Hong Kong Speed Figures

[The Hare]

So I have been messing around , with the help of @Outlander for a while on these Hong Kong Ratings. As i say the information provided is excellent and easy to collect. But the Speed Figures compiled have been less than perfect! Myself and @Outlander compile the ratings in different ways but with some similarities. I have changed the methods but there is little difference.

The biggest problem is working out a correlation between the Official Ratings (or Racing Post Ratings) which are on a similar scale to the actual ratings my Speed Figures calculate. Today I have gone back to a method that at the 1000m (approx 5f) distance gives around 22lbs per sexonf, this is using the Timeform method (or at least the one they used 5 or so years ago). This would produce similar results to @Outlander or @TheBluesBrother or @davejb throughout the Forum. I have used higher figures but they don't seem correct.

So the figures produced have not been great, I would say the top rated produce lower than 20% success in races averaging about 12 runners per race. I realise these are usually competitive handicaps but it is not a great return. I have also messed around with calculating pure form figures which are better but still only around 20% for yop rated. I will have to analyse the results I have for this season and post!

So I was using all of the last 10 runs for a horse if it was in the last year - this may be a problem, could be going back to far for this. I know @Outlander uses a deterioration for runs further away but I think he would admit the results are similarly poor.

If i get time I will analyse my results with varying numbers of last runs for this season and post later.

So today I have been messing around with my new calculation method and compared the Official Ratings of winners with the ratings these produced. Now the structure of tacing in Hong Kong are strict Classes that are based on the Ratings -
Class 5 are 0-40 rated
Class 4 40-60
Class 3 60-80
Class 2 80-100
Class 1 above
Then Group Races above this.

So I compared the results of all the winners in the last 5 and a half seasons with there Official Rating and found the figures were very much determined by Class yet within the Class the ratings were very close to 1lb on Official Ratings equalled to 1lb on my speed Figures. This can be seen on the attached with clear breaks at 40, 60 and 80 - the Class 2 is more complicated as the top ratings can vary from 90,95 and 100.

So if we say Class 3 and 4, the average OR of the winner is 18.58 different whilst the SF differnece is only 6.39. Yet looking at the ratings between 41 and 60 and 61-80 the differences are very close to a similar scale, using Linear Regression I had 1.03 SF to 1 OR for Class 5 and 1.09 for Class 4.

This has completely confused me as it suggests a 40 rated in Class 5 is equivalent to say a 50 or even higher rating in Class 4 and so on.

Kevin

​

 

[Outlander]

It’s one of the great mysteries speed ratings work better in Irish jump racing where the times are considered accurate if they are within 10 seconds and the distances if they are within 2 furlongs, than in Hong Kong where there are just 2 flat courses with precision timing and distance data.
The Hong Kong OR scale is alien to anyone used to the UK / IRE scale, the best horse can be rated around 136 but the same horse would be worth 125 ish in the international classifications, it the reverse of that say an established 100 horse in the U.K. transfers to HK they will get a mark around 75-80. So where there could be 25 between a 100 rated and 125 rated horse in U.K. , that could easily end up being 55 lb + in Hong Kong.
Also they start off newcomers in handicaps off 52 , they could be future 120 horse just put in off 52 ( don’t know why 52)
A lot of the top horses run in handicaps off like 130+ ratings and you would think they would have no chance off their inflated rating but they seem to get on ok.
From all the ratings I’ve tried to compile HK I was expecting to be the best and they are the worst, if it took any amount of my time I would have given up by now, but with 2 meetings a week keep persevering in the hope maybe one day a tweak or two from better understanding of the system over there will bring about worthwhile improvements.
Looking at your data The Hare its gone over my head but the way speed relates to the official weights and class bands is the key to improvement , just think it’s a head scratcher over there.

 

[The Hare]

So thought I would put up a tale of the efforts I have made with Hong Kong ratings.

Firstly a couple of years ago I found a load of past data on the web. I then realised the Hong Kong Jockey club site had extensive information and details on how they operated racing there. Using this I updated and added to the data I had and have maintained it ever since. The information is easy to collect an keep in Excel or a database, which I haven't mastered yet!

I did it for interest, only the 2 meetings a week and long summer break means it doesn't take a lot of time. I don't bet on the racing, as far as I know the only option is Tote betting and if you look through the results you will see that the market is very accurate, much more than in the UK I suspoect.

The original plan was to determine how i could use speed figures and if I could find a way to use them! As mentioned above @Outlander and myself have attempted to compile ratings for a year or so with limited success. The next post will explain some methods I have tried.

 

[The Hare]

So as the racing in Hong Kong is very similar you would think compiling Speed Figures would be very simple and accurate. However results suggest otherwise.

The first issue with Speed Figures is compiling a method to calculate the Standard Times for the courses. Although there are only 3 tracks in Hong Kong, Happy Valley turf (HV), Sha Tin turf (ST) and Sha Tin all weather (STA). this should be a relatively simple task. I have tried many variant ways of doing this and although the figures I came up with were different the important thing is that the margins between the distances are very similar. The issue with Speed Figures is the inaccuracy that incorrect Standard Times can cause.

So I have decided to finally do it properly with my current list of Standards, the differences in them in the past were based on how many lbs/second I used and what level of ability the Standard was set to. Below I explain the class structure for my standards. For UK and Irish RAcing many people base their standards on the ability of top grade horses, in the US I believe they use the low class Claimers as the most common type of race.

So Classes are:
Group 1,2,3 - the Group 3 races can be handicaps when they apply to just local horses.
Class 1 - Virtually every non Group race in Hong Kong is a handicap. These are the top horses.
In a post above I explain the structure of Class 2-5 below this.
The very few stakes races they have are usually 2yo's, maybe 10 a season called Griffins and 5 or 5 races for 3YO and $YO only to determine Champions.
As @Outlander mentions the majority of unraced horses in Hong Kong have been imported from other countries and are usually given a starting Handicap Rating of 52 but other ratings of 68,74,80,85 and 90 are given to better horses but very rarely. Most horses are imported from Australia, South Africa, UK, Ireland and France.
There are only around 22 trainers and jockeys with occasional short visits from a foreign based jockey in their off season. Licenses are restricted so the only time other jockeys and trainers appear is in Group races.
Most racing takes place on Turf - ST on a Sunday (occasional Saturday) and HV on a Wednesday although again they sometimes are replaced by ST to allow the grass time to recover. ST also has occasional All weather races usually up to 4 a meeting though they do have a couple of All Weather only meetings. There are between 8 and 11 races a meeting.

So like I say it should be easy to compile Speed figures, but the failure of them so far suggests that there are maybe better ways to analyse the racing. Certainly when compiling UK jumps racing many years ago they were very useful. Though I always struggled with the Flat racing then though more due to less interest.

The next post I will finally explain the issues with the Speed Figures!

 

[The Hare]

So looking through the internet and this Forum there are many differing ways to calculate Speed Figures. But when looked at closer they are all very similar. The main difference relates to whether you use weight as most do or as in USA ratings weight is usually not included. In the USA though the weight spread is much less than in the UK, Ireland and Hong Kong.

As I said in a previous post when compiling ratings for UK racing my National Hunt ratings were much more accurate, backed up by @Outlander above. The ratings compiled by @TheBluesBrother and the thread Compiling Speed figures give an excellent rundown on how he compiles his figures, which is basically the method I use. On that thread @Outlander gives information on his method with a superb template to help on UK All Weather Speed Ratings Template. Also on the Early days thread @davejb compiles in a similar manner to @TheBluesBrother.

Earlier in this thread @ USFIGS suggests his method for Hong Kong was to take all the winners times at each distance for a year and remove top and bottom 10% then take the median. I decided that I would take the 25th percentile as my Standard Time. It shouldn't make a lot of difference but just removes the slower races.

Then to calculate my standard times I followed the following method:

1. I used the results of the last 5 seasons and current season, around 55,000 results. Split these into Distances and Classes for each track. I decided to use all the finishers times as most horses in Hong Kong are ridden to their best and the actual finishing time is given for all horses.
2. From these figures I compiled a different figure for each class, to ensure that there was a reasonable correlation between class and distance. This seems so as shown below for Happy Valley.

DISTANCE	Class 1	Class 2	Class 3	Class 4	Class 5
1000m		57.09	57.46	57.71	58.03
1200m	69.62	69.83	70.23	70.63	70.89
1650m	99.44	100.02	100.39	100.91	101.28
1800m	108.67	109.39	109.91	110.34	111.00
2200m			137.12	137.90	138.63

3. So as you can see these differences are fairly reliable. I decided that using the Class 2-5 races I would adjust the finishing times based on these differences, I averaged them out based on the number of runners that produced each figure to work out a Class difference. As mentioned above the Classes cover around 20 points on the Official Hong Kong scale.
4. This is where the problems start! The first time I performed the ratings I used a figure of 1200lbs, around the average weight of a horse and jockey. This means that a race run in 60 seconds has an allowance of 20lbs per second for times beaten. I then messed around using 1800 as my constant or 30lbs/second which seems too high, also tried a different figure for the All Weather racing around 25% lower as horses are generally beaten further on the sand, whether kickback or faster races I'm not sure. I have now settled on using the Timeform method as described on this link which uses 1500 and hence 25lbs/second, a split of my 2 previous attempts.
Horse Racing Times Explained: How to analyse times of horse races
5. So this is a very simple method to calculate a raw rating, I deduct for weight differences and include jockey allowance in the weights. I did mess around with non weight ratings but they were not as successful.
I then take the meeting and compare each horses last 3 ratings with the raw rating for today and take the median of the differences as my going assessment for the meeting. I have removed finishers beaten over 40 lengths as these have clearly not shoen their recent form and would give 3 low figures if they had run 3 previous races. This removal makes only a small difference and is more if I take a look at averages rather than medians in the future.
6. So then adjusting for the Going allowance gives me the rating for this race.

So the problems in the next post!

 

[The Hare]

So then the problems! Attached below are my calculations and eventual Standard Times for the different courses.

The first table gives my different standards for the Classes at each distance - those in yellow have very few runners and are unreliable.
Using the number of races at each distance I produced the Class distance in pounds, the second table, for each course. Now here is the first problem, as you can see I have a different set of figures for the 5f races at Sha Tin. All I can think of is that this is the sole straight course in Hong Kong and is usually run on a different side of the track. All other races go round bends, which can be placed in a variety of positions and differing distances from the inside line. These are clearly explained on the HKJC website. AS similar to UK racing and the random rail movements with Clerks of the Course not understanding how to calculate. I suppose the Hong Kong measurements are very accurate. I have yet to decide whether they change the starting position for these races and may have to watch some videos to decide. This can obviously affect the times of the races due to extra distance and the angles of the bends. That is something to look at, though I did compare the Ratings of the different courses once and the only affect was with the 5f sprint time against certain bend positions.

So after that ramble as you can see the 5f races at Sha Tin are very different class wise to the others. Later on I will attach a table showing the average Hong Kong Official ratings, my form ratings and my speed ratings for each class. As you can see from the Class differences here the speed figures produce approximate 10lbs/points difference between Class 4 and 3 and Class 3 and 2. These are the best to use as on Hong Kong official ratings the band covers 20 points and very similar on the Racing Post Ratings. On my own form ratings, calculated using a similar raw rating to the Time ratings and then adjusted on a race by race basis, my Class differences are around 19lbs between these bands so very close to the Hong Kong ratings.

Any explanations for this are helpful! Perhaps the bends make races slower but they are usually all run at breakneck speed so Speed ratings should match Form?

So the 5f ratings at Sha Tin are much closer to expected than the others, as mentioned above.

The last table are my new standards and the average number of seconds per 200m, a good way to check whether Standards are reliable. Then the number of lbs per second for each distance to calculate beaten horses.

So I have tried many methods of calculating Standards but they are similar.

Later I will show the difference between Form and Speed at each distance.

 

[Outlander]

So then the problems! Attached below are my calculations and eventual Standard Times for the different courses.

The first table gives my different standards for the Classes at each distance - those in yellow have very few runners and are unreliable.
Using the number of races at each distance I produced the Class distance in pounds, the second table, for each course. Now here is the first problem, as you can see I have a different set of figures for the 5f races at Sha Tin. All I can think of is that this is the sole straight course in Hong Kong and is usually run on a different side of the track. All other races go round bends, which can be placed in a variety of positions and differing distances from the inside line. These are clearly explained on the HKJC website. AS similar to UK racing and the random rail movements with Clerks of the Course not understanding how to calculate. I suppose the Hong Kong measurements are very accurate. I have yet to decide whether they change the starting position for these races and may have to watch some videos to decide. This can obviously affect the times of the races due to extra distance and the angles of the bends. That is something to look at, though I did compare the Ratings of the different courses once and the only affect was with the 5f sprint time against certain bend positions.

So after that ramble as you can see the 5f races at Sha Tin are very different class wise to the others. Later on I will attach a table showing the average Hong Kong Official ratings, my form ratings and my speed ratings for each class. As you can see from the Class differences here the speed figures produce approximate 10lbs/points difference between Class 4 and 3 and Class 3 and 2. These are the best to use as on Hong Kong official ratings the band covers 20 points and very similar on the Racing Post Ratings. On my own form ratings, calculated using a similar raw rating to the Time ratings and then adjusted on a race by race basis, my Class differences are around 19lbs between these bands so very close to the Hong Kong ratings.

Any explanations for this are helpful! Perhaps the bends make races slower but they are usually all run at breakneck speed so Speed ratings should match Form?

So the 5f ratings at Sha Tin are much closer to expected than the others, as mentioned above.

The last table are my new standards and the average number of seconds per 200m, a good way to check whether Standards are reliable. Then the number of lbs per second for each distance to calculate beaten horses.

So I have tried many methods of calculating Standards but they are similar.

Later I will show the difference between Form and Speed at each distance.

The Hare below my standards for comparison all pretty uniform around 0.11-0.13 sec/200m lower than yours, with the obvious exception of the 5f at Sha Tin and the longest distance at HV and ST.
I compile standards totally differently than most , but one of us is wrong on these particular 3 distances by about what amounts to 8-12 lbs

Probably not a simple as it seems as you suggest with the way the straight 5f relates to the turn races on the different racing lines in Sha Tin turf.

The performance of my ratings in HK would suggest my standards can‘t be perfect, so I’m not suggesting that it is your standards that are wrong.

STANDARD TIMES
		The Hare	sec/200	LBS/SEC	Outlander	sec/200	LBS/SEC	Correlation
HAPPY VALLEY	1000​	57.61	11.5220​	26.04​	56.94	11.3880​	26.34​	0.1340​
HAPPY VALLEY	1200​	70.46	11.7433​	21.29​	69.75	11.6250​	21.51​	0.1183​
HAPPY VALLEY	1650​	100.71	12.2073​	14.89​	99.57	12.0691​	15.06​	0.1382​
HAPPY VALLEY	1800​	110.22	12.2467​	13.61​	109.02	12.1133​	13.76​	0.1333​
HAPPY VALLEY	2200​	137.58	12.5073​	10.90​	135.35	12.3045​	11.08​	0.2027​
SHA TIN AWT	1200​	69.56	11.5933​	21.56​	68.74	11.4567​	21.82​	0.1367​
SHA TIN AWT	1650​	99.54	12.0655​	15.07​	98.31	11.9164​	15.26​	0.1491​
SHA TIN AWT	1800​	109.05	12.1167​	13.76​	107.87	11.9856​	13.91​	0.1311​
SHA TIN	1000​	57.13	11.4260​	26.26​	56.01	11.2020​	26.78​	0.2240​
SHA TIN	1200​	70.00	11.6667​	21.43​	69.27	11.5450​	21.65​	0.1217​
SHA TIN	1400​	82.59	11.7986​	18.16​	81.85	11.6929​	18.33​	0.1057​
SHA TIN	1600​	95.43	11.9288​	15.72​	94.59	11.8238​	15.86​	0.1050​
SHA TIN	1800​	108.33	12.0367​	13.85​	107.37	11.9300​	13.97​	0.1067​
SHA TIN	2000​	122.86	12.2860​	12.21​	121.73	12.1730​	12.32​	0.1130​
SHA TIN	2200​	134.88	12.2618​	11.12​	133.66	12.1509​	11.22​	0.1109​
SHA TIN	2400​	150.10	12.5083​	9.99​	147.01	12.2508​	10.20​	0.2575​

 

[The Hare]

So then attached are comparisons between the Official Ratings, My form ratings and my speed figures.

I discarded the first season of results from my figures to allow the ratings to settle. As I say I calculate my Going allowances as just the Median of the difference between today's raw rating and the last 3 ratings given to the horse. This is a little rough and ready, I suppose the last rating is more significant then the one before and so on. This is something I could look into to fine tune the differences but generally it is close. My form ratings are on a similar scale, with the raw differences for beaten runners calculated using the same formula above. The difference is that for these I just use the Median differences based on a race by race basis.

So I took all finishers for the last 4 and a half seasons, deleted all that hadn't finished or were beaten far enough to record a rating of less than -60 for form. I also deleted those horses with no official rating, all Griffin races and some of the Group runners have no rating.

The attached gives the average Official Rating of each horse in each class, those in yellow are were there are less than a 100 runners in the sample. As can be seen there is usually a gap of between 17 and 19 ranking points between Classes 2 and 3 and 3 and 4, which with a spread of 20 rating points is fine.
The average form ratings are my ratings for the same runners as the official ratings. These are slightly different as they include distance beaten and weight but show a close match to the Official ratings. Perhaps indicating the formula is a little incorrect as the table of differences suggests my ratings increase a little compared to the official ratings as distance increases. Maybe just over 1 rating point/pound per distance.
I am using the actual finishing time of each horse in calculating the ratings, so (1500*(difference Standard-finishing time)/finishing time). I think this can be addressed a little by using the standard time or the winners time to calculate this.
But the big difference is with the Speed Figures. As we can see there is a close correlation between the form and official ratings yet my speed figures are much different. So the differences between Classes don't match!

So the Totals page averages the runners in each class using weightings for the runners. It seems that the Speed Figures increase by 1 point for every 2.4 points of Official ratings and Form ratings.

As I say my form and speed use the sane method of 1lb in weight = 1point beaten and the same scale for beaten distances. The only difference is how the Going Allowance is calculated. Maybe i need more manual work on calculating the Going adjustment for speed figures, perhaps a different one for 1000m races at Sha Tin would help but clearly the speed figures can never match the form ones.

I did try non weight ratings and may have another look at those to see if i can work out a better adjustment for weight carried.

When compiling the meeting ratings I use the last 10 runs in the last 365 days, perhaps this is the problem? I will test the effect of looking at different number of races in the past at a later date.

Kevin

 

[The Hare]

Attached are speed and form ratings for tomorrow. Last 10 runs this year for all horses. Ratings adjusted in 1st column.

 

[The Hare]

This attachment gives the runners and there last 10 ratings adjusted for the races tomorrow, Again just the last 365 days, apologies there are a few in the previous Speed ratings page that are a little longer ago than that.

These are just the weight adjusted ratings from the previous attachment. If i go back through my old date then I can analyse the ideal number of races to look back at.

 

[The Hare]

Well not a great performance using the last 10 runs again. Deleting the non runners I make 2 in the top 3 rated Speed and 3 Form. Though the new format of laying out the adjusted ratings for each of last 10 runs shows clearly the best performers latest runs are much better. As we would expect. I think the Form Ratings are much better as the Speed need a better way of calculating the Going Allowance which will be the next task.

Attached are the adjusted last 10 runs in the last 365 days for Wednesday.

 

[The Hare]

Sorry errors! The bain of life. Those figures above are unadjusted for weight Speed and Form ratings.

The full sheet attached here with corrections.h

 

[The Hare]

Well the Speed figures return 1 Top, 1 2nd and 1 3rd top. But the Form was much better at 2-3-2. This gives me the idea that I need a different adjustment for the 1 point for 1 lb for speed figures asa the scale for distances beaten is much less, maybe try that next.
Still form has improved since my method of calculation was adjusted so that's good.
Sunday's card attached.

 

[The Hare]

Reality returns, not sure about the attachments. When I checked after race looked as if picked up wrong figures,

Anyway for the 10 races the Speed figures had 2 top rated 1 2nd and 1 3rd whilst the Form was all Zero!!

It seems looking back 10 races is too many, certainly just the last 3 is much better even for those Form figures.

Card for Wednesday attached and bugs of last meeting removed.

 

[The Hare]

An average day.Over the last 10 races, Speed had 2 Top, 1 "nd and 2 Third. Again the form struggled with 0-1-1 on the same basis. Will eventually check back over last year and give total performances based on number of runs back, but a longish job and need to amend for the change in format to older meetings. Won't be backfitting as the Formula is the only thing changed, all ratings are calculated by computer. Only manual work the first few meetings without ratings.

Anyway card attached for next meeting on Monday the next one. Last 10 runs adjusted for weights set to carry.

 

[The Hare]

Speed returned 1 top, 2 Second and 1 Third top rated. Form poor again at 1-0-1. Much better if don't go back so many runs though.

Card for Thursday attached.

 

[The Hare]

Another poor effort for the last 10 runs. Speed 0-2-1 and Form 0-1-1.

Sunday card attached

 

[The Hare]

Another poor effort - all round! I missed the final race, thought the usual 10 but they had 11.

So 9 rated races and only 1 Top and 1 3rd on speed, the other race was top on speed, honestly! For form just had the 1 top on form and nothing else, missed race included.

All i can think is for speed figures the very sharp bends and shortish run in give less effect of weight and that could be another area to conisder with different values for weights. So many things to try. TO be fair though if I don't go back 10 runs the ratings perform much better.

Card for Wednesday attached.

 

[The Hare]

An amended set of ratings for tomorrow. Thanks to @Outlander for spotting an error in my results for a Dead Heat and all the ensuing ratings in error..

 

[The Hare]

If anyone has been using the results the error is in Race 406 this season with the times against the wrong horses. New attachment for this season results. If anyone would like these updating after each meeting let me know.

Kevin