SPEED RATINGS

[MCLARKE]

I know some of us are interested in speed figures and it has been suggested that we have a separate thread where we can discuss these and publish speed figures / results / methodolgy etc.

[Trotter]

Nice one ! ...... 👍

Any chance you could move our recent posts over just to get it started ?

[MCLARKE]

This is the first post that was posted in the daily chat by @Trotter

 

I don't think there's any way I'm going to keep this up but thought I'd do some speed figures for todays Curragh meeting. I'd like to do ratings for every meeting that includes a Group race but I don't imagine I will .... It'll soon become a chore that I give up on

Here's the ratings of todays winners - highest to lowest

Procrastinate 102, 
Real Force 101,
Brilliant 98,
Money Dancer 93,
Arizona Blaze (2) 90,
Chazzesmee 90,
Merisi Diamond 88,
Global Energy 83

This was run on Heavy ground and my going allowance was 88 ...... so the bulk of all the above figures is going allowance

eg Procrastinate ran a figure of 14 then got his 88 going allowance added on

note that my ratings are in lengths per mile ...... so Procrastinate would be 18 lengths per mile quicker than Global Energy ....... in a mile race I'd expect him to beat him by 18 lengths

[Trotter]

thanks ! ....... 👍

[MCLARKE]

Speed figures are on my to do list. I think I'll start with the all weather, should be simpler than the turf.

[MCLARKE]

@Trotter replied

that's true ! ...... I usually do AW when I bother with them 

In fact I did do AW speed ratings for a good 3 or 4 years a few years ago and found it a good basis for assessing races. It's OK over the winter but during the summer horses flit between AW and Turf so you never have the full picture of how they're running from race to race. The main problem is that there is far too much racing ...... it would be a full time job to rate all races. It might be practical to do it for Irish racing where they have a lot fewer meetings

If you do fancy a crack at compiling your own ratings it might be an idea to have a separate thread where all the speed ratings posts can be grouped together ..... it's a topic that crops up every now and then but the posts get lost in the daily threads

[MCLARKE]

Good idea with a separate spread, I know @Zilzalian produces his own speed ratings. 

Out of interest, where do you get the winning times from ?

[MCLARKE]

@Trotter replied

Just the Racing Post and I use their standard times

I think the 'art' in the process comes from working out allowances ...... as for race times the Post has a huge database so their times should be as good as any and as long as you use the same set of times all the time it doesn't really matter ....... it's not absolute figures that matter it's relative figures between horses !

[Trotter]

I feel a bit obliged to do some more speed ratings now ! ....... 😜

I think I'll do the Doncaster meeting at the weekend in case I decide to compile ratings for all the main meetings which I have to say is a bit unlikely as the season gets underway ........ too much racing !

  ....... and maybe any Irish flat meetings this week just in case I decide to concentrate on the much more manageable Irish flat season with it's much lower number of meetings

Over the winter, say October to March,. the AW is ideal ......... but not much point starting that now

[MCLARKE]

The first big question

Do we adjust for weight !

[Trotter]

9 minutes ago, MCLARKE said:

The first big question

Do we adjust for weight !

personally I don't ........ 

I can see the argument for doing it  to get more precise figures but I reckon you can take that into account when assessing a race...... if you have a couple of horses that look around the same level and one is carrying a stone less than the other today then it's got an advantage. But in practise I don't even take account of the weight carried in the race I'm looking at to make a selection. I tend to stick to handicaps and  let the handicapper do his job and conveniently group together horses that are in the same ballpark to run in handicap races

With all the chaos involved in a horse race and all the factors which affect the final time....... class, field size, going, wind, pace of race etc ....... I reckon a few pounds here or there are probably not too relevant

Obviously I can see why people would be attracted to trying to get figures to the nearest pound but I never have taken weight into account. 

 

[MCLARKE]

I'll have to dust off my Mordin On Time book !

I seem to recall that he concluded extra weight can slow down fast horses but not speed up slow horses

[Trotter]

14 minutes ago, MCLARKE said:

I'll have to dust off my Mordin On Time book !

I seem to recall that he concluded extra weight can slow down fast horses but not speed up slow horses

I've got that book as well .........  I think it was Mordin that first got me interested in speed figures. I think he used to do a column in The Weekender ? ........ late 80s, early 90s maybe

 

 

 

 

[Zilzalian]

12 minutes ago, MCLARKE said:

I'll have to dust off my Mordin On Time book !

I seem to recall that he concluded extra weight can slow down fast horses but not speed up slow horses

You are correct that is exactly what he (Mordin) said.

Initially Mordin didn't believe in taking weight into account but he did eventually concede that he was wrong and weight should be included. (I have a different angle).

My experience with speed figures goes back over 35 years and has served me well.

I will write my thoughts down and post when i am sure i have said all i need to say on the subject.

One thing i will say before everyone starts hacking away at producing speed figures is this. Lets say for the sake of argument that speed figures are 30% accurate for graded races, then i can tell you that for handicaps it would be 8%. So if you want to maximise your time the pool of horses in the group/graded races is very manageable.

Bear in mind there are 14,000 horse in training (jumps and flat) just in Britain.

As to your AW suggestion i am guessing but there maybe more aw runners than turf nowadays which is a big number but serves you 12 months of the year.

 

[Zilzalian]

5 minutes ago, Trotter said:

I've got that book as well .........  I think it was Mordin that first got me interested in speed figures. I think he used to do a column in The Weekender ? ........ late 80s, early 90s maybe

 

 

 

 

Yes you are correct, he stated off his "Systems" column when it was the sporting life weekender and carried on when it was taken over by the Racing post.

[Trotter]

8 hours ago, Zilzalian said:

One thing i will say before everyone starts hacking away at producing speed figures is this. Lets say for the sake of argument that speed figures are 30% accurate for graded races, then i can tell you that for handicaps it would be 8%. So if you want to maximise your time the pool of horses in the group/graded races is very manageable.

Bear in mind there are 14,000 horse in training (jumps and flat) just in Britain.

As to your AW suggestion i am guessing but there maybe more aw runners than turf nowadays which is a big number but serves you 12 months of the year.

 

I do find 'problems' with every attempt to find a sub-set of the horse population to concentrate on

For grade races you'll always find runners that have been running in maidens, handicaps, conditions & listed races, in Ireland, in France ....... horses that you have no ratings for unless you rate every meeting which I'd find impossible

My main focus has always been the Winter AW ...... say October to March. The horses that run there are not running anywhere else so you can be pretty confident of having the full picture for those 6 months

(My usual practice is to rate the meetings day by day and store the winners rating which I can refer back to and take lengths off for the beaten horses ratings. I don't start to rate the race I'm looking 'today' from scratch)

this is how I store my ratings on a text document - 

24/10 Wolv (hcap par 85, F7) - Vespasian 100 (6f c4), Triggered 97 (7f c5), Intoxicata 89 (7f c6), Bankrupt 89 (5f c6), Hannah's Return 88 (7f c6), Master of Combat 88 (9f c5), Elliott 87 (5f c5 2yo), Winterfair 78 (8f c5 2yo), Serengeti Sunrise 70 (7f c5 2yo)

Hcap par figure is the expected average figure that should be run in the meeting's handicaps according to the class of horse. I only use handicaps for compiling allowances which I then apply to the other races on the card.

F7 ------- this is my going allowance. In this case I reckon the track is riding Fast and on average the horses are running 7 lengths per mile faster than expected for horses of this class. So I take 7 points off the raw rating.

Horses name and rating ...... then in brackets the distance, class of race and if it's a 2yo

 

 

 

[harry_rag]

Wish I could find anyone else to talk about a ratings system for goal (or try/touchdown) scorers but I'm more or less ploughing a solo furrow with that one, barring a guy I used to work with who is effectively my "colleague" in the endeavour.

Slightly obvious question (maybe) but I'll ask it to see if any of you fellow stats nerds think differently.

I've started afresh with my data since the new "one spread firm" era began in November and made a slight tweak to my formula to allow for there no longer being two firms offering prices. T'other fella has carried on as before. Doing it his way gives 573 bets with a profit of 37.27 points and an ROI of 6.5% while my new approach gives 414 bets and +29.26 points with an ROI of 7.07%. Those are qualifying selections from an overall data sample of 2348 players.

So I "win" in terms of having the better ROI but have passed up 159 bets with a profit/ROI of 8.01 points and 5.04%.

Would you start backing those selections immediately or stick to the slightly less bets/better ROI criteria for a while longer and see how the data develops. The 159 bets is still a relatively small sample size.

[Torque]

14 minutes ago, harry_rag said:

Wish I could find anyone else to talk about a ratings system for goal (or try/touchdown) scorers but I'm more or less ploughing a solo furrow with that one, barring a guy I used to work with who is effectively my "colleague" in the endeavour.

Slightly obvious question (maybe) but I'll ask it to see if any of you fellow stats nerds think differently.

I've started afresh with my data since the new "one spread firm" era began in November and made a slight tweak to my formula to allow for there no longer being two firms offering prices. T'other fella has carried on as before. Doing it his way gives 573 bets with a profit of 37.27 points and an ROI of 6.5% while my new approach gives 414 bets and +29.26 points with an ROI of 7.07%. Those are qualifying selections from an overall data sample of 2348 players.

So I "win" in terms of having the better ROI but have passed up 159 bets with a profit/ROI of 8.01 points and 5.04%.

Would you start backing those selections immediately or stick to the slightly less bets/better ROI criteria for a while longer and see how the data develops. The 159 bets is still a relatively small sample size.

Generalising massively, and from a purely statistical standpoint, I'd be a lot happier with a 5% return after 1000 bets of one system than a 10% return after 500 bets of another. The figures you mention above suggest there's not much in it between the two systems, at least at this stage anyway. They could very well come together more or less with enough bets.

[Zilzalian]

12 minutes ago, harry_rag said:

Wish I could find anyone else to talk about a ratings system for goal (or try/touchdown) scorers but I'm more or less ploughing a solo furrow with that one, barring a guy I used to work with who is effectively my "colleague" in the endeavour.

Slightly obvious question (maybe) but I'll ask it to see if any of you fellow stats nerds think differently.

I've started afresh with my data since the new "one spread firm" era began in November and made a slight tweak to my formula to allow for there no longer being two firms offering prices. T'other fella has carried on as before. Doing it his way gives 573 bets with a profit of 37.27 points and an ROI of 6.5% while my new approach gives 414 bets and +29.26 points with an ROI of 7.07%. Those are qualifying selections from an overall data sample of 2348 players.

So I "win" in terms of having the better ROI but have passed up 159 bets with a profit/ROI of 8.01 points and 5.04%.

Would you start backing those selections immediately or stick to the slightly less bets/better ROI criteria for a while longer and see how the data develops. The 159 bets is still a relatively small sample size.

Seems to me you have the solution already Harry, you both carry on until you are convinced one is more practical/profitable than the other. I had a similar dilemma, i have two options with my speed figures the first is my own speed figure, the second is to add  "stopwatch" (Dave Edwards) speed figures from the R Post, so who's to use? Mine on their own or the combo? Mine beats "stopwatch" (because he publishes his in the RP and that in itself affects the price) but many times the combined beats mine so i often split the stakes depending on my confidence in the selection or if mine seem too far away from his or common sense. Having said all that i do prefer using my own.

[harry_rag]

Thanks @Torque. I've taken it a stage further, so I'd be interested how you feel now (and welcome any other take on which is the best option). Basically we did 2 things when SPIN disappeared and left us with a single set of prices. We increased the "magic number" from 72 to 73 and I adjusted how the figure was calculated to try and more accurately reflect what it would have been if there were still 2 firms. In simple terms imagine 72 or 73 being the rating figure required for you to have a bet.

Scenario	Rating	Basis	Bets	P/L	ROI
1	73	Old	573	£37.27	6.50%
2	73	New	414	£29.26	7.07%
3	72	Old	690	£41.77	6.05%
4	72	New	510	£39.72	7.79%

Scenario 1 is what my mate does now while scenario 2 is what I'm doing. I'd say his approach is best as he's making more money from a greater number of bets even if the ROI is slightly lower.

Scenarios 3 and 4 are what if you left the number at 72 with either the old or new calculation basis. I'm inclined to rule out scenario 3 due to it being too many additional bets for too little benefit. I already feel like I place enough bets for an increase of 66.66% to be over the top (this is from 6th November onwards for context). I'm attracted by option 4 though. Best ROI, close to best £ return, 2nd lowest bet volume but still an increase of 23%. There is a logical argument to say either increase the number OR adjust how it's calculated but not both as you duplicate the effect.

So my conclusion is to change from scenario 2 to scenario 4 for now and see how it goes while monitoring all of the above as the data sample gets bigger. (Bear in mind I'd accumulated data for just over 6000 players back when there were 2 sets of spread prices to play with.)

 

[harry_rag]

 Forget about the above, I've sent myself down a bit of a rabbit hole! All of those numbers are true in terms of the criteria in question but there's another criteria that comes first so I'll already be betting on quite a few of the supposedly "missed" bets anyway. I need to look at it again, sorting the data rather than just analysing it the way I've done today. There may be some uplift to be had but it won't as significant as I first thought.

One thing I would say is that the initial criteria I refer to above is now less profitable than the secondary ones I waffled about in the previous posts whereas it always used to be the other way round. I need to get my head round that. Bottom line is it's still profitable despite losing one of the spread firms for input but this was a cack-handed first attempt to analyse the "new era" data.

Anyway, about those speed figures... 

[harry_rag]

After all that kerfuffle and waffle I'm on Smith at 21/10 to score for Bradford tonight. Old rating is 74 but new one only 72 so it's a perfect example of a bet I'd do as a result of the mooted change. Surely it's in the stars? 

[MCLARKE]

5 hours ago, harry_rag said:

Wish I could find anyone else to talk about a ratings system for goal (or try/touchdown) scorers but I'm more or less ploughing a solo furrow with that one, barring a guy I used to work with who is effectively my "colleague" in the endeavour.

Slightly obvious question (maybe) but I'll ask it to see if any of you fellow stats nerds think differently.

I've started afresh with my data since the new "one spread firm" era began in November and made a slight tweak to my formula to allow for there no longer being two firms offering prices. T'other fella has carried on as before. Doing it his way gives 573 bets with a profit of 37.27 points and an ROI of 6.5% while my new approach gives 414 bets and +29.26 points with an ROI of 7.07%. Those are qualifying selections from an overall data sample of 2348 players.

So I "win" in terms of having the better ROI but have passed up 159 bets with a profit/ROI of 8.01 points and 5.04%.

Would you start backing those selections immediately or stick to the slightly less bets/better ROI criteria for a while longer and see how the data develops. The 159 bets is still a relatively small sample size.

I would calculate the probability value (you need to know the average odds for this)

[harry_rag]

9 minutes ago, MCLARKE said:

I would calculate the probability value (you need to know the average odds for this)

Thanks, one of those things I do occasionally then forget how to do! In reality it turns out I missed out on 55 bets that showed a profit of 0.4 points to 1 point level stakes so I could have backed 110 bets with an ROI of 18% but instead backed 55 with an ROI of 36%! Obviously we’re into tiny sample size territory there. Overall the indicators support the change I was thinking of but the impact in terms of additional bets will be much smaller than I originally thought.

[MCLARKE]

22 minutes ago, harry_rag said:

Thanks, one of those things I do occasionally then forget how to do! In reality it turns out I missed out on 55 bets that showed a profit of 0.4 points to 1 point level stakes so I could have backed 110 bets with an ROI of 18% but instead backed 55 with an ROI of 36%! Obviously we’re into tiny sample size territory there. Overall the indicators support the change I was thinking of but the impact in terms of additional bets will be much smaller than I originally thought.

With my AW systems last year I had plenty that had shown profits in the "test" year and therefore I could use live. I ended up with loads of selections and overall made a loss.

This year I have used the system that had the highest p value and I have ended up with much fewer selections but a substantial profit.

[calva decoy]

How do you take into account a national hunt chase race where for instance 2 horses of the 8 entered in their last run only jumped 4 of the 8 fences due to a low sun compared to the other 6 who jumped all of the 8 fences & say over a 2 mile trip ? 

Whatever the answer it'll turn out to be a false narrative because of the 25% from the last runs inequalities & being speedier because of the less jumping needed .

The only time I used speed figures is via Equibase in US dirt sprint races between 5-6f & it's been quite productive along with low draws in those races but , each to their own & a good thread for discussion 👍

[MCLARKE]

6 minutes ago, calva decoy said:

How do you take into account a national hunt chase race where for instance 2 horses of the 8 entered in their last run only jumped 4 of the 8 fences due to a low sun compared to the other 6 who jumped all of the 8 fences & say over a 2 mile trip ? 

Whatever the answer it'll turn out to be a false narrative because of the 25% from the last runs inequalities & being speedier because of the less jumping needed .

The only time I used speed figures is via Equibase in US dirt sprint races between 5-6f & it's been quite productive along with low draws in those races but , each to their own & a good thread for discussion 👍

I think I'll concentrate on the flat, there are enough question marks there !

[Trotter]

I wouldn't bother with speed ratings for Jump races to be honest ......... the final time would be so dependant on the pace of the race that they would be meaningless. The same applies to an extent on the flat as well but over much shorter distances the relative importance is less

I guess speed figures should be most useful over sprint trips on the flat

An analogy with human athletes ........... 100 metre races are run pretty flat out so 'time' is the most important factor in who should win ...... just who runs the fastest !

But over 1500 metres you also have race tactics, pace ..... will it be a slow run race to suit the athlete with a finishing kick or will some competitors set a fast pace to draw the sting out of the fast finishers

The longer the horse race you'd expect the race pace and tactics to be more important relative to pure speed

The way I look at speed figures is that all horses can run slow but not all horses can run fast ......... high speed figs indicate horses that can run fast relative to other horses and they'll always have an advantage because it opens up all the race tactics if you know you can run faster than the opponents (I can see the value of sectional times here to identify horses who can run the last couple of furlongs really fast)

[Zilzalian]

34 minutes ago, Trotter said:

I wouldn't bother with speed ratings for Jump races to be honest ......... the final time would be so dependant on the pace of the race that they would be meaningless. The same applies to an extent on the flat as well but over much shorter distances the relative importance is less

I guess speed figures should be most useful over sprint trips on the flat

An analogy with human athletes ........... 100 metre races are run pretty flat out so 'time' is the most important factor in who should win ...... just who runs the fastest !

But over 1500 metres you also have race tactics, pace ..... will it be a slow run race to suit the athlete with a finishing kick or will some competitors set a fast pace to draw the sting out of the fast finishers

The longer the horse race you'd expect the race pace and tactics to be more important relative to pure speed

The way I look at speed figures is that all horses can run slow but not all horses can run fast ......... high speed figs indicate horses that can run fast relative to other horses and they'll always have an advantage because it opens up all the race tactics if you know you can run faster than the opponents (I can see the value of sectional times here to identify horses who can run the last couple of furlongs really fast)

"I guess speed figures should be most useful over sprint trips on the flat" The most useful speed figures are not really in the sprints (usually all very similar) what you are looking for is outliers beyond 8F.  Just one Great example of this was Alpenista, go look back at its form, it came to my attention when it beat a horse i really fancied in a time that was exceptional for that day so it went into my notebook, go forward and Prescotts big mistake was not going for the arc the year before it won it, the arc, the one that torquato tasso won. (my big mistake was not backing Torquato to win the arc in the absence of Alpenista, 100/1 in places)  Alpenista had beaten that horse comfortably on two occasions in Germany. the point i am making here is that i made a lot of money off Alpenista just by noting the time anomaly (big speed figure) and following it thereafter.

[harry_rag]

For any of you stats types who swing both ways (footy as well as racing) I've just posted this on the subject of the value of xG and would welcome any thoughts. It's been puzzling me for a while.