating.
This part highlighted in Blue with the oversized numbers


On that day there was a horse with an Ability Rating of 105 - Scriptwriter and when normalised for weight , maturity, lbs above/below standard etc had a total lbs value of 142 - which means for that horse to achieve that ability rating on "time" i would have to make the going allowance 142 - like this

But look what happens to the other horses 69 rated Man On A Mission has just ran a timefigure of 145 and equalled Frankel!! , 67 rated He's Our Star has ran a 143!! and of course with a Going Allowance of 142 i would have to make the Going Description "Slow" - severely Slow in fact!. Of course i'm jesting here and this would never happen - the point is that "Total lbs" column represents what each horse would need the going allowance to be, to equal it's ability rating (which is assessed after the race btw and not a pre-race figure) therefore the "minimum"in that column is a decent starting point to work from and then it's just a matter of finding the "best fit" around all horses on the card. The 105 rated Scriptwriter actually ran a "raw" 38.67lbs below standard and the race was really a crawl - after normalising for weight , maturity etc he ended up with a measly timefigure of 33, around 72lbs below his AB-Rat. Of the two races of interest ,both running +4lbs above their AB-Rat, Man On A Mission went on to win his next two races, whilst 2nd placed Glorious Charmer over his next 5 races had form figures of 31121, 4th placed Nellie French's next 6 runs went 321721. The other race won by He's Our Star worked out not as good as Man On A Mission's but the winner's next four runs went 4252 (both seconds beaten less than a length , one a neck so possibly unlucky) , 2nd placed Eye Of The Water won next time out, 4th placed Crocodile Power's next two runs ran 2nd both times , Lightly raced 5th placed Romanovich next 3 runs went 621 (the win at 7-1, BFSP 10.9) , whilst 9th placed Swiss Rowe's(only beaten 5L) next two runs went 5th and then won at 12-1, BFSP 17.2. There are those that diss "times" and believe that sectionals have all the answers and whilst they have a use their accuracy in this country has been abysmal. The company TPD(Total Performance Data) who supply sectionals to an ever increasing number of tracks in the UK have made big money in the US but the punters there demand accuracy and noticed errors and discrepancies right away and 10 of the 12 tracks they supplied sectionals to in the States have now reverted back to the good old beam technology and now use TPD as a "back-up" - the other mob CourseTrack don't even respond to any queries or emails and their error margin is far beyond the tolerances stated on their Racing TV Pdf's. So there is still plenty of mileage left in final timefigures especially in truly run races compared to ability.

 

[ARAZI91]

Wow ARAZI91 must be a very professional set up you have, makes my stuff look amateurish.
I would never use anyone else’s standard time calculations such as RP or HJKC, agree with you regarding different standards for different classes being a nonsense.
my Sha tin standards are

1000. 55.56 or alt. 56.02
1200. 68.58
1400. 81.01
1600. 93.57
1800. 106.22
2000. 120.28
2200. 133.24
2400. 146.55

so looks like mainly 1200, 1400 and 1600 most notable differences
thanks for sharing

And you mate -Thanks I see you have noticed the anomaly that is Sha-Tins 1000 metres which can produce some strange times when compared to other races on the card. I once read a 32 tweet thread based on Sha Tin's 1000 mtrs on twatter but was still none the wiser.

 

[Outlander]

O Outlander - Forgot about compiling standard times - basically since starting to compile timefigures i originally used the last 5 years of data - i only do the Flat and AW and add to that original 5 years of data. So when starting a course i only use Hcap races (pace is generally more genuine and you have the added bonus of having an OR next to each time) - i adjust all raw times to 140lbs and also Class adjust them to a figure of 90 using a constant. After that i run the courses batch of times through some software that omits obvious outliers. I then use Std.Deviation*1.5 around the mean of those times. From that i then use "percentiles" based on distance. It's true that there are more truly run races at 5f than there is at say 12f so the percentiles reflect that decreasing all the time as distance increases. I do a distance at a time starting at the least and working upwards until i have rough standards at all distances at a course. If the sample is low i may go back and take individual times of close up 2nd placed horses and even close up third placed horses. i then look at that set of "rough" standards and the relationships between distances - i may make some manual adjustments here to reflect the relationships. I then run them against the distance range using linear regression looking at the R2 and the std error - again from this i may make some manual adjustments based on the linear regression outputs. After that the last step is to sort from fastest to slowest within each year and give each year a weighting. Originally when i started with the 5 years of data, i used weights of 1.00,0.80,0.60,0.40 & 0.20 for each year from recent to oldest. Again i ran them through some software (R) to achieve a final recency weighted time for each distance. 2nd Final check was again to look at the relationships within each distance - again i may make some small manual adjustments here. Then back to linear regression where through iteration i'm looking to get an R2 of at least 0.95+ . When adding a new years worth of data i repeat the above process just on that years data and compare them to the original standards, particularly looking at the relationships between the times - here i just use correlation (Last year v Original Standards) - usually the correlation is on the high side and i just integrate them into the original data but changing the weights so they are adjusted for recency - a quick linear regression check to make sure the R2 is above 0.95+ and that's about it. Because my standards are adjusted to 10 stone(140lbs) and a BHA 90 rated horse whilst using "percentiles" they are a lot looser than say the RP's but it does not really matter where you pitch your standards - you could adjust them to a 75 horse for example and 126lbs if desired - what matters the most in standards is the time relationships between distances. What matters the most in the whole timefigure process is the all encompassing Going Allowance -you could have the most accurate standards on the planet but your figures would be all over the place if you cannot "best-fit" the GA - the good thing about using "ability pars" such as Official Ratings or Private Handicap Ratings and having the going in pounds is that when you normalise for weight , maturity etc with "ability" the Total lbs column is what the Going would have to be for that horse to equal it's ability rating.
This part highlighted in Blue with the oversized numbers

View attachment 134536
On that day there was a horse with an Ability Rating of 105 - Scriptwriter and when normalised for weight , maturity, lbs above/below standard etc had a total lbs value of 142 - which means for that horse to achieve that ability rating on "time" i would have to make the going allowance 142 - like this
View attachment 134537
But look what happens to the other horses 69 rated Man On A Mission has just ran a timefigure of 145 and equalled Frankel!! , 67 rated He's Our Star has ran a 143!! and of course with a Going Allowance of 142 i would have to make the Going Description "Slow" - severely Slow in fact!. Of course i'm jesting here and this would never happen - the point is that "Total lbs" column represents what each horse would need the going allowance to be, to equal it's ability rating (which is assessed after the race btw and not a pre-race figure) therefore the "minimum"in that column is a decent starting point to work from and then it's just a matter of finding the "best fit" around all horses on the card. The 105 rated Scriptwriter actually ran a "raw" 38.67lbs below standard and the race was really a crawl - after normalising for weight , maturity etc he ended up with a measly timefigure of 33, around 72lbs below his AB-Rat. Of the two races of interest ,both running +4lbs above their AB-Rat, Man On A Mission went on to win his next two races, whilst 2nd placed Glorious Charmer over his next 5 races had form figures of 31121, 4th placed Nellie French's next 7 runs went 321721. The other race won by He's Our Star worked out not as good as Man On A Mission's but the winner's next four runs went 4252 (both seconds beaten less than a length , one a neck so possibly unlucky) , 2nd placed Eye Of The Water won next time out, 4th placed Crocodile Power's next two runs ran 2nd both times , Lightly raced 5th placed Romanovich next 3 runs went 621 (the win at 7-1, BFSP 10.9) , whilst 9th placed Swiss Rowe's(only beaten 5L) next two runs went 5th and then won at 12-1, BFSP 17.2. There are those that diss "times" and believe that sectionals have all the answers and whilst they have a use their accuracy in this country has been abysmal. The company TPD(Total Performance Data) who supply sectionals to an ever increasing number of tracks in the UK have made big money in the US but the punters there demand accuracy and noticed errors and discrepancies right away and 10 of the 12 tracks they supplied sectionals to in the States have now reverted back to the good old beam technology and now use TPD as a "back-up" - the other mob CourseTrack don't even respond to any queries or emails and their error margin is far beyond the tolerances stated on their Racing TV Pdf's. So there is still plenty of mileage left in final timefigures especially in truly run races compared to ability.

ARAZI91
i did have my own standards for all the U.K. / Irish Tracks, and did compile ratings for them all, but the processing time on excel was getting me down, coupled with premium charge and being banned restricted from bookies, also the straw that broke the camels back I was getting more joy from National Hunt speed ratings and my standards were miles better than RP but I became too frustrated with the fact that they appeared to be guessing at the actual finishing time it could be many seconds out and too frustrating knowing you are trying to get an accurate ratings and they are guessing at the winning time.
I just didn’t think it warranted my time anymore, I became quite good at calculating standards in general, certainly much better than available standards you could find in publications.
I just now want to fine tune all my ratings for HK , the data is incredible, everything is there, no excuses, so just want to see how accurate I can predict results using the rating, truth is The Hare is keeping records and my speed rating I think are in good profit to HK tote odds, but can’t match the form ratings he calculates that I use as the base for all my standard calculations and ratings.
When I get something I’m happy with will post ratings on this thread like AustinDillon75 and The Hare etc. and get back to having a few small bets n the HK stuff. Maybe with a few people interested we can help each other along.
I had my standards on an old thread but they would be out of date now

https://www.theukbettingforum.co.uk/XenForo/attachments/no-weight-speed-template-xlsx.84107/

 

[ARAZI91]

ARAZI91
i did have my own standards for all the U.K. / Irish Tracks, and did compile ratings for them all, but the processing time on excel was getting me down, coupled with premium charge and being banned restricted from bookies, also the straw that broke the camels back I was getting more joy from National Hunt speed ratings and my standards were miles better than RP but I became too frustrated with the fact that they appeared to be guessing at the actual finishing time it could be many seconds out and too frustrating knowing you are trying to get an accurate ratings and they are guessing at the winning time.
I just didn’t think it warranted my time anymore, I became quite good at calculating standards in general, certainly much better than available standards you could find in publications.
I just now want to fine tune all my ratings for HK , the data is incredible, everything is there, no excuses, so just want to see how accurate I can predict results using the rating, truth is The Hare is keeping records and my speed rating I think are in good profit to HK tote odds, but can’t match the form ratings he calculates that I use as the base for all my standard calculations and ratings.
When I get something I’m happy with will post ratings on this thread like AustinDillon75 and The Hare etc. and get back to having a few small bets n the HK stuff. Maybe with a few people interested we can help each other along.
I had my standards on an old thread but they would be out of date now

https://www.theukbettingforum.co.uk/XenForo/attachments/no-weight-speed-template-xlsx.84107/

Yes in NH racing the RP start timing when the horses as a pack start running which could be a good bit away from the starter on the ladder with the flag or as the tape breaks. I know Tom Heslop from Timeform who actually spends all day video timing race replays on the software Kinovea - mostly for two reasons - to get a final 3 or 2f sectional for their sectional archive and so they can calculate Simon Rowlands Finishing Speed % (which he stole from the Sartin Group in the US , who were studying fractional split times as far back as the early 80's - Simon also claims he invented PRB (% of Rivals Beaten) in the mid 2000's but it was a metric on the old Adrian Massey site as far back as 1997) - the other reason Tom Heslop of TF video times races is for NH racing as he starts the timer on the software as they pass the starter. I've seen comparisons of same race times between TF and the RP and there can be up to 25lb differences between the two. The last two Cheltenham Festivals were supposed to introduce "electronic" timing within NH racing that started as they passed the starter but having seen both Timeforms and the RP's versions there was still differences.
HK is a different ballgame as far as racing data goes (although you need to learn how to scrape) it really is the "Gold Standard" of racing data - i have around 9 years of older HK data , sectionals and all but they still round up their beaten margins which can make a difference cumulatively to your ratings. I posted earlier how to find the conversion point in seconds per length and then you can get more accurate margins but i know from private conversations with Steve Blume who was contracted to do some "time" work for Alan Wood's betting syndicate that they are the most accurate sectionals out of all the major racing jurisdictions. Add in Body weights , 3 types of replays including the stewards tape and very detailed stewards reports and it puts our country to shame - the thing is they are able to do this because they are the industry leaders at returning the highest % of the betting mans dollar back into their game with Japan a close second. Regarding the outstanding Prize Money this excerpt from our sports national "rag" the RP says it all


Racing in HK, Japan and even Australia are in a much healthier state than the predicament we find ourselves in - decreasing Field sizes, very large fixture lists, Hcap racing approaching 70% of total output with around 50% of that Class 5 to 6 , horses in training increasingly sold off abroad, Whip debates, Groups like Peta on our radar and with negative intention, Power Laws within the training ranks where the top 20 or so win around 70% of all prize money.
Look at the increase in Australia's prize money over the last 20 years or so - they must be doing something right/differently



I don't like negativity and prefer to remain positive personally as i'm a better person for it but our game at home needs some radical change and much like in business it must start from top-down - we need to attract more people to the game and a big % of that "more" must be younger. It's not just a question of sustainability but improving sustainability and progressing.

 

[ARAZI91]

Wow ARAZI91 must be a very professional set up you have, makes my stuff look amateurish.
I would never use anyone else’s standard time calculations such as RP or HJKC, agree with you regarding different standards for different classes being a nonsense.
my Sha tin standards are

1000. 55.56 or alt. 56.02
1200. 68.58
1400. 81.01
1600. 93.57
1800. 106.22
2000. 120.28
2200. 133.24
2400. 146.55

so looks like mainly 1200, 1400 and 1600 most notable differences
thanks for sharing

Meant to ask, do you use weight ? And what do these standards represent - a 100 horse on the HKJC scale?,something else?

 

[ARAZI91]

O Outlander having re-read your Golden Sixty post again i noticed that we may be talking about different pars regarding Golden Sixty - my 800m pars are from the start to the 800m juncture - you mentioned 800m-F in the Golden Sixty Post. Because my pars are "proportional" then my 800-F "par would be (Standard Time- Start-800 mtr time) so (93.36-46.36)=a 800-F par of 47.00. This changes things -actually quite a lot. We could revert back to Simon Rowlands work and calculate a Finishing Speed % but having studied his methods and know that he stole/ plagiarised the Finishing Speed % from The Sartin Group in the US, i tend to stay away from his methods as every man and his dog that use sectionals seem to use his methodology nowadays. I have frittered with them in the past but no longer use them for UK racing having done some work with my good friend Barry O'Neal in regards to error rates and just plainly wrong data. Truth is GPS or GNSS was not designed to use to obtain sectional times or positional points in horse racing - the whole process uses an algorithm called a Kalman Filter to "Estimate" these data points. Barry is an ex-equine therapist with over 30 years experience who worked in various yards , Stuart Williams, Jonjo's and also freelance for many trainers and has worked with Group and Graded horses such as Gordon Lord Byron and many others - Arthritis in his hands caused him to retire from the game but he's one of the best "physicality" experts i've ever known and for a couple of years made his living betting on paddock reviews all over tracks in Ireland. The Irish seem to have "horses" in their DNA and it's amazing the amount of trainers, jockeys , bloodstock agents etc that have came from his hometown of Fermoy. Barry lives just up the road from Coolmore Stud. Anyway back to Golden Sixty where we know the sectionals are accurate - his final timefigure would not change on my scale /standards which still stands at 111 but his 800-F rating is now uplifted to 126 where as before i was using the start to 800mtr par but unbelievably if you use the 400-F Par - basically (Standard time - Start to 400mpar) = 70.39 , Golden Sixty ran 67.48 (2.91 fast) - because this is over a longer distance (last 1200mtrs) this equates to an Internal Sectional Rating of 145!
Having looked at his sectionals from that race more closely and used the correct pars his internal pace/sectional ratings would go like this
Start to 400m - 24 - Very Slow Early Crawl - is running 107 lbs below his ability par.
Start to 800m - 119 - stepping it up now as after that early slow crawl to 400m this rating incorporates a very Fast 400 to 800 section so an uplift of 95 lbs compared to the start to 400 mtrs - by now is running 12lbs below his ability par of 131
Start to 1600m - 111 - Slow Early pace compromised the final time by 20lbs - Golden Sixty's ability par post race was 131

800-Finish - 126 - ran basically 9lbs below scale (0-135) but still a 15lb uplift on a final time which was compromised by a very slow early crawl and only 5lbs below ability par
400-Finish - 145 - ran basically 10lbs above scale (0-135) as it also incorporates the fast 400-800 section and 14lbs above his ability par of 131.

This ties in with the theory that a genuine top class "Elite" horse even in a slowly run race should at least sectionally at some point run close to or above his ability.

Rob

 

[Outlander]

Meant to ask, do you use weight ? And what do these standards represent - a 100 horse on the HKJC scale?,something else?

ARAZI91 the standards represent a 100 rated horse on neutral going , weight is included and I use The Hare private handicap ratings as the base which are dynamic and change with new evidence .
They are similar to HKJC OR but scale would would have a slightly lesser range.
For the purpose of my sectional experiment I was using the formulae you gave me a while ago I think you obtained from an Australian friend who specialised in HK racing, so although I used the same standards I converted everything on to that scale first that does not include weight.
The figures using the Australian guys formula and my standards were 89.64 for overall and 164.81 for the last 800m
Horse capable of between 113-114.
I thought I was strange on one of your earlier post that G60 rating was devalued by his last 800m , confused me at the time .

 

[ARAZI91]

ARAZI91 the standards represent a 100 rated horse on neutral going , weight is included and I use The Hare private handicap ratings as the base which are dynamic and change with new evidence .
They are similar to HKJC OR but scale would would have a slightly lesser range.
For the purpose of my sectional experiment I was using the formulae you gave me a while ago I think you obtained from an Australian friend who specialised in HK racing, so although I used the same standards I converted everything on to that scale first that does not include weight.
The figures using the Australian guys formula and my standards were 89.64 for overall and 164.81 for the last 800m
Horse capable of between 113-114.
I thought I was strange on one of your earlier post that G60 rating was devalued by his last 800m , confused me at the time .

Yeah there's a lot more to the Aussie's method concerning Mr SB than one formula and several equations. You don't get to to do contract work for one of the greatest gamblers that ever lived if you don't know what you are doing - i'll explain more of his formulae as we go -for instance ( STD_TIME / WIN_TIME ))-0.9)*1000 is for "winners" only - if you look at it closely it outputs (MTRS_PER_SEC)...... to treat the also rans (those in behind the winner) he uses the formula (1-(WIN_TIME/HORSE_TIME)^(RACE_DIST/1000))*1000 again the output is (MTRS_PER_SEC) but you deduct the resultant value from the winners value from the first formula and everything is normalised to a 1000 mtrs
to keep everything proportional. With sections you need to use the "relaionalship" values within another formula.(we will get to that if this is the road you want to go down) . SB also uses the official HKJC Class Ratings as "pars"
I also use "proportion" within my stuff just that my outputs are in pounds than in metres per second and the figures i put up earlier revising Golden Sixty's outputs at various sectional junctures as well as 800-F and 400-F are all done using proportion. That is why i bolded in the difference between Golden Sixty's 800-F rating and the 400-F rating as proportionally he achieved a faster time over the last 1200 metres which was 0.75 or 75% of the race compared to the slower time achieved over 800-F which was the last 800 metres of the race which equalled 0.50 or 50% matters - Ratios and Proportion are important
Back to "pars" and as said he uses the HKJC official Ratings as Pars but i'm sure he uses a weighted mean as a "Race Par" as SB say's himself

One thing that SB discovered was that "Beyer Figures" as from his method in Beyer's first book way "Picking Winners" way back in 1976 were simply "proportional ratios"
Beyer worked off of "par charts" and "beaten length" charts creating zillions of times for each class at each track and they were just plain wrong.
But in "Picking Winners" he gives an example of a set of pars at Santa Anita at 6f and on his "charts" a horse running 70.00 sec was worth a Beyer figure of 106 and a horse running 74.00 was worth a Beyer Figure of 55 meaning there was 51 between both horses- but Steve worked hard on that first Beyer book and eventually came up with this little equation that ((74/70)-.9)*1000)) = 157 now subtract the 106 and what do you get?? 51. This is when he realised that "Beyer" numbers were simply "ratios no matter what the scale -if that 157 represented 135 then the 84 represented 51. If that 157 represented 100 then of course the 106 represented 51. Instead of Beyer needing to have all these charts etc , he only needed that equation.

 

[Outlander]

Yeah there's a lot more to the Aussie's method concerning Mr Steve Blume than one formula and several equations. You don't get to to do contract work for one of the greatest gamblers that ever lived if you don't know what you are doing - i'll explain more of his formulae as we go -for instance ( STD_TIME / WIN_TIME ))-0.9)*1000 is for "winners" only - if you look at it closely it outputs (MTRS_PER_SEC)...... to treat the also rans (those in behind the winner) he uses the formula (1-(WIN_TIME/HORSE_TIME)^(RACE_DIST/1000))*1000 again the output is (MTRS_PER_SEC) but you deduct the resultant value from the winners value from the first formula and everything is normalised to a 1000 mtrs
to keep everything proportional. With sections you need to use the "relaionalship" values within another formula.(we will get to that if this is the road you want to go down) . Steve also uses the official HKJC Class Ratings as "pars"
I also use "proportion" within my stuff just that my outputs are in pounds than in metres per second and the figures i put up earlier revising Golden Sixty's outputs at various sectional junctures as well as 800-F and 400-F are all done using proportion. That is why i bolded in the difference between Golden Sixty's 800-F rating and the 400-F rating as proportionally he achieved a faster time over the last 1200 metres which was 0.75 or 75% of the race compared to the slower time achieved over 800-F which was the last 800 metres of the race which equalled 0.50 or 50% matters - Ratios and Proportion are important

Another thing I don’t understand to be honest, I think the formula is brilliant but don’t understand why you have to change it for beaten horses if you have the times for each individual horse .
Been using ( STD_TIME / HORSE_TIME ))-0.9 ) *1000 for each horse.
Probably makes some kind of mathematical sense above my understanding but if the winner wasn’t in the race the 2nd horse would be the winner and you would use the winners formula with that horses time.
As far as I can tell in Horse A wins r1 over 1200m in 68.75 and Horse B comes 5th in r2 in exactly the same time of 68.75 then using the different formula for beaten horses Horse B would rate different than Horse A despite the same time ?
Will try the beaten horse formula when at my PC , see what effect it has, very interesting as usual, thanks

 

[ARAZI91]

Another thing I don’t understand to be honest, I think the formula is brilliant but don’t understand why you have to change it for beaten horses if you have the times for each individual horse .
Been using ( STD_TIME / HORSE_TIME ))-0.9 ) *1000 for each horse.
Probably makes some kind of mathematical sense above my understanding but if the winner wasn’t in the race the 2nd horse would be the winner and you would use the winners formula with that horses time.
As far as I can tell in Horse A wins r1 over 1200m in 68.75 and Horse B comes 5th in r2 in exactly the same time of 68.75 then using the different formula for beaten horses Horse B would rate different than Horse A despite the same time ?
Will try the beaten horse formula when at my PC , see what effect it has, very interesting as usual, thanks

Simple answer is "proportion"
Your example of Horse A and Horse B would depend on the WINNERS TIME - the formula you have been using(for winners only btw) should actuallybe ((STD_TIME)/WIN_TIME))-0.90)*1000 not "HORSE TIME"

So your two example races are over 1200mtrs
Assuming in Horse A's Race the winner came home in for simplicity beat your standard 69.58 by 0.50 - using your standard for 1200m and a neutral Going Allowance so came home in 68.08
Horse A's Rating would be ((68.58/68.08)-0.90))*1000 = a rating of 107.34 or 107 rounded (which is actually "Mtrs per second)
then your other nag in that race came home in 68.75
so (1-(68.08/68.75)^(1200/1000))*1000 = 11.68 or 12 rounded
Which would give Horse A a rating of 107-12 = 95

Now let's assume Horse B came home winning in 68.65
again assuming a neutral Going Allowance
Horse B's rating would be ((68.58/68.65)-0.90)*1000 =98.90 or 99 rounded
then your other nag in that race came home in 68.75
so (1-(68.65/68.75)^(1200/1000))*1000 = 1.74 or 2 rounded
Which would give Horse B a rating of 99-2 = 97

This is how it should be as in the faster first race the mtrs per second value is worth a lot more than in slower second race so to finish 0.67 off a winner that records a 107 would have a slightly lesser rating then to finish 0.10 off a winner that records a 99.

In Steve's own words
"there should be NO constant value for a length, because beaten 1 second on a heavy track is not the same as being beaten 1 second on a fast one"

"beaten 1 second on turf here(AUS) is not the same as beaten 1 second on dirt there (US)"

"beaten 1 second in a fast time race is not the same as beaten 1 second in a slow time race.
it's all about "relativities" and "proportion""

"that calculation ((STD_TIME)/WIN_TIME))-0.90)*1000) should only be used comparing winner times not also run times or you would have the absurd thing that beaten 1 second over 1000 metres would have about twice the value of being beaten 1 second over 2000 metres, and that is plainly nonsensical"

Take my Sha Tin standards for example and compare the range of " mtrs per second" when proportion is used over the full distance range compared to something that Timeform would use in pounds per length or Poundage per 60 secs. Here i'm using the standard as the "Win Time" and the "Standard + 1 second" as the beaten horse, so in effect obtaining the worth of "one second" over the distance range using Steves formula , Timeform's constant of 1500lbs per 60 secs and their equivalent Pounds Per Length using a standard 6 Lengths Per second.

And here's using 8 years of Sha Tin data using 2nd to 7th place that shows length values relatively stay the same, over most distance ranges compared to the second way of using a fixed pounds per length system

pounds per lengths (values generally decrease as distance increases)


By the way Steve treats "sectionals" completely differently and does not use that formula that you are using which is for Final Times and "winners only"
I really should not be putting this stuff up in a public forum as it is the guy's life's work but when your contracted to "sort out" "Time based work" for at the time Globally the second largest betting syndicate on the planet (Benter's at the time was first) for Alan Woods then you would have to assume he knows his onions. Woods told Steve that variables used and constructed from Steve's figures accounted for around 30% of their models output. They were not just betting into Hong Kong but Singapore, Japan, Kranji, Korea - basically the whole Asian circuit.

 

[Outlander]

Simple answer is "proportion"
Your example of Horse A and Horse B would depend on the WINNERS TIME - the formula you have been using(for winners only btw) should actuallybe ((STD_TIME)/WIN_TIME))-0.90)*1000 not "HORSE TIME"

So your two example races are over 1200mtrs
Assuming in Horse A's Race the winner came home in for simplicity beat your standard 69.58 by 0.50 - using your standard for 1200m and a neutral Going Allowance so came home in 68.08
Horse A's Rating would be ((68.58/68.08)-0.90))*1000 = a rating of 107.34 or 107 rounded (which is actually "Mtrs per second)
then your other nag in that race came home in 68.75
so (1-(68.08/68.75)^(1200/1000))*1000 = 11.68 or 12 rounded
Which would give Horse A a rating of 107-12 = 95

Now let's assume Horse B came home winning in 68.65
again assuming a neutral Going Allowance
Horse B's rating would be ((68.58/68.65)-0.90)*1000 =98.90 or 90 rounded
then your other nag in that race came home in 68.75
so (1-(68.65/68.75)^(1200/1000))*1000 = 1.74 or 2 rounded
Which would give Horse B a rating of 90-2 = 88

This is how it should be as in the faster first race the mtrs per second value is worth a lot more than in slower second race.

In Steve's own words
"there should be NO constant value for a length, because beaten 1 second on a heavy track is not the same as being beaten 1 second on a fast one"

"beaten 1 second on turf here(AUS) is not the same as beaten 1 second on dirt there (US)"

"beaten 1 second in a fast time race is not the same as beaten 1 second in a slow time race.
it's all about "relativities" and "proportion""

"that calculation ((STD_TIME)/WIN_TIME))-0.90)*1000) should only be used comparing winner times not also run times or you would have the absurd thing that beaten 1 second over 1000 metres would have about twice the value of being beaten 1 second over 2000 metres, and that is plainly nonsensical"

Take my Sha Tin standards for example and compare the range of " mtrs per second" when proportion is used over the full distance range compared to something that Timeform would use in pounds per length or Poundage per 60 secs. Here i'm using the standard as the "Win Time" and the "Standard + 1 second" as the beaten horse, so in effect obtaining the worth of "one second" over the distance range using Steves formula , Timeform's constant of 1500lbs per 60 secs and their equivalent Pounds Per Length using a standard 6 Lengths Per second.
View attachment 134579
And here's using 8 years of Sha Tin data using 2nd to 7th place that shows length values relatively stay the same, over most distance ranges compared to the second way of using a fixed pounds per length system
View attachment 134581
pounds per lengths (values generally decrease as distance increases)
View attachment 134582

By the way Steve treats "sectionals" completely differently and does not use that formula that you are using which is for Final Times and "winners only"
I really should not be putting this stuff up in a public forum as it is the guy's life's work but when your contracted to "sort out" "Time based work" for at the time Globally the second largest betting syndicate on the planet (Benter's at the time was first) for Alan Woods then you would have to assume he knows his onions. Woods told Steve that variables used and constructed from Steve's figures accounted for around 30% of their models output. Thy were not just betting into Hong Kong but Singapore, Japan, Kranji, Korea - basically the whole Asian circuit.

Excellent clearly a genius, couldn’t get my head around it , how does a horse running the same time on the same day over the same distance rate differently, I see now, this and other rating methods that don’t understand the ‘proportion ‘ aspect is probably the main reason for bias towards lower weights who have graduated from lower class, as the beaten horse in the higher class are underrated.
Thanks

 

[AustinDillon75]

This is superb stuff. When I get home I'll be applying this across a range of experiments. I'm sure that a speed rating added to current metrics will improve strike rates further.