Bookies and team ratings

[ARAZI91]

I don't make anecdotal stories.
They 're true.
With the jockey's mom case I spent a summer in Karystos - better than Mykonos with Mc Guire and his hippie company.

Now those guys you mention I do believe of course they are real.
But with the kind of money you talk about can you assume there is no hanky-panky stuff involved ?
Hanky panky stuff removes us even further from the premises of our debate.
We should n't count that.

Anyway I never knocked other peoples methods when they claim successes.
Some other friends I have are now into scanners so there may be something in that too.
What I 'm doing is try to locate abnormal patterns in the way some prices are structured.

But you implied they are "Groups of Losers" - most of the people who play at that level you would not even know their names, as it's a secretive world - Zeljko's partner David Walsh( a genuine math prodigy) has made that much money from "locating abnormal patterns in the way ALL prices are structured" has even built his own conceptual museum , MONA back in his homeland in Hobart
Mona – Museum of Old and New Art | Hobart, Tasmania all funded from maximised value betting.
Tony Bloom and Matthew Benham made so much money from Asian handicap betting in football , they bought their own football teams, Brighton and Brentford and use the same "Moneyball" principles and a "Value" approach to expand those teams using data and analytics - Now it's common for every football team of note to have a data/analytics research department.
Alan Woods had companies and shares in tech,media,manufacturing,and played the financials the same way he approached horse racing by using a "Value" orientated approach - he was also one of the biggest players on Betfair before he died. Zeljko is one the top 4 biggest players on Betfair and has 4 account managers but because of the advantage he holds he is on a bespoke commission rate as well as top rate Premium Charge. Don't know about scanners but i do know of a few Irish mobs who were using "drones" to find an edge within "in-running" betting on Betfair.
You talk of "hanky panky" but that "hanky panky" is only true Value betting by having better and more accurate probabilities than the market makers.The work that goes into creating those probabilities involves advanced math, statistical analysis and sheer hard graft and when they spot a "rick" or an error within the market makers product they maximise it till it's bled dry. They are advantage gamblers through and through and no matter the game they will find a way to beat it legitimately. I have corresponded with ex-Benter and ex-Woods employees as well as watch a syndicate work up close - they play a straight game as there is no honour amongst gamblers when the game is rigged or cheating is involved - they dismiss inside information as most of it is bullshit anyway as within one race their could be 3 to 4 horses at least trying or ready to run to their maximum ability or improve.

 

[ARAZI91]

And i'm in no way doubting your anecdotes about your home land are true or not ( anecdotes are not defined by truth or falsehood - look up the dictionary meaning - "a short amusing or interesting story about a real incident or person". i am doubting the sample size of "mean nothing" games once a team has qualified for the next round - in no world does that stand up to statistical scrutiny - your basing a theory on what ?11 games - i would bet that there would be plenty of teams who once qualified would still go on to win their remaining games within football tournaments even sending out fringe players. Have you ever seen the Spanish or Italian press when one of their teams draw or lose a game they are expected to win - they are hounded, embarrassed even. And if you think Market makers such as bookmakers are "following" your commercial open source ratings then think again - the whole idea between punters and market makers is to use bespoke non-public information, the lesser known the better. In the long ago bookmakers plastered the Racing Post on every wall of every betting shop ? - why ? Because most of the information contained within the past performances were already factored into their odds.

 

[cavelino]

Tremendous replies there, Robert.

 

[cosmicsports]

But you implied they are "Groups of Losers" - most of the people who play at that level you would not even know their names, as it's a secretive world - Zeljko's partner David Walsh( a genuine math prodigy) has made that much money from "locating abnormal patterns in the way ALL prices are structured" has even built his own conceptual museum , MONA back in his homeland in Hobart
Mona – Museum of Old and New Art | Hobart, Tasmania all funded from maximised value betting.
Tony Bloom and Matthew Benham made so much money from Asian handicap betting in football , they bought their own football teams, Brighton and Brentford and use the same "Moneyball" principles and a "Value" approach to expand those teams using data and analytics - Now it's common for every football team of note to have a data/analytics research department.
Alan Woods had companies and shares in tech,media,manufacturing,and played the financials the same way he approached horse racing by using a "Value" orientated approach - he was also one of the biggest players on Betfair before he died. Zeljko is one the top 4 biggest players on Betfair and has 4 account managers but because of the advantage he holds he is on a bespoke commission rate as well as top rate Premium Charge. Don't know about scanners but i do know of a few Irish mobs who were using "drones" to find an edge within "in-running" betting on Betfair.
You talk of "hanky panky" but that "hanky panky" is only true Value betting by having better and more accurate probabilities than the market makers.The work that goes into creating those probabilities involves advanced math, statistical analysis and sheer hard graft and when they spot a "rick" or an error within the market makers product they maximise it till it's bled dry. They are advantage gamblers through and through and no matter the game they will find a way to beat it legitimately. I have corresponded with ex-Benter and ex-Woods employees as well as watch a syndicate work up close - they play a straight game as there is no honour amongst gamblers when the game is rigged or cheating is involved - they dismiss inside information as most of it is bullshit anyway as within one race their could be 3 to 4 horses at least trying or ready to run to their maximum ability or improve.

By hanky-pankies I mean fixes.
In the old race course sometimes such intel reached me (now I don't go - I don't want to travel the distance and also the environment has changed).
In football never ever did something reach me. By inference maybe but nothing ever reached me by word of mouth.
So what I say is let's discount such form the conversation as if don't exist.
We talk about prediction theory, market analysis and the experts in those fields.

 

[cosmicsports]

And i'm in no way doubting your anecdotes about your home land are true or not ( anecdotes are not defined by truth or falsehood - look up the dictionary meaning - "a short amusing or interesting story about a real incident or person". i am doubting the sample size of "mean nothing" games once a team has qualified for the next round - in no world does that stand up to statistical scrutiny - your basing a theory on what ?11 games - i would bet that there would be plenty of teams who once qualified would still go on to win their remaining games within football tournaments even sending out fringe players. Have you ever seen the Spanish or Italian press when one of their teams draw or lose a game they are expected to win - they are hounded, embarrassed even. And if you think Market makers such as bookmakers are "following" your commercial open source ratings then think again - the whole idea between punters and market makers is to use bespoke non-public information, the lesser known the better. In the long ago bookmakers plastered the Racing Post on every wall of every betting shop ? - why ? Because most of the information contained within the past performances were already factored into their odds.

Again you misunderstand.
I did n't say public domain ratings are super.
Surely the bookies have something better, even though I don't see gross differences.

Then those 11 matches was all I could do in five minutes.
It takes more than that.
And surely I 've been to such matches myself in the terraces and sometimes the favourite plays really well.

Why the "mean nothing" games have been used as a test group by me ?
Because they are representative of a bigger group of matches in which some important external (to the ratings) factor influences things.
Also because I don't have to remember things, situations and events. I just take a note of the scoreline.

 

[ARAZI91]

With Selby v. Wilson you say they started 1.70 - 2.43 and became 1.48 - 3.03.
In terms of probabilities that makes 59% - 43% and then 67% - 33%.
Do you mean that every food poisoning means I lose 25% of my initial probability, if I go and play snooker under the effect of food poisoning ?
Medical science can't do that, so what do you actually do to go from (59-43) to (67-33) ?
I mean to do it with statistical factors. How else ?
Also if you say "it's from how the market evolved" it's no good as I 'm supposed to be ahead of the market.

With horses it goes like this sometimes:
Suppose from speed considerations the win probability is 30%. But the jockey is an auxiliary, one used to carry the horses to and from the stables.
An attenuation factor applies and the 30% becomes 10%.

With this logic of course one can err.
Recently there was one such match. An derby match and many people, everyone rather, thought the home team does n't care and will lose.
We were mugged. No laziness or anything like that was exhibited and it was a home win in the end.
But the total score counts.

No not every "food poisioning" scenario means you lose 25% of your initial probability. I used "simplicity" to show how the technique works as it is Bayesian in nature - your also accounting for Selby's statement that "he has no chance of winning this match" , his practise session where he was hitting the balls all over the shop - obviously his usual standard had dropped - much depends on the actual market prices which on Betfair are as close to "true odds" we will ever have.As it is Bayesian you use a prior first and work from there - you could run a linear regression and use iteration till Kyren Wilson's "new" updated price has an R2 of 0.9999 and an standard error of 0.0001 - this would also give you the probabilistic factor of what to use in the Degenerate Distribution. Based on the new information i implied that Wilson must now have a greater than 0.50 chance than originally implied and added on 10% which gives a Degenerative value of 0.55 (i only done this to show how the concept works) - that 0.55 mass of probability is multiplied by the Degenerative Value of 0.55 and and then added to his original probability which i compiled from the ratings.
As for "compound probability" - that is not the correct way to "combine" probabilities as it should be an "additive" process
Imagine i wanted to combine my final "model of models" probabilities to the current markets probabilities and i had a live feed of the markets prices that refresh every minute. The true nature of this process is logistical and involves the use of exponents and logits - to combine one set of probabilities with another looks like this - here imagine my final rating is my "Model of Models" final Probs and i want to combine that with the current Market Odds (Betfair) here im using a weighting of 0.40 model and 0.60 market

So it's a horse race and the first part has a confidence interval built in which is horse specific or can be race specific - these are my final probs and all in i'm using a Margin Of Error of 14.53% - my models odds sum to a 85.47% book (it's actually 17% but because of proportion to odds it works out to 14.53 , ie it counts for more at the top end of the probability spectrum and lesser the further you go down the probability spectrum)



I then want to "combine" them with the current market Odds which in this case is Betfair's -this the most optimal way of doing it using the exponential distribution and the logistical distribution -so using a Model /Market weight of 0.40/0.60 respectively they would look like this



As you can see both probs(as a % chance) and odds combined sum to 100% or 1.00
Looking at +/-EXPEV, QRTKELLY and FULLKELLY the top two in the betting are still "overlays" , and the most optimal bets , i may also bet Queen Olly as i use a variable stakes approach with "edge obtained" layered over that variable stakes.But as you go down the rank of +/-EXPECTED EV the more the signal of information decreases. If you use the metric BFSP(p)/FS(p) where (p)=probability and 1.00= Exact Field Size - eg 0.10/0.10 =1.00
then only 15-17% of all winners have a value from that metric that is <1.00 over many years of horse racing data
Also within any of my horse racing models would i use win% as it reduces a multi-competitor sport to a coin toss where that coin toss is binary and the two outcomes are 1 win and 0 Loss - this is where "optimisation" comes in and you can use a fin.pos metric such as normalised finising position that is more granular and uses all data points and not just two binary data points




In this example i combined my final base probability with the markets probability but it could have been combining a jockey model's final probability with a trainers model final probability - the process using the exponential distribution and the logistical distribution remains the same

 

[ARAZI91]

By hanky-pankies I mean fixes.
In the old race course sometimes such intel reached me (now I don't go - I don't want to travel the distance and also the environment has changed).
In football never ever did something reach me. By inference maybe but nothing ever reached me by word of mouth.
So what I say is let's discount such form the conversation as if don't exist.
We talk about prediction theory, market analysis and the experts in those fields.

That is why the big syndicates combine their final base probabilities with the markets probabilities as there is a signal there in market info that cannot be accounted for in models .

 

[ARAZI91]

Again you misunderstand.
I did n't say public domain ratings are super.
Surely the bookies have something better, even though I don't see gross differences.

Then those 11 matches was all I could do in five minutes.
It takes more than that.
And surely I 've been to such matches myself in the terraces and sometimes the favourite plays really well.

Why the "mean nothing" games have been used as a test group by me ?
Because they are representative of a bigger group of matches in which some important external (to the ratings) factor influences things.
Also because I don't have to remember things, situations and events. I just take a note of the scoreline.

Why use scoreline - when you can use xG

 

[ARAZI91]

By hanky-pankies I mean fixes.
In the old race course sometimes such intel reached me (now I don't go - I don't want to travel the distance and also the environment has changed).
In football never ever did something reach me. By inference maybe but nothing ever reached me by word of mouth.
So what I say is let's discount such form the conversation as if don't exist.
We talk about prediction theory, market analysis and the experts in those fields.

Sure there is gaming that goes on - especially in jurisdictions that use official ratings to frame handicaps but there are metrics that can account for this , whilst combining your base final outputs as probs with the markets probs usually show up any "gaming" as most trainers who do this actually bet ,as do owners and most people connected with the yard.

 

[ARAZI91]

Why use scoreline - when you can use xG

xG is more predictive than "scoreline" as it is an aggregated metric of who should have won - there are at least about 12 inputs that go into xG but only one in the scoreline. Studies have shown that xG is more predictive of future scoreline than scorelines itself. It;s the equivalent of PRB2 in horse racing v Win % - studies have shown that PRB2 has a better chance of predicting future Win% than Win % itself

 

[cosmicsports]

xG is more predictive than "scoreline" as it is an aggregated metric of who should have won - there are at least about 12 inputs that go into xG but only one in the scoreline. Studies have shown that xG is more predictive of scoreline than scorelines itself. It;s the equivalent of PRB2 in horse racing v Win % - studies have shown that PRB2 has a better chance of predicting future Win% than Win % itself

Hrmph.
By scoreline I only mean the 2nd leg was identified as non-affair.
E.g. City - Porto 5-0 in Porto then in Manchester they were playing to pass the time.

 

[cosmicsports]

No not every "food poisioning" scenario means you lose 25% of your initial probability. I used "simplicity" to show how the technique works as it is Bayesian in nature - your also accounting for Selby's statement that "he has no chance of winning this match" , his practise session where he was hitting the balls all over the shop - obviously his usual standard had dropped - much depends on the actual market prices which on Betfair are as close to "true odds" we will ever have.As it is Bayesian you use a prior first and work from there - you could run a linear regression and use iteration till Kyren Wilson's "new" updated price has an R2 of 0.9999 and an standard error of 0.0001 - this would also give you the probabilistic factor of what to use in the Degenerate Distribution. Based on the new information i implied that Wilson must now have a greater than 0.50 chance than originally implied and added on 10% which gives a Degenerative value of 0.55 (i only done this to show how the concept works) - that 0.55 mass of probability is multiplied by the Degenerative Value of 0.55 and and then added to his original probability which i compiled from the ratings.
As for "compound probability" - that is not the correct way to "combine" probabilities as it should be an "additive" process
Imagine i wanted to combine my final "model of models" probabilities to the current markets probabilities and i had a live feed of the markets prices that refresh every minute. The true nature of this process is logistical and involves the use of exponents and logits - to combine one set of probabilities with another looks like this - here imagine my final rating is my "Model of Models" final Probs and i want to combine that with the current Market Odds (Betfair) here im using a weighting of 0.40 modeland 0.60 market

So it's a horse race and the first part has a confidence interval built in which is horse specific or can be race specific - these are my final probs and all in i'm using a Margin Of Error of 14.53% - my models odds sum to a 85.47% book

View attachment 135078

I then want to "combine" them with the current market Odds which in this case is Betfair's -this the most optimal way of doing it using the exponential distribution and the logistical distribution -so using a Model /Market weight of 0.40/0.60 respectively they would look like this

View attachment 135079

As you can see both probs(as a % chance) and odds combined sum to 100% or 1.00
Looking at +/-EXPEV, QRTKELLY and FULLKELLY the top two in the betting are still "overlays" , and the most optimal bets , i may also bet Queen Olly as i use a variable stakes approach with "edge obtained" layered over that variable stakes.

View attachment 135080


In this example i combined my final base probability with the markets probability but it could have been combining a jockey model's final probability with a trainers model final probability - the process using the exponential distribution and the logistical distribution remains the same

Compounding goes something like that:

P ~ (P1^Q1) . (P2^Q2) . (P3^Q3) .... (Pn^Qn), normalised to 1.

where Pi = Probability of network i, Qi = importance index of network i

We compute those by requiring the total information from the post-posterior probabilities to become maximum (the Shannon number).

 

[ARAZI91]

Compounding goes something like that:

P ~ (P1^Q1) . (P2^Q2) . (P3^Q3) .... (Pn^Qn), normalised to 1.

where Pi = Probability of network i, Qi = importance index of network i

We compute those by requiring the total information from the post-posterior probabilities to become maximum (the Shannon number).

I know how compound probability works but this is not coin tosses or drawing cards from a deck. It's horse racing where the true odds are never known (arguably you could assign a probability mass of 1.00 to the winner and 0.00 to the also rans after the race ) - I take it "importance index" means "significance" - Benter has already solved the horse racing problem by using a logistic function and an exponential function, further to that other creative people like Edelman and Silverman added Bayesian hierarchical "updating" to Benter's work which is now seen as essential given today's efficient markets. Certain people made a helluva lot of money by applying it to horse racing and still are!!! - i'd suggest you go read all three author's published work - there is no "compounding" involved. Not one mention!! - How would/could you compound two logits and two exponents in the traditional sense of the word when it's those two functions "working under the hood" to create a new probability??????
As for the "Shannon Number" are we talking entropy here or his chess experiment / work of 1950 - Shannon number of Chess game-tree complexity = 10120 ???????

 

[ARAZI91]

Hrmph.
By scoreline I only mean the 2nd leg was identified as non-affair.
E.g. City - Porto 5-0 in Porto then in Manchester they were playing to pass the time.

"Playing to pass the time" - imagine getting paid £175,000 for 90 minutes just to "pass the time".Just how efficient is Man City's wage function.??

 

[Outlander]

I know how compound probability works but this is not coin tosses or drawing cards from a deck. It's horse racing where the true odds are never known (arguably you could assign a probability mass of 1.00 to the winner and 0.00 to the also rans after the race ) - I take it "importance index" means "significance" - Benter has already solved the horse racing problem by using a logistic function and an exponential function, further to that other creative people like Edelman and Silverman added Bayesian hierarchical "updating" to Benter's work which is now seen as essential given today's efficient markets. Certain people made a helluva lot of money by applying it to horse racing and still are!!! - i'd suggest you go read all three author's published work - there is no "compounding" involved. Not one mention!! - How would/could you compound two logits and two exponents in the traditional sense of the word when it's those two functions "working under the hood" to create a new probability??????
As for the "Shannon Number" are we talking entropy here or his chess experiment / work of 1950 - Shannon number of Chess game-tree complexity = 10120 ???????

I read this and it’s like it’s been written by a different species, I must stop reading your posts, as it just confirms I am a complete idiot who doesn’t even understand or is even capable of understanding the things that most interest me, which is a special kind of torture.

 

[ARAZI91]

I read this and it’s like it’s been written by a different species,

Soz mate - I promise i'm human O Outlander - just a bit of a "geeky" human when it comes to numbers etc but a human all the same. Btw i mentioned i ran a correlation from fastest to slowest between both our final time figures for that day Golden Sixty ran and the coefficient came out at 0.93 - that is a seriously high number considering we use different standards, ability pars, methodology etc -in layman's terms it's 7% away from "perfection" (ie we both had the same rank of horse as fastest to slowest) - A mean squared error of 0.34 is also excellent over the 10 race sample - your no idiot in my book mate that's for sure.

 

[cosmicsports]

"Playing to pass the time" - imagine getting paid £175,000 for 90 minutes just to "pass the time".Just how efficient is Man City's wage function.??

Well they lost to Brentford just prior to the cup final when they were already champions.
It's called resource management.

 

[ARAZI91]

With Selby v. Wilson you say they started 1.70 - 2.43 and became 1.48 - 3.03.
In terms of probabilities that makes 59% - 43% and then 67% - 33%.
Do you mean that every food poisoning means I lose 25% of my initial probability, if I go and play snooker under the effect of food poisoning ?
Medical science can't do that, so what do you actually do to go from (59-43) to (67-33) ?
I mean to do it with statistical factors. How else ?
Also if you say "it's from how the market evolved" it's no good as I 'm supposed to be ahead of the market.

This part shows you have misread the whole concept - yes they started at Selby 1.70 and Wilson 2.43 those were calculated very simply by the "imaginary ratings" of 901 Selby , 629 Wilson but with "updating" became 1.87 Wilson and 2.15 Selby which brings more equalisation into the ratings (remember "food poisioned" or not Selby does not have a rating nearly a third more than Wilson for nothing) - the market remained at 1.48 Selby and 3.03 Wilson(0.35 rounded between them) - we used a degenerative factor of 0.55 which accounts for his condition, his comments and his practise session BUT he's still Mark Selby , a top 5 player (top 3 at best) , winner of two ranking events this season whilst Wilson's form has dipped a little pre-tournament, so the updated odds look about right (at best in betting you should always be conservative) the new updated odds make Wilson now favourite with around 0.06 between them - he's still a massive overlay though compared to the market which still has him at 3.03 but instead of there being 0.35 between them towards Selby there is now 0.06 between them towards Wilson. Your way ahead of the market.

 

[cosmicsports]

Another one from old time I just remembered is Man United v. Legia Warsaw for the europa cup.
United beat Legia in Warsaw by 3-0.
The return leg at Old Trafford was attended by Margaret Thatcher and Lech Walesa.
Could United go all out to thrash the Poles ? That ended in a 1-1 draw.

 

[ARAZI91]

Well they lost to Brentford just prior to the cup final when they were already champions.
It's called resource management.

Resource management or not, recently there was an Old Firm game (Rangers v Celtic) and with the title wrapped up Celtic fielded basically a reserve, fringe player team. Celtic having had the upper hand against the Gers all season lost 3-0 - you try telling the many 1000's of Celtic fans that paid good money to watch what was an embarrassment of a game about "resource management" and you would have been in the Glasgow Royal Infirmary within the hour getting stitched up and a few broken bones reset.