June 6, 2007 at 13:52 #1871
Not sure this is the right area of the forum to post this but here goes anyway.
IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢m trying to come up with a means of assessing probabilities by using RPR as a guide. The idea is to come up with an opinion on the rating a horse is likely to be able to run to given the conditions and then by twiddling about with the Poisson function on Excel try to work out the probabilities of each horse running to that rating or above. It is all a bit rough and ready at the moment.
My question is this: are the RPR recorded by each horseÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s name in the results on the RP website adjusted for weight and if so, is it to 10 stone?
Thanks<br>June 6, 2007 at 15:33 #63553
Sounds like a good embryonic idea. Converting ratings into probabilities is something that has often been discussed on this part of the forum without any definitive conclusions.
My own rule of thumb is to reckon that one ratings point on the RPR scale is equivalent to four percentage probability points. I’ve covered this in more depth on the ‘RP Ratings Revisited’ thread.
To answer your main question:
The RP ratings that appear in the results on the RP website are on the scale 0 to 140.
The 2yos are given ratings relative to other 2yos and the 3yos are given ratings relative to other 3yos.
When 2yos and 3yos(and early season 4yos) meet older horses they receive a weight for age allowance(WFA) to compensate for their immaturity.
If you look at the RPRs on the racecard, IN HANDICAPS and MIXED AGE races, they are adjusted to 10st for mature horses(4yos from as late as July onwards depending on distance) and to 10st minus WFA allowance for all other horses, in practice mainly 3yos.
In races where all horses are 3yo or 2yo, they are adjusted to 10st.June 6, 2007 at 15:56 #63554
Thanks for that response Artemis
So just to clarify: lets say IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢m looking at a 4yo horse in a handicap for 4yos+ From his previous results, I can see that the best RPR heÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s achieved is 67. IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢m happy that under the current conditions he can reproduce that RPR again. In the race in question, he will be carrying 9-10. What I have been doing is then adjusting that rating to take into account the 4lb less than 10-0 he will be carrying in todayÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s race ÃƒÂ¢Ã¢â€šÂ¬Ã¢â‚¬Å“ so his rating for the race I am looking at is 71 and making similar adjustments for all of the others too.
Does that sound sensible?<br>June 6, 2007 at 17:54 #63555carlisleMember
- Total Posts 772
A horse’s best RPR can sometime have been achieved on soft ground, 7 furlongs at Ponefract…..when asked to run a fast ground 5 furlongs at Epsom.
In this context the word "best" would have a very hollow ring to it.
Artemis, thanks for the Topspeed stuff.
cheers from<br>carlisle<br>June 6, 2007 at 21:24 #63557
I’d be interested on your views on the Poisson function, there are several ways of pricing races up very much dependant on your unit of measure .. as Artemis says; everyone pretty much has their own beliefs and ways of going about things, if you want a few opinions you’ve come to the right place.June 6, 2007 at 21:28 #63558
I’d be interested on your views on the Poisson function, there are several ways of pricing races up very much dependant on your unit of measure .. as Artemis says; everyone pretty much has their own beliefs and ways of going about things, if you want a few opinions you’ve come to the right place.June 7, 2007 at 08:33 #63559
That’s how it’s done.June 7, 2007 at 09:45 #63560
Carlisle, you are right, I’m not just looking at taking the best RPR, but the best I think the horse is capable of in the current race conditions, which won’t always be the best it has achieved in the past. And I’m adding a few pounds for improvement, taking a few away for a poor draw, making my own judgement, essentially.
I expect to fine tune the adjustments over time, but for the moment am working with the idea (taken from Braddock’s book) that an average horse can improve maybe 10-14lb as a 3yo and making smaller adjustments for draw, jockey, pace bias etc. All of this is pretty rough and ready at the moment.
On the poisson idea, I should stress that my mathematical knowledge stopped developing the day after my Maths GCSE exam. However, after much brain-straining I managed to get a rough idea of the way the function worked on Excel.
Let’s say my ratings for a 3-runner race, after all adjustments for conditions, weight carried etc. have been made are:
Horse A 110<br>Horse B 105<br>Horse C 101
I would try to use Poisson to find the probability of each horse in the race running equal to or above the top rating (110). As I understand it, the following formula
POISSON(110,110,TRUE) in Excel will give you the probability of the 110 rated horse running to a rating between 0 and 110. I then deduct this from 1.0 to find the probability of it running over 110. Doing the same for the other horses gives their probabilities of running over 110. Then all I would have to do is divide the individual probabilties into the total probability for each so that I have a 100% book.
There was a far more rigorous approach suggested by Steve a while back which involved calculating the probability of each horse beating each other horse for each of the possible ratings from 1-140 (as I remember it) but it looked horrendously complicated) so this is my attempt at a rougher but speedier version, which I can then tweak to produce a tissue.
I am hampered by only having Excel at work – at home I have Works Spreadsheet which doesn’t seem to have a Poisson function.
I have tried this for a couple of races, and came up with a reasonable tissue, but it is in its early stages. I might have a go at the Sandown race mentioned on the other thread.
Any suggestions or criticisms from the more mathematically minded on here would be well received.June 7, 2007 at 14:09 #63561
Here is my attempt at the race mentioned on the other thread
7:55 Sandown Rating Ãƒâ€šÃ‚Â Decimal Odds<br>Hassad 107 9.52<br>Full Victory 106 10.36<br>Master Pegasus 105 11.35<br>Marajaa 104 12.53<br>Fabrian 104 12.53<br>Merry – 104 12.53<br>Killena Boy 103 13.94<br>Tumbleweed 103 13.94<br>ItÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s a – 102 15.62<br>Voliere 101 17.66<br>Langford 101 17.66<br>General Knowledge 101 17.66<br>Councellor 99 23.14<br>The Snatcher 97 31.45<br>Pintle 94 53.67
(Some of the missing names are due to my abbreviating them at home before completing the tissue at work)
As you can see the ratings are all fairly tight. But in a handicap, IÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢m thinking that should be expected. I may well have underestimated Hassad. But what is clearly wrong is the conversion from ratings to prices. Maybe in the Grand National an 8-1 favourite is not impossible, but this is merely a fairly competitive flat handicap and it just looks wrong.
More work required I think <br>
(Edited by Aranalde at 3:12 pm on June 7, 2007)June 7, 2007 at 18:32 #63563
To get a tissue, ratings by themselves are not enough.<br>You nearly always end up with something similar to your odds for the 7.55.
The best way around this, IMO, is to use the ratings and also factor in the market. I do this by awarding each horse a number of points according to its best available price at as near to the off as possible. If you are leaving the house at 10.00 am, the time to do this is probably about 9.00am. If you are home all day, leave it as late as possible.
The scale I use is:
current price(expressed as a % probability) divided by 4
So: Evens = 12pts(50/4), 6 to 4(40/4) = 10pts, 3/1(25/4)= 6pts, and so on.
Add these to your RPRs and then recalculate and you will get something closer to a tissue. The original ratings are tempered by the actual market.June 7, 2007 at 19:28 #63564
That would certainly give a more realistic tissue, a useful idea. I’ve never been keen in factoring in the market, still chasing the idea that I can somehow produce a more accurate tissue than the market often enough to make a profit. However, that involves more work and time than I have so far been able to muster, so perhaps factoring the market in will make things more realistic.
I am still fascinated by the fact that people like Dave Nevison are able to create a tissue through form study and their own knowledge/observations, that enable them to identify enough mispriced horses to make a living. When I first read Nevison’s column, I found his lumpen prose style and drinking stories tedious. But bits and pieces I’ve read about his methods have intrigued me. It sounds to me as though he (and presumably others) draw up their tissue the night before and so probably don’t take the market into account. What also fascinates me is that he is (apparently) able to price up a race so accurately he is prepared to back 50-1 shots that he reckons should be 33-1 and so on.
Ultimately I would like to do away with even using RPR and produce my own ratings but that requires time I can’t spare at the moment.
I have in the past had a go at predicting SPs and found it fiendishly difficult. Guess there is nothing for it but continual practiceJune 7, 2007 at 20:18 #63565
I don’t particularly like factoring in prices, not saying that it’s wrong .. but the price must be derived from something else and not from itself.
A way you could price a race up is like this…
Set the top rated value as 20 and then deduct the difference of the second rated from 20 until you reach 0 or you have covered the top 6 or so of the card. So your ratings would be ..
Hassad 107 20 9/2<br>Full Victory 106 19 9/2<br>Master Pegasus 105 18 5/1<br>Marajaa 104 17 5/1<br>Fabrian 104 17 5/1<br>Merry – 104 17 5/1<br>Killena Boy 103 <br>Tumbleweed 103 <br>ItÃƒÂ¢Ã¢â€šÂ¬Ã¢â€žÂ¢s a – 102 <br>Voliere 101 <br>Langford 101 <br>General Knowledge 101 <br>Councellor 99 <br>The Snatcher 97 <br>Pintle 94
20 + 19 + 18 + 17 + 17 + 17 = 108<br>20/108 = 18.5% 9/2<br>.. and and and on.
To get a proper price you must have a good rating, I would suggest that your ratings need more work because the spread of the price is only as good as the rating.June 7, 2007 at 20:40 #63566
I have tried something similar, but from the other direction, starting on 1 and going upwards, though it wasn’t much better. Your idea sounds sensible though so I’ll give it a try. I could adjust the number to one outside of which no horse can have a chance, maybe 28, or perhaps a different number for different types of races, or sizes of fields.
I would like to produce my own ratings, so I’m also having another look at Artemis’s excellent thread on the ratings he produced, to get some pointers. I’d like to produce my own but I also want to cross that threshold in my mind where I can be fairly happy they are as good if not better than those produced by Raceform/Timeform (in the area that I want to concentrate on)June 8, 2007 at 07:44 #63567
It’s a very fair point dave raises about using current prices to prepare a tissue. It can certainly compound any errors which are contained in the prices you use.
Obviously, the closer to the off time, the more dependable these prices will be. You can make the points awarded somewhat less than I do if that is your preference.
I would doubt that many tissue compilers who are not doing it for a living prepare their lists without looking at the exchanges and other tissues. The RP tissue is available at 7.00 pm the evening before the race, so that is a good starting point.
Compiling your own ratings from results would be a mammoth task, even if you specialised in one area such as 2yo. It would also be fascinating and absorbing, but I would not be eager to recommend it.
RPR and Topspeed are quite a good foundation but have the major disadvantage that hundreds of people use them. As a result, selections based on these ratings are often overbet.
Really, there is no sustitute for reading the form book and watching as much racing as you possibly can. The professionals are disdainful of most ‘systems’, although they might use a systematic approach. The problem for most of us is that we don’t have the time to do the job properly, so we look for short-cuts and instant fixes.<br>I’m afraid these do not work – there is no holy grail or magic formula waiting for a Newton of the punting world to discover it. Horse races are chaotic events and difficult to analyse and predict, but it’s fun to try.June 8, 2007 at 09:01 #63568
That’s a good post Artemis and I agree with all of that, you should take any system betting with a large pinch of salt … systems generally don’t work over time, but you can be lucky for a short while, even a year or so and make a profit. Systems tend to be overwhelmed by waves of luck, good and bad. If you don’t know why you are winning then it will invariably be luck. If your overall profit is higher than around 10%, that will probably be luck also.
I would also advise against the rating route .. unless you have loads of time and/or good programming skills you could be at it forever.
A good starting point is to copy someone else’s ratings and use their historical data and make your own adjustment for future races, to get a feel for it.
- You must be logged in to reply to this topic.