General Betting
Welcome to Live View – Take the tour to learn more
Start Tour
There is currently 1 person viewing this thread.
WIN PROBABILITY WITH HORSE RACING USING RATINGS
I would like to use KELLY STAKING method on one of the betting strategies applicable to ratings, but I can see different views on arriving at win probability. I have set out some thoughts below and if anyone can/will express views on this subject, then it will be much appreciated.
=====================================================================
Out of 626 races , 444 winners were in top 50% + 1 of field size and those winners were as follows :-
TOP RATED - 110 ( 17.57% of all races)
2ND TOP RATED - 104 ( 16.61% of all races)
3RD TOP RATED - 58 ( 9.26% of all races)
OTHER TOP RATED - 172 (27.47% of all races)
Using above %s as win probability for horses ranked TR/2TR etc does not take account that if a horse is ranked in top 50%+1 of field size then , based on past results, there is a 70.92 % chance the winner is within the top 50%+1.
Average field size for the 626 races is 11 approximately…..
Using top 6 (50%+1 of average field size figure) from where say 71% of winners can be achieved using ratings, then if a horse was 6th rated this would be a 20% ( 5 to 1) if 100% of winners were in top 50%+1 of field size.
Is it then reasonable to say 14.20 % ( 71% x 20 ) is the average win probability for top 6 rated horses?
No appears to be the answer to above question…..as the win strike-rate for top 2 rated horses is nearer 18 for top rated and 17% for 2nd top rated and 9% for 3rd rated is based on past results for ratings. This leaves 27% for 3 other horses in top rated……,but 14% is perhaps the win probability to use for horses rated 4, 5 and 6.……..
3rd favourite win stats are 14% ,year in and year out, per AM website, so perhaps when backing them this win probability % is the one to use, although using KELLY you are supposed to be as accurate as possible re % used based on each race…presumably, as opposed to a blanket % for all races.
===================================================================
Initially I used a very crude method to weigh up whether there were horses at top end of market (favs etc) that had “value” in odds using RANK position they were in in ratings.
E.g. CHEPSTOW14:30
1,Diamond Vine 12
2,Sole Danser 3RD 28-1
3,Ghostwing 8 3RD F
4,Lady Bayside 1ST 11-2 2ND F
5,Bateleur 25
6,Belle Bayardo 14
7,Bermondsey Bob 3.33 F
8,Spirit Of Gondree 14
=============================================
9,Huzzah 14
10,Festival Dance 28
11,Nubar Boy 2ND 20-1
12,Noble Four 8.5
13,Going French 14
14,Colourbearer 18
15,Brandywell Boy 40
The 3rd fav is 3rd ranked, therefore this is a 2 to 1 win possibility……then the available odds of 8-1 are clearly VALUE and should be backed.
The 2nd fav is 4th ranked, therefore considered a 3 to 1 win shot……..so with available SP odds at 11-2 this should also be backed. It won.
The fav is ranked 7th and is considered a 7 to 1 shot……therefore as available odds at SP were 3.33 it is clearly a VALUE LAY!
In effect the RANK position determines what the win probability is deemed to be.
The reality though is ……the stats shown at start of this post reflect what past results show for the system’s ratings.
It would be “dangerous” for bank if the RANK position win deemed % method was used for KELLY though……due to it not being as accurate as the historical results from previously rated races !
=======================================================================
Currently I am using the win probability to check past rated races as set out below :-
Where horse is -
RANKED
1.18%.
2. 17%
3. 9%
Others ranked above the line (drawn after 50%+1 on field size) but not in top 3 rated…..9%.
As average field size for the 626 races checked is 11 approx., the above application of win % for top 6 rated gives …..18+17+9+9+9+9 = 71 %. The 71% is in line with the over all stats for wins in those ranked in top 50%+1 of field size ( 444 winners out of 626 races for one period checked).
Using the above KELLY win probabilities to check out the difference in PROFIT and LOSS between -
Backing using LEVEL STAKES bets of 1 point for 1 week in JULY 2008... PROFIT 39.62 points.
Backing using KELLY STAKING, using 100 point bank, …………………PROFIT 52.44 points.
In the week checked it is “interesting” that using KELLY STAKING there were :-
57 NO BETS compared with LEVEL STAKING method.
29 BETS using KELLY.
6 winners selected.
10 horses won that were NOT selected ( in 57 NO BETS).
and more PROFIT would have been made compared to LEVEL STAKES crude way of betting.
===========================================================================
Any views on the above would be much appreciated.
GL with bets today.
I would like to use KELLY STAKING method on one of the betting strategies applicable to ratings, but I can see different views on arriving at win probability. I have set out some thoughts below and if anyone can/will express views on this subject, then it will be much appreciated.
=====================================================================
Out of 626 races , 444 winners were in top 50% + 1 of field size and those winners were as follows :-
TOP RATED - 110 ( 17.57% of all races)
2ND TOP RATED - 104 ( 16.61% of all races)
3RD TOP RATED - 58 ( 9.26% of all races)
OTHER TOP RATED - 172 (27.47% of all races)
Using above %s as win probability for horses ranked TR/2TR etc does not take account that if a horse is ranked in top 50%+1 of field size then , based on past results, there is a 70.92 % chance the winner is within the top 50%+1.
Average field size for the 626 races is 11 approximately…..
Using top 6 (50%+1 of average field size figure) from where say 71% of winners can be achieved using ratings, then if a horse was 6th rated this would be a 20% ( 5 to 1) if 100% of winners were in top 50%+1 of field size.
Is it then reasonable to say 14.20 % ( 71% x 20 ) is the average win probability for top 6 rated horses?
No appears to be the answer to above question…..as the win strike-rate for top 2 rated horses is nearer 18 for top rated and 17% for 2nd top rated and 9% for 3rd rated is based on past results for ratings. This leaves 27% for 3 other horses in top rated……,but 14% is perhaps the win probability to use for horses rated 4, 5 and 6.……..
3rd favourite win stats are 14% ,year in and year out, per AM website, so perhaps when backing them this win probability % is the one to use, although using KELLY you are supposed to be as accurate as possible re % used based on each race…presumably, as opposed to a blanket % for all races.
===================================================================
Initially I used a very crude method to weigh up whether there were horses at top end of market (favs etc) that had “value” in odds using RANK position they were in in ratings.
E.g. CHEPSTOW14:30
1,Diamond Vine 12
2,Sole Danser 3RD 28-1
3,Ghostwing 8 3RD F
4,Lady Bayside 1ST 11-2 2ND F
5,Bateleur 25
6,Belle Bayardo 14
7,Bermondsey Bob 3.33 F
8,Spirit Of Gondree 14
=============================================
9,Huzzah 14
10,Festival Dance 28
11,Nubar Boy 2ND 20-1
12,Noble Four 8.5
13,Going French 14
14,Colourbearer 18
15,Brandywell Boy 40
The 3rd fav is 3rd ranked, therefore this is a 2 to 1 win possibility……then the available odds of 8-1 are clearly VALUE and should be backed.
The 2nd fav is 4th ranked, therefore considered a 3 to 1 win shot……..so with available SP odds at 11-2 this should also be backed. It won.
The fav is ranked 7th and is considered a 7 to 1 shot……therefore as available odds at SP were 3.33 it is clearly a VALUE LAY!
In effect the RANK position determines what the win probability is deemed to be.
The reality though is ……the stats shown at start of this post reflect what past results show for the system’s ratings.
It would be “dangerous” for bank if the RANK position win deemed % method was used for KELLY though……due to it not being as accurate as the historical results from previously rated races !
=======================================================================
Currently I am using the win probability to check past rated races as set out below :-
Where horse is -
RANKED
1.18%.
2. 17%
3. 9%
Others ranked above the line (drawn after 50%+1 on field size) but not in top 3 rated…..9%.
As average field size for the 626 races checked is 11 approx., the above application of win % for top 6 rated gives …..18+17+9+9+9+9 = 71 %. The 71% is in line with the over all stats for wins in those ranked in top 50%+1 of field size ( 444 winners out of 626 races for one period checked).
Using the above KELLY win probabilities to check out the difference in PROFIT and LOSS between -
Backing using LEVEL STAKES bets of 1 point for 1 week in JULY 2008... PROFIT 39.62 points.
Backing using KELLY STAKING, using 100 point bank, …………………PROFIT 52.44 points.
In the week checked it is “interesting” that using KELLY STAKING there were :-
57 NO BETS compared with LEVEL STAKING method.
29 BETS using KELLY.
6 winners selected.
10 horses won that were NOT selected ( in 57 NO BETS).
and more PROFIT would have been made compared to LEVEL STAKES crude way of betting.
===========================================================================
Any views on the above would be much appreciated.
GL with bets today.
Show More
I think your sample size ( 626 races ) is far to small.
Horse racing has so many seemingly small variations, but these nuances
significantly affect the data.
I would suggest using a data set of at least 15,000+ races as a valid starting point.
Otherwise any systematic edges you find will more likely be noise.
best of luck with it
Horse racing has so many seemingly small variations, but these nuances
significantly affect the data.
I would suggest using a data set of at least 15,000+ races as a valid starting point.
Otherwise any systematic edges you find will more likely be noise.
best of luck with it
Thanks Escapee for post.
Have carried out about 2,000 ratings on races......but only this week decided to consider KELLY STAKING.
Pretty sure the results for the 2,000 races are roughly in line with the 626 races re where winners are ranked.....
15,000 appears to be "overkill"....in that i might be dead before completing checks
Have carried out about 2,000 ratings on races......but only this week decided to consider KELLY STAKING.
Pretty sure the results for the 2,000 races are roughly in line with the 626 races re where winners are ranked.....
15,000 appears to be "overkill"....in that i might be dead before completing checks
only basing it on my own experiences.
I've lost count of the many times I thought I was onto something over
a period of 1,000->3,000 races only to find its just a blip when put
into the context of 15,000 races.
Now I have a few things working over 15k races, but I'm ultra aware that these could just be blips when
run over 50,000 races.
Most people just don't comprehend quite how much variation there can be in probabilistic data sets.
Size is everything
I've lost count of the many times I thought I was onto something over
a period of 1,000->3,000 races only to find its just a blip when put
into the context of 15,000 races.
Now I have a few things working over 15k races, but I'm ultra aware that these could just be blips when
run over 50,000 races.
Most people just don't comprehend quite how much variation there can be in probabilistic data sets.
Size is everything
If your "edge/s" is/are working well over about 2,000 races.......being all types of races whether AW/NH or TURF, then if they are genuine "edges" they should stand up with future races, provided no major changes come in re handicapping etc.
If you only have 1 "edge"..........then I would agree it needs more races checked out.......but if you have NUMEROUS "edges", then it is more likely your method of rating races will stand up to future races, as it is the complexity of the system of rating races that negates the need for carrying out as many checks etc.
I know the system will find VALUE backs, as that is what it was devised to do, and the results of numerous checks on past rated races support my belief.
Using KELLY for backing ONLY on certain horse selections is what I am weighing up at present......and it appears to work well SO FAR.....though I agree 1 week's checks are not enough. The improvement in PROFIT using KELLY makes it worth further consideration/checks.
I agree that most people do not comprehend the complexity necessary to rating a race......but the better your system is at rating a race, then the less variation there should be on results. Therefor is it really necessary to check out 15,000 races when it works generally well over about 2,000 races?
If you only have 1 "edge"..........then I would agree it needs more races checked out.......but if you have NUMEROUS "edges", then it is more likely your method of rating races will stand up to future races, as it is the complexity of the system of rating races that negates the need for carrying out as many checks etc.
I know the system will find VALUE backs, as that is what it was devised to do, and the results of numerous checks on past rated races support my belief.
Using KELLY for backing ONLY on certain horse selections is what I am weighing up at present......and it appears to work well SO FAR.....though I agree 1 week's checks are not enough. The improvement in PROFIT using KELLY makes it worth further consideration/checks.
I agree that most people do not comprehend the complexity necessary to rating a race......but the better your system is at rating a race, then the less variation there should be on results. Therefor is it really necessary to check out 15,000 races when it works generally well over about 2,000 races?
I let handicapper do the ratings, they vastly more experienced than me.
My "system" does not use such "ratings" R46........never considered there was any need to do so....but perhaps , at some stage in future, I might consider weighting such ratings!
Just after completing a check using KELLY on races rated on 23.07.08 and following "interesting" stats arise ....
14 NO BETS compared with LEVEL STAKING method.
3 BETS using KELLY.
1 winners selected.
4 horses won that were NOT selected ( in 14 NO BETS).
and more PROFIT would have been made compared to LEVEL STAKES crude way of betting ( 50.71 points compared to 26.48 points!)
14 NO BETS compared with LEVEL STAKING method.
3 BETS using KELLY.
1 winners selected.
4 horses won that were NOT selected ( in 14 NO BETS).
and more PROFIT would have been made compared to LEVEL STAKES crude way of betting ( 50.71 points compared to 26.48 points!)
24.07.08 was a LOSS of 10.61 points using LEVEL STAKES method........the KELLY METHOD result will be "interesting" on this day!
The check using KELLY on races rated on 24.07.08 had following stats ....
16 NO BETS compared with LEVEL STAKING method.
1 BETS using KELLY.
0 winners selected.
1 horse won that was NOT selected ( in 16 NO BETS).
and the LOSS was lower than LOSS compared to LEVEL STAKES crude way of betting ( 4.52 points compared to 10.61 points!)
16 NO BETS compared with LEVEL STAKING method.
1 BETS using KELLY.
0 winners selected.
1 horse won that was NOT selected ( in 16 NO BETS).
and the LOSS was lower than LOSS compared to LEVEL STAKES crude way of betting ( 4.52 points compared to 10.61 points!)
No more checks tonight...time for teaaaaaaaaaaa
Cheers the nooooooooooooo
Cheers the nooooooooooooo
Kelly staking is perfect if you know what your % advantage is ,which in racing can change all the time up and down. If your
% advantage isn't accurate then it won't maximise it and if it is well out, will inevitably lead to the wiping out of the betting bank.
The best way to tackle it is probably to work out % advantage over a long period of time (lets say its 8%) then use Kelly criterion
on a 5% gain. Although this won't maximise profits if an 8% gain is gained in the future ,it lessens the chance of being wiped
out should the advantage % drop below 8%.
Personally I think kelly criterion is best used on an individual race basis. If in one race a person thinks a 5/1 shot should
be 2/1 this should lead to a lot bigger bet than a 4/1 shot that you think is a 7/2 shot. Some people will bet level stakes on both
but if the assessment of the true odds is accurate then it cannot be right and profits aren't maximised.
If you wanted to invest your money in a bank and 2 were offering interest at 2% pa and one at 10% pa and you were forced to have
money in each of them then you'd have very little in the 2% banks and a big chunk in the 10% bank maximising your rate of return.
% advantage isn't accurate then it won't maximise it and if it is well out, will inevitably lead to the wiping out of the betting bank.
The best way to tackle it is probably to work out % advantage over a long period of time (lets say its 8%) then use Kelly criterion
on a 5% gain. Although this won't maximise profits if an 8% gain is gained in the future ,it lessens the chance of being wiped
out should the advantage % drop below 8%.
Personally I think kelly criterion is best used on an individual race basis. If in one race a person thinks a 5/1 shot should
be 2/1 this should lead to a lot bigger bet than a 4/1 shot that you think is a 7/2 shot. Some people will bet level stakes on both
but if the assessment of the true odds is accurate then it cannot be right and profits aren't maximised.
If you wanted to invest your money in a bank and 2 were offering interest at 2% pa and one at 10% pa and you were forced to have
money in each of them then you'd have very little in the 2% banks and a big chunk in the 10% bank maximising your rate of return.
How did it go tonight DFC?
Had a nice lay on Reading when they were 3-0 up and 4-0 up..4-0 lay was at 1.03 (for extra £20). That's when you wish you punted some more at 1.03!..lol They were never going to come out 2nd half lying down. Can't complain..netted me £50 profit
Had a nice lay on Reading when they were 3-0 up and 4-0 up..4-0 lay was at 1.03 (for extra £20). That's when you wish you punted some more at 1.03!..lol They were never going to come out 2nd half lying down. Can't complain..netted me £50 profit
tinkler - I don't believe that I know the exact % advantage I have, other than the hard past stats that indicate 18% top rated ones win and 17% of 2nd top rated win...and 9% of those rated 3rd rated win........and 71% of those rated in top 50%+1 of field size win in top rated.
It does not change "up and down " re the general stats.....over approx 2,000 races.
Don't believe ANYONE can be "perfect" on knowing the win probability %...
However, based on checking results of applying KELLY STAKING to 9 days in JULY 2008 rated races the difference between KELLY STAKING and LEVEL STAKING is 43.15 points extra PROFIT!
Not sure why you select BANKS as investment vehicles....after their total lack of any sensible investment policies over the years up to the collapse of their illusions....BTW.....you probably would LOSE your money choosing the idiotic 10% return bank.....when market generally is 2% on returns....unless you had an "edge" re your investment which MOST people don't have when investing through banks.
JBG....watched READING game in pub...best place for such a dodgy game .....Glad you won though
AL trying to "fix" problem.......so hopefully he will.
Watching poker a wee bit now.....time to chill out!
G M the nooooooooooooooooooooooooo
It does not change "up and down " re the general stats.....over approx 2,000 races.
Don't believe ANYONE can be "perfect" on knowing the win probability %...
However, based on checking results of applying KELLY STAKING to 9 days in JULY 2008 rated races the difference between KELLY STAKING and LEVEL STAKING is 43.15 points extra PROFIT!
Not sure why you select BANKS as investment vehicles....after their total lack of any sensible investment policies over the years up to the collapse of their illusions....BTW.....you probably would LOSE your money choosing the idiotic 10% return bank.....when market generally is 2% on returns....unless you had an "edge" re your investment which MOST people don't have when investing through banks.
JBG....watched READING game in pub...best place for such a dodgy game .....Glad you won though
AL trying to "fix" problem.......so hopefully he will.
Watching poker a wee bit now.....time to chill out!
G M the nooooooooooooooooooooooooo
DF
Using banks isn't the greatest example but if you put up to £50,000 in one you've got the government guaranteeing you'll get it back.
If a person does the same with a bookie there is no such guarantee even the big boys could go bust due to acute mis-management.
The supposed safe guard of ring fenced accounts (such as Bf) was blown apart by the Full tilt debacle.
If a person makes say 7% return on there average bet then the natural inclination is to increase bet sizing in line with an
increased betting bank, if the 7% advantage is maintained the betting bank will get bigger and bigger . Once it reaches a certain
point new problems come about, firstly if your betting mainly with the bookies they'll eventually close your accounts not to mention
if you were intially betting £50 a point and now betting £1,000 a point it's not that easy to get the money on at the required price
unless the price is available on BF 5 mins before the off. There is also in most people a choke point where there judgement is effected due to the size of the amount being bet. If detachment can be kept from the amounts being bet and the % advantage maintained plus liquidity to get it on then there is no reason why a betting bank shouldn't mushroom.
Good luck with your project.
Using banks isn't the greatest example but if you put up to £50,000 in one you've got the government guaranteeing you'll get it back.
If a person does the same with a bookie there is no such guarantee even the big boys could go bust due to acute mis-management.
The supposed safe guard of ring fenced accounts (such as Bf) was blown apart by the Full tilt debacle.
If a person makes say 7% return on there average bet then the natural inclination is to increase bet sizing in line with an
increased betting bank, if the 7% advantage is maintained the betting bank will get bigger and bigger . Once it reaches a certain
point new problems come about, firstly if your betting mainly with the bookies they'll eventually close your accounts not to mention
if you were intially betting £50 a point and now betting £1,000 a point it's not that easy to get the money on at the required price
unless the price is available on BF 5 mins before the off. There is also in most people a choke point where there judgement is effected due to the size of the amount being bet. If detachment can be kept from the amounts being bet and the % advantage maintained plus liquidity to get it on then there is no reason why a betting bank shouldn't mushroom.
Good luck with your project.
This is a good document concerning converting ratings into probabilities
http://www.football-data.co.uk/ratings.pdf
http://www.football-data.co.uk/ratings.pdf
Thanks for post DB.
HORSES rating is a different animal to FOOTBALL ones. I'm pretty sure I glimpsed at that PDF before, fine for FOOTBALL, but not much help re - HORSE RACE RATINGS tough
Having reviewed the top rated horses odds and IR odds, and considered the use of KELLY where higher VALUE level exists/appears to exist, I now have strategies that have produced, in theory , on past results 19 points per day on average profits over relatively short period checked.
Not gone through the almost 2,000 rated races to check whether such strategies would have worked, only glimpsed at relevant horse results and , at a glance can see the strategies would generally work well. 19 points a day on average may be difficult in practice to achieve, though 12 points looks the likely lowest average daily profit when sufficient number of races being run in Britain.
Currently waiting on someone fixing a "problem" re RACE CARDS before I can put strategies into practice. Pretty technical "problem" needs fixing, so may take a few weeks.
Any other views on use of RATINGS for HORSE RACING and deriving probabilities for win etc would be appreciated.
GL with bets.
HORSES rating is a different animal to FOOTBALL ones. I'm pretty sure I glimpsed at that PDF before, fine for FOOTBALL, but not much help re - HORSE RACE RATINGS tough
Having reviewed the top rated horses odds and IR odds, and considered the use of KELLY where higher VALUE level exists/appears to exist, I now have strategies that have produced, in theory , on past results 19 points per day on average profits over relatively short period checked.
Not gone through the almost 2,000 rated races to check whether such strategies would have worked, only glimpsed at relevant horse results and , at a glance can see the strategies would generally work well. 19 points a day on average may be difficult in practice to achieve, though 12 points looks the likely lowest average daily profit when sufficient number of races being run in Britain.
Currently waiting on someone fixing a "problem" re RACE CARDS before I can put strategies into practice. Pretty technical "problem" needs fixing, so may take a few weeks
.Any other views on use of RATINGS for HORSE RACING and deriving probabilities for win etc would be appreciated.
GL with bets.
Post Your Reply
Please login to post a reply.
Wonder
Instance ID: 13539