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

Early days

@TheBluesBrother

I've rerun the Stratford races using the times in your Standard time file, and get very similar comptime figures to kick off with that then diverge when I do similar calcs ... if I use your -0.25 figure for the allowance I get very similar values right across the board for adj/adjcomp/lbsLength and div/0.2 (or x by 5 as I tend to figure it) but my final ratings are rather different, Mister Universum for example is coming out at 21 (my ratings put him at 69, so I suppose I'm circling you at the moment....)

Going back manually over the calculation, because I don't see how I can have all the different values within a hair of yours and end up with half the final rating.......isn't your calculation basically to take the div/0.2 figure, multiply it by lbs per length, subtract the result from 100 and then compensate for weight? For Mr U I'd make that 100-(86.63 * 0.92) which is 100- 79.7 = 20.3, with the horse carrying 11-0 there's no weight to add/subtract, so isn't that a final rating of 20? Are you then adding a correction in for the difference between the average OR and the top 3 ratings or something?

Sorry to be a pain, I'm obviously missing something simple here - probably as a result of staring at the screen too long again....
Enjoy the fishing!

After 45 mins in bed I simply had to fire the PC up to explain that I had a lighhtbulb moment.
It's all okay, you are working to a base rating of 130 of course while I was still using 100 - suddenly we are in agreement.... Dohhh again!

Dave
 
Last edited:
Hi davejb davejb ,
Good luck with this project.
My 14 year old son has just started to learn Python.
I may have to Bribe him to help me with my Ratings.
Donny.
 
Python's okay,
just a bit annoying when you want to use multi dimensional arrays and find they're not built in.... it's a nice enough language, quite easy to get into etc and very much like many others, computer languages get easier each time you learn a new one.

I started when I was a mere lad in my mid 20's back in the early 1980's, and although I've done the odd Open Uni course in the distant past I still rely on a blunt instrument approach to programming for most of what I do - I think I self taught myself most of the bad habits pretty quickly :)

Dave
 
you are working to a base rating of 130 of course while I was still using 100

You are correct, flat 100 (9-0) and national hunt 130 (11-0).

The major problem you will face, especially when it comes to the Irish scene is the advertised race distances,
for example, when Galway had a private company in to measure their race distances, they found that their 2m1f chase distance was 2m2f.

I just reminded horse racing Ireland this week, that SIS were supposed to be supplying data from GPS for the correct race distances for all Irish
racecourses, this was supposed to start on Jan 1st 2017 but never happened.

Off fishing again.

Mike.
 
Last edited:
Irish racing - not a problem, you just rely on the RP standard time and comptime values for that and....
.... well, maybe not then:eek:

Actually although I include the Irish results and cards in what I do, I didn't initially - I just found so many Irish raiders popping up that they became too frequent to ignore, so I had to start trying to account for them. I treat Irish ratings with a degree of suspicion, it's difficult enough to make the british form ratings reliable after all. I think in general the racing industry (like many others) is sometimes over protective of its traditional bad practises, we just have to do what we can to mitigate their effect.

Dave
 
is sometimes over protective of its traditional bad practises

The racing professional who I refer to as "the sheep" wouldn't even realise that yesterday meeting at Bellewstown was so bad,
the 2 x 2m½f(H) races had to have been run over a furlong short.

Mike.
 
Sorry if this is a pain,
could I just check a couple of things from your Stratford example above? Assuming I am using the same spreadsheet as yourself, which I believe I am, there's a timing query and one on your method compiling this example,
The timing - as far as I can tell you are using the screenshot copy from the RP site, as per your video, but compiling your own comptime from your standard times for Stratford, then applying the railmove correction - in the first race for example the RP comptime is 15.2, using their racetime of 4:1.20 and your standard time of 3:37.0 produces a comptime of +24.2 (9s slower than the RP version) which then reduces to +21.4s with a railmove allowance of 2.8s for 42 yards? I'll be somewhat amazed if this isn't your calculation, I just wanted to confirm that.
As ever, disregarding the RP standard times and using yours, between RP, Timeform, and HRB each race gets three different comptimes (the fifth race is timed at 5:33.9, 5:36.0, and 5:37.3 across the three, so up to almost 3.5s difference depending on where you look).
I'm using HRB, I guess it would be very nice to know if anybody has accurate figures, you'd expect that might be Timeform, but ultimately I have to hang my hat on one of them!

The second query is again a case of 'I think you did so and so' and I'm just checking that's so, otherwise I can't figure out why your numbers and mine are noticeably adrift. I think it was yourself who posted a youtube video on the compilation of the ratings in Excel using the 'Analysis of winning times' from the RP site, and during it you said you sometimes moved some of the OR ratings out of the averages calculation if you thought, based on experience, they were going to unduly slew the results. I can only get close to your figures, with an allowance of -0.29, if I remove 4 of the 6 ratings there - specifically the 103, 136, 122, 114 - leaving the program to average the 80 and the 85. Could you confirm that is either how you ran this particular meeting calculation, or at least close to it, as it would explain why my program (and my own manual attempts in Excel) are so different.... leaving all the ORs in I get Mr U to be 70 and the allowance at least -0.68 I recognise that this is an unusually high figure, leaving me to wonder if Stratford has become the latest AW track based on concrete.... without those 4 high ORs my numbers pretty much match yours.

Dave
 
davejb davejb

The video I put up years ago is not the way I work out the going allowances these days, I use a few more tools now, the video was only a rough guide for somebody who asked.

Since Apr 2013, Racing Post will time a national hunt race when they cross the starters line, which is the way Timeform also do it.

With regard to Stratford I could not follow what you were asking me.

Rail Movements:
Races 1 & 6 +42 yards (subtract 2.8s)
Race 2 -6 yards Races (add 0.4s)
Races 3 & 5 -24 yards (add 1.6s)
Race 4 +72 yards (subtract 4.8s)

My standard times adjustments (comparison per furlong)

Strat1.png

Rail movement adjustment.

Strat2.png

Going allowance adjustment of -0.25s/f (good)

Strat3.png

If you are still confused just ask one question at a time.

Mike.
 
Last edited:
Sorry,
too involved in it all I guess! I think I've figured the 2 things out by steadily bashing the computer until it squeaked....

I've been trying to figure out why my ratings for Stratford are up to 20 higher than yours. I worked out which times you were using okay, it's RP race times + your own standard times + railmove calculation, my first question was asking you to confirm that but it'd be highly unlikely to get the same values any other way so no worry.

My second query was based on your method of calculating ratings, which adjusts the allowance until the top 3 ratings match the OR average for the meeting - this is the method I've been using, initially via Excel then via my program. I think it's simply that you are judging the going allowance separately, then plugging this value in, so you get -0.25s for the meeting while I'm getting -0.51s. which is causing the ratings to be so different. Mystery solved! I thought originally you were removing some of the winner OR's from the average calculation but you have the same value of 113 that I do, so it couldn't be that.

As I'd be absolutely guessing when it came to applying a going allowance any other way I'll keep plugging along - as I've made a few small tweaks along the way (not least shifting NH to a base of 130) I'll run my database back through the PC to update all the ratings. Thanks for the help, although my ratings are still higher than yours they're not quite as astronomical at Startford as they were...

I'm attaching my spreadsheet for july 4, it's a sort of intermediate file my program produces after calculating ratings, so you can see the numbers that I've been getting, in case you are interested.
Dave
 

Attachments

  • daily_ratings.csv
    4.1 KB · Views: 8
Got a big bug to squash, won't be doing ratings this evening - if I get it sorted in the morning I'll post then, but I've had enough for one night!
Today's ratings were down on level stakes, 56 races and only about 47 pts back overall, the Larkspur ratings made something like 67 pts.
Dave
 
you get -0.25s for the meeting while I'm getting -0.51s.

The major problem with National Hunt, the majority of races are not true run (slow pace), so you have to fit the going allowances to suit.
If for arguments sake, your top rated horse has a speed figure of 120, and you look back at his ratings and the best figure you have for him
is only 105, look at other horses and races, and you more than likely see a pattern emerge, your going allowance is too generous giving the
horses inflated speed figures.
You can use Dave Edwards speed figures as a guide, the only difference on the flat is he has to use WFA.

Mike.
 
Hi,
crikey, you're up later than I was - or is that a bit of dawn fishing?
Yes, that's the disadvantage of having a program doing most of the work (though the time I've spent programming and reprogramming it you'd not consider this labour saving at the moment), it can be difficult to tweak things, although it would be possible to amend stored ratings after doing a manual check - I have a section of code trying to spot the really silly ratings, but often it depends on context (a 120 rating isn't that astounding, but it's a bit unlikely in a class 5 handicap at Chelmsford).

I should say "A" disadvantage, as I've been working away for about 16 of the last 24 hours reprogramming to put the NH to base 130, then recompiling the database - most of that time squashing bugs, I even had to watch the Eclipse as a replay I was so caught up in hammering things into a semblance of shape. Mind you, even as a replay it was a damn fine race...

Right, having got everything working again - and removed the 450 rated runner (some random number crept in there) - I finally noticed how little racing is on tomorrow, so here's Ayr and Market Rasen, whilst the ratings file covers these two plus Limerick. You'll all know when things are running more smoothly, I'll not be having t post at midnight....

Dave
 

Attachments

  • market_rasen.csv
    29.8 KB · Views: 1
  • market_rasen2.csv
    20.7 KB · Views: 1
  • ayr.csv
    71.9 KB · Views: 2
  • ayr2.csv
    34 KB · Views: 1
  • ratings.csv
    3.4 KB · Views: 5
On another note,
yes, Irish racing looks bad - it's surprising that there isn't at least some interest from those involved in getting at least some of the information close to reality - I presume they figure that they are getting some sort of advantage from it all, or are just lazy. I suspect if you turned up at a meeting with a trundle wheel and started pacing along the course they'd clap you in irons....

I don't have any RI jumps form pre April, but I do include the HRB form since that date as I process the daily results file each day. I do have RI flat for years and years, but have to edit it quite dramatically to fit it into my database format, and it's very noticeable that when the program is running results in it's the Irish meetings that get flagged up for having problems - the RI data for foreign racing can also have problems to resolve, there's an occasional shortcut where the race time and the standard time for a race are identical (presumably to avoid having an empty data field there). The good thing is that you get used to most of the more common fudges quickly, and get very good at spotting and correcting them....

Although most of my time is of necessity spent programming and fixing stuff, there should be a point in the not too distant future where I can leave the program to do the donkey work which will free me up to spend time actually looking at what I've got, and I would hope to improve my ability to adjust going allowance and similar as a result - for now my program is probably doing it as well as I would be likely to do manually.

Thanks for the interest, I know you get quite a lot of queries as folk like me pop up, so I'll try to avoid bothering you too much!
Dave
 
Yet another rearrangement - I DID say this was work in progress,
I'm uploading the two versions of each card for Ayr, Ripon and the jumps at Worcester - and instead of a large ratings listing for every race I've incorporated the ratings section into the top of each 'card2' csv, so for example ayr.csv has the runners listed with their single lines of info (wt, headgear, rating, OR, trainer, jockey etc) for each race, followed by all form lines going back to 1/4/16 in a section under each racecard.
ayr2.csv is the cut down form version, which now has all the ratings for that meeting (so that'll be ratings for ayr then) the same single line info as the ayr.csv has, but the form lines underneath are only the ones where the horse has won or placed - the idea, as described earlier, being to make it easier to spot whether a horse has win/place form over distance, going, class etc. I forget who the genius was who said this, but I hold firmly to it, (I misquote dreadfully however) 'never bet on a horse to do something it's never done before'.... now obviously some horses win first time out etc, but I think the principle is sound when assessing form.

Anyhow, enjoy the cards.

I've spent the day filtering my data a bit, so recompiling the trainers/runners/races database (not that bad a job, it's all fairly slick now) for a couple of hours. I added alerts to identify when timing information looked unduly suspect, and I'm sure Mike TheBluesBrother TheBluesBrother will not be amazed to hear that I spent most of the checking time going over Irish timings. The program checks for the comparison time being within a fairly generous fraction of the standard time, generally speaking errors in reporting times (I found a few) look to be mainly typos on behalf of whoever compiled them, and easy to find with even a gross error check - things like a 12f race being run a minute faster than standard, or a 5f sprint that took 2m 15.0s. Needless to say a few (yes Irish) races had truly humungous variations, I'm pretty sure 5s a furlong slow is not a reflection of the truth, no matter how heavy they say it was....
There are also checks for stupendously good or bad ratings - these often resolve back to timing errors also, so (sadly) I won't be finding any more Secretariats in the ratings....I've been cross checking database times with Timeform and RP times, tending to go with TF when in doubt, so I'm pretty hopeful that the ratings are getting closer to where they ought to be.

Yesterday's ratings, in case anyone was wondering, had selections in 22 of 23 races (2 x NR in the other) and the ratings had 6 winners, although one was at 1/12 so I doubt that will set the blood racing. A couple at 7/1 really helped, the return was about +50% overall. I've noticed that sometimes the ratings throw up a big priced outsider, yesterday's first runner at Limerick came third at 33/1 for example, so I'm keeping an eye out for what the big priced selections are up to. Obviously even though backing all the selections seems to do reasonably for ROI it's a big outlay and an odd bad day would wipe you out, so I'm going to be looking more closely at filters in the near future to try to reduce things to a manageable size.

Now, dinner awaits....
Dave
 

Attachments

  • ayr.csv
    79.1 KB · Views: 6
  • ayr2.csv
    36.7 KB · Views: 1
  • ripon.csv
    30.2 KB · Views: 1
  • ripon2.csv
    18.1 KB · Views: 3
  • worcester.csv
    32 KB · Views: 2
  • worcester2.csv
    21.5 KB · Views: 2
I spent most of the checking time going over Irish timings. The program checks for the comparison time being within a fairly generous fraction of the standard time

What Irish racecourses did your program flag up as suspect?

I have ask Dave Edwards many times to do something about this but all he does is sit on the fence, he just prefers the quite life.

There was a major problem with Market Rasen yesterday, the clerk of the course rail movements calculations are suspect.

Mike.
 
Last edited:
They're mostly from the heavier going in April over jumps, for example:
1/4/17 14:15 Navan nearly 51s slow (about 25%) over 2m hurdles, others at same meeting also all about 25% slow
14:50 52.1s slow 2m hurdle again
16:00 2m 41/2f hurdle 85.1s slow
16:35 2m 4f chase 69.2s slow
17:10 2m 4f chase 64.7s slow
17:40 2m NH Flat 56.8s slow
- 25% slow seems a bit excessive to me, unless they were in a complete bog

Cork on the 2nd April has a similar time in the 14:35, a 5f flat race that ran 12.68s slow

I've got similar for a bunch of flat times from my 2016 data, Limerick a couple of times on 7/4/16 an odd one for Gowran Park and Sligo, and just about every race at Tipperary on 10/5/16
Where I've had the same sort of thing flagged up for British courses a quick check of RP and Timeform has shown that my downloaded data has had the odd typo in it for the most part, which has been easy to correct - as my program will flag these issues up on loading the daily results it'll be an occasional quick check to fix and that's no bother (it was just a bit annoying to do it for 15 months of data in one go - the RI data covering the first 12 months), but when I check the Irish times out everyone agrees these were the times posted on the day.

My conclusion is that the going was incredibly heavy (heavy going for UK races is not producing anything like the same amount of braking to the runners), they've been running over anything up to a furlong further than stated, or they've put a PP9 battery into their clock instead of a couple of AA's and the thing is going round 25% faster than it should.

Dave
 
Cork on the 2nd April has a similar time in the 14:35, a 5f flat race that ran 12.68s slow

Cork.png

You might notice a difference in the comparison per furlong times for 5f, this is because the standard times have recently been changed.
The new standard times for 5f at Cork is 56.7s.

55.9
56.68
56.89
57.0
57.2
57.43
57.56
57.6
57.88
58.12

30th perecentile = 56.9s

The going on the day was heavy @ -1.23s/f (heavy).

Turf going allowance table:
Firm +0.55s/f to +0.63s/f
Good/firm +0.20s/f to +0.53s/f
Good -0.25s/f to +0.18s/f
Good/soft -0.55s/f to -0.28s/f
Soft -1.00s/f to -0.58s/f
Heavy -1.58s/f to -1.03s/f
 
TheBluesBrother TheBluesBrother I'm attaching the ratings my program produced for yesterday, based on HRB data (except last race time changed to 4m 7.80s as per TF, HRB had it at 4m8.70) in case you find it at all useful for Market Rasen - applying my limited abilities to that card the first 2 races (2m 4.5f Hurdles) both return very low ratings. I can understand the high rating for the class 2 race over 2m 7f is feasible, and the Novice chase (4th race) looks okay as the winner was quite a decent horse, in race 5, the 2m 5.5f chase, the winner ran a bit of a low rating but that might still be okay. The first 2 however seem abnormally low considering the winners' are both rated almost 100 higher on OR whilst races 2,3 and 4 all ran within a reasonable distance of their ORs.... in the last race an amateur riders race with a winner getting within 20 of their OR doesn't strike me as too odd.

Dave
 

Attachments

  • daily_ratings.csv
    2.6 KB · Views: 24
Back
Top