Elo rating system (my thoughts)

[ELO Evangelist]

The fact that you have entered this thread means you probably know what an ELO rating system is but in case you do not here is a brief explanation. Arpad Elo a physics professor and chess master realised that chess needed an accurate ranking system and in the 1970 devised a system based on statistical estimation over a natural distribution curve. This system of statistical estimation is widely regarded as the most accurate mathematical model for predicting probabilistic outcomes. It is used today by many sports but has been improved many times along the way, and herein lies the problem, not all ELO rating systems are created equal. Different organisations use different ELO rating systems with differing degrees of accuracy. You would think that an organisation would want it's rating to be accurate but as we know this is not the case. Take FIFA's 3 points for a win and 1 point for a draw regardless of oppositions strength or whether the match was home or away. We know this system can at best only be described as moderately accurate. The same goes for many ELO systems used at the moment. Ease of use overrides mathematical complexity and as a consequence accuracy suffers also. Of the 3 most widely used natural distribution curves

Gaussian (normal), Logistic and Poisson, Arpad chose the normal curve to model chess but this turned out to be not the most accurate and the Logistic is now used by some chess federations and the normal is used by others. Another shortcoming of Arpad's original ELO system was the K factor it was found that early in a ranking estimation the value should be large and then as the team or player's rating (μ [pronounced miu] value) is known then this factor should be reduced. The answer to this problem became ELO systems with clumsy stepped K factors as the rating became higher or constant K factors. Whilst at university reading computer science I found myself helping one of my maths professors code a custom ELO algorithm. The work was being done by the university for a major Internet IT company for the primary purpose of page ranking. The System the maths team created was mind numbingly complicated but featured algorithms that create it's own custom distribution curve and a variable K factor. One more thing that this system could do that most ELO algorithms can not is take into account draws and the affect they have on the win loss distribution variance accurately. Since being exposed to this algorithm and writing most of the code for it I have become an evangelist for ELO based systems ever since and have been using them to my advantage for a few years. You would not believe the number of punters I know that think the existing soccer league tables are accurate. I regularly search for ELO based research papers to see if the system that my university devised and is in daily use by millions of Internet users has been surpassed but it seems still to be the best. It is a strange fact that organisations not not require their ELO systems to be as accurate as possible. Take the FIFA world rankings or most domestic ELO systems you will come across. They average together home and away performance to give and average ELO rating because they think your average soccer fan can not understand a system more complicated than a single figure. So keep up the good work guys because I think in a few years time everyone will know about ELO based ranking systems because the computing power needed to use them is now on every one's desktop. You can guarantee that the pro syndicates and the bookmakers have been using ELO systems for years.

[carlos50]

Re: Elo rating system (my thoughts) Is there a research paper about the algorithm you helped build? Other main research papers you can recommend?

[ELO Evangelist]

Re: Elo rating system (my thoughts) There is a great deal of research matter out there on the internet. In this particular case I would suggest starting with wikipedia. Take a look at the following articles and scroll down to the bottom of each page and you will find links to the research and from there you will find more links etc. etc. You might even read the research on the algorithm I helped code ;) http://en.wikipedia.org/wiki/Elo_rating_system http://en.wikipedia.org/wiki/Bayesian_inference http://en.wikipedia.org/wiki/Bayesian_network But be warned you are going to run into a wall of mathematics. I understand hardly any of the maths in these papers but I spend a lot of time on maths forums getting help. If you are interested in programming your own work then few people know Microsoft have effectively developed a language to help you do it. The original work I did was in FORTRAN but recently I ported it to .NET http://research.microsoft.com/en-us/um/cambridge/projects/infernet/docs/Two%20coins%20tutorial.aspx

[methuselah]

Re: Elo rating system (my thoughts) Your Post was interesting ELO but probably all of us here would be unable to follow the Maths you mention. I should imagine that the out- come was another method of preparing League Tables. Is this so.? If so, what are the results you have come up with. Your Post leaves me looking for the end of the Rainbow.:wall

[carlos50]

Re: Elo rating system (my thoughts) The math is less scary than the programming, so i'm not to worried about the wall of math. I usually prefer a top-down approach, starting where I want to end, hence: please show me the end of the rainbow :D Just kidding. It seems Glickmans webpage is a good place to start, I'll start there. I was hoping though you could give me a title of the paper you assisted? Thx

[pacey]

Re: Elo rating system (my thoughts) Hi, Betting Genius Is there a program that processes the bets with the theory Elo. Someone knows it is reliable?

[ELO Evangelist]

Re: Elo rating system (my thoughts)

Your Post was interesting ELO but probably all of us here would be unable to follow the Maths you mention. I should imagine that the out- come was another method of preparing League Tables. Is this so.? If so, what are the results you have come up with. Your Post leaves me looking for the end of the Rainbow.:wall

You are correct IMO an accurate ELO algorithm will give you a more accurate indication of a teams ability. Here are my league standings for the end of last season. I Rank all the worlds major leagues game by game. You will notice some interesting results from the top of the league. I think Man U is better than Chelsea, Liverpool are a top four team and Man C are not as good as people think they are. 29.944 Man U 29.393 Chelsea 28.644 Arsenal 27.975 Liverpool 26.222 Everton 26.061 Aston V 25.946 Spurs 25.254 Man C

[Lardonio]

Re: Elo rating system (my thoughts) I'm skeptical towards this until someone shows me some proof that ELO ratings actually produce profitable betting results over time. I'm sure it won't produce league tables that are 'accurate', as an 'accurate' league table is a highly subjective concept and there are so many variables to be considered that nobody can produce an 'accurate' league table, and especially not a mathematician. However I'm sure it will produce a league table that is different, as any other ratings system would. You can gift wrap all the fancy math in the world and use a lot of big words, but it's all just about calculating numbers from a formula and if it does not automate the process of producing consistent profit in betting then it's all just playing with numbers. Also there is no reason why a ratings system based on the ideas of Arpad Elo, created for chess, a vastly different sports from football, would perform any better than a ratings system produced by anyone else. Sure his ideas are worth reading but producing a rating system is not necessarily about complexity and professor-level math, it's about using the information available in a way that consistenly produces an alternative and profitable way of approaching betting based on value. I'm sure lots of people already use Elo-like ideas in their betting systems today, and most of the ratings systems I've seen or created myself use the same theories as Elo's. I think some people's love of complexity and numbers overshadow the target of making a profitable betting system, and I know from experience that it's easy to be fooled into thinking you are a genius when you compile huge formulas and gigantic databases producing lots of numbers that look good. Sorry to be negative, but unless you start posting profitable selections I can't get excited, but I do welcome a non-mathematical debate that could improve any ratings system. As for the info you published above this post, it's not exactly revolutionary. I think almost everyone from simply observing would agree that Liverpool were unlucky last season and that Man City were over-rated. After 38 games you have the exact same 7 teams in the top as the actual league table and I would find it hard to believe that you could use that info for something useful. How would you profit from "knowing" after the whole league is finished that Manchester United are a tiny, tiny fraction better than Chelsea (1 point separated them!!)? There is something many ratings systems share; even the best ones produce information that 99% of football interested people could get just by looking at the league table or the form data.

[ELO Evangelist]

Re: Elo rating system (my thoughts)

The math is less scary than the programming, so i'm not to worried about the wall of math. I usually prefer a top-down approach, starting where I want to end, hence: please show me the end of the rainbow :D Just kidding. It seems Glickmans webpage is a good place to start, I'll start there. I was hoping though you could give me a title of the paper you assisted? Thx

The strength of ELO based models are that once you have an accurate set of ratings then you can calculate the chances of a win, loss or draw accurately. The problem comes when you have to decide who's ratings are the most accurate because they all seem to be different from my experience. This is where confidence in a strong algorithm is the key. There is a lot of argument as to whether Glickman's more complex algorithm is more accurate than Elo's original simple algorithm. IMO it is but this is based on reading hundreds of arguments for and against rather than empirical evidence. The time factor used in Glickman's algorithm may not lend itself to soccer games, aso since Glickman lost his Patent request for his algorithm he has stopped developing it so you would have to assume he intended it to be developed some more.

[ELO Evangelist]

Re: Elo rating system (my thoughts)

Sorry to be negative, but unless you start posting profitable selections I can't get excited, but I do welcome a non-mathematical debate that could improve any ratings system.

I was of the same mind as you until a few years ago. As a punter I have been bombarded with 'bookie bashing systems' and 'secret staking plans' at every turn. If you can ignore the complexity of the maths and read through texts on Bayesian maths you will see that the maths world is aware of established and proven ranking models.

[Lardonio]

Re: Elo rating system (my thoughts) I'm afraid you're in the same league as the 'bookie bashing systems' until you show me the money ;)

[Machine]

Re: Elo rating system (my thoughts)

rating system is not necessarily about complexity and professor-level math, it's about using the information available in a way that consistenly produces an alternative and profitable way of approaching betting based on value.

And how would this 'value' be found ? please no: flipping coin,dealing cards rolling dice and "in the long run" analogy's. (I've always wondered how long 'in the long run' actually is --a day--a year --a lifetime) were talking sports betting here. :ok

[Lardonio]

Re: Elo rating system (my thoughts) The only way to establish that what you're doing - whether it be ratings, analyzing form and injury news, consulting psychics, whatever - produces value is, unfortunately, by looking at your betting history summary (for, let's say, 1000 bets) and see whether you are making money or not.

[Machine]

Re: Elo rating system (my thoughts)

The only way to establish that what you're doing - whether it be ratings' date=' analyzing form and injury news, consulting psychics, whatever - produces value is, unfortunately, by looking at your betting history summary (for, let's say, 1000 bets) and see whether you are making money or not.[/quote'] but what if those 1000 bets were a profitable statistical blip and your next 1000 bets you lost a shed load of cash is that still value? The point Im making is 'value' is bandied about like confetti on message boards by all and sundry and yet not one person can say what it is. One year you make a load of cash then crow like a rooster 'im a value punter' The next year doing exactly the same you lose a load of cash, thats when the old "well in the long run " statement comes out. No one will ever no because you never know when your 'value bets' are going to hit such a losing run that you cant afford to bet them any more and they will. The moral is be gratefull for your winnings but never think you have mastered 'value'

[Datapunter]

Re: Elo rating system (my thoughts) There is indeed no substitute for a real life betting record. And by that i mean bets actually placed, nothing less. I remember a long long time ago.... ( do i sound old now, fair enough ) i researched a horse racing system which showed excellent promise after the first set of backtesting, i then proceeded to papertrail to see if i could get the prices like the ones in the backtest, so far so good. Just one catch, as it often goes with horse racing there where a series of losers and then a nice winner at a price of 8.00 or even 14.00 yammy. And of course these where all included in the backtesting. In real life however i was not able to bet each and every day and as luck would have it the 1 day in 10 i could not bet happened to be the day with the 8.00 winner which i missed. Missing that single winner meant the system was overall showing a loss and not a profit. Totally throwing out any expectation from the backtesting. Didn't see that one coming. I know it's not directly related to the subject of value here but it's just one example how real life betting can differ from what you might expect from any form of research. (It doesn't invalidate the research btw.) Same thing can happen with value, value shown up in research may or may not be present in real life, only one way to find out.

[Lardonio]

Re: Elo rating system (my thoughts) I see where you are coming from monkeys nest and I agree with you to some extent, but if I had such a defeatist outlook towards betting I would stop altogether. It's trying to find value and beat the bookmakers that make betting interesting, and of course it is bloody difficult.

[Datapunter]

Re: Elo rating system (my thoughts) Value exists when the price you get is higher than the price implied by the change of an outcome happening. That part is quite simple. What is not simple is the determination of the chance of an outcome happening. In fact you can never determine a number as in sports betting there is alway a number of randomness to the event. You can only get an approximation. (which totally invalidates any kind of Kelly staking, but that's a different conversation) So the only way to actually establish if you are any good at identifying value is to measure the results over a period of time. I.e. you need to be showing a profit after X bets. The second tricky thing is that value is not fixed over time. Circumstances change so you must constantly measure and adjust your methods. The nasty bit here is that by the time you find out your methods no longer produce value bets you will already have lost a bundle. Hey, nobody said this stuff was easy. The concept of value however should be understood to the bone by anyone serious about sportsbetting. It's like gravity or the alphabet. For starters it's the reason bookmakers make a profit. Think about it. By applying an overround of on average 10% they create their own value. Same with Betfair really as the 5% comission is a negative value right from the start. If you don't understand that you've got no chance. if you do then maybe, just maybe, good luck.

[ELO Evangelist]

Re: Elo rating system (my thoughts)

... I would find it hard to believe that you could use that info for something useful. How would you profit from "knowing" after the whole league is finished that Manchester United are a tiny, tiny fraction better than Chelsea (1 point separated them!!)?

This is where Linear Regression, the next piece of the puzzle goes. http://en.wikipedia.org/wiki/Linear_regression http://en.wikipedia.org/wiki/Regression_analysis Again this is not 'Bookie Bashing' rhetoric, but sound and established maths. A linear regression line has an equation of the form Y = a + bX, where X is the explanatory variable and Y is the dependent variable. The slope of the line is b, and a is the intercept (the value of y when x = 0). From analysis of 12,342 results we get: Home win: Slope = 8.28, Intercept = 31.92 Away win: Slope = -5.65, Intercept = 34.76 Draw: Slope = -2.63, Intercept = 33.31 X = Rating difference Y = Predicted outcome %

[ELO Evangelist]

Re: Elo rating system (my thoughts)

Hey, nobody said this stuff was easy.

:notworthy

[carlos50]

Re: Elo rating system (my thoughts)

And how would this 'value' be found ? please no: flipping coin,dealing cards rolling dice and "in the long run" analogy's. (I've always wondered how long 'in the long run' actually is --a day--a year --a lifetime) were talking sports betting here. :ok

How can you say: no flipping coin analogies etc? Betting on sports is based on the same theory, as betting on a coin flip or in a casino. The difference is when flipping a coin you know the actual probabilities. When sports betting you are trying to estimate the probabilities, to find value in the odds. Wether you do it by crunching numbers or a qualitative analisys is not important, it is basicly the same thing. @Lardonio As above quantitive or qualitative analisys is just two different methods of acomplishing the same thing. I don't see a reason to neglect the math heavy methods just because you don't like them. As you say basicly it is to try to automate the process. It would surprise me a lot if nobody have done that, they just havent published the results risking making the value disappear. At least thats what I think. I wouldnt look for them in the secret bookie bashing systems though :D Personally I do it because I enjoy the process, wether I actually find the holy grail is less important.

[Machine]

Re: Elo rating system (my thoughts)

How can you say: no flipping coin analogies etc? Betting on sports is based on the same theory, as betting on a coin flip or in a casino. The difference is when flipping a coin you know the actual probabilities. When sports betting you are trying to estimate the probabilities, to find value in the odds. Wether you do it by crunching numbers or a qualitative analisys is not important, it is basicly the same thing.

SIGH :eyes (thats why there are so many losers) Betting on coin flips etc you know the probabilities..... Sports betting you dont its all a guess how people estimate that guess differs, whether you do it by maths,intuition or ouija board but its still a guess. Also you only know whether it is value or not after the event in sports betting, as it is a 1 off event. Every single sports bet is an independent event so you cannot say "in the long run". Prove me wrong and show me, its as simple as that Lardinho I am not being negative, any man jack can come on here and spout about value as if you just pick it off a tree. The problem that I have with people like that is they can never tell everone how easily they find it.(maybe because they cant or am I just being negative):ok

[Vilamoura04]

Re: Elo rating system (my thoughts) Im an ELO believer but Im not sure I understand the point of this thread :unsure Who are the millions of internet users using this brilliant system every day?

[Machine]

Re: Elo rating system (my thoughts) the thread went off track when Lardinho said the only way was value betting , so I asked for this mystical system which everyone seems to know. But alas yet again I do not get the answer to my quest of someone who can actualy show me this brilliant system rather than just say 'value betting the way to go' ;)

[carlos50]

Re: Elo rating system (my thoughts) @ Monkeys Nest Nobody said it was easy. Of course value betting is the only way. Obviously it will be a guess, but an educated guess can be good, not only within sportsbetting. You dont know the wether it was a value bet after a single bet, the actual outcome isnt important. Every single coin flip is a single independent event, still the only difference is that you know the probabilty.

[Machine]

Re: Elo rating system (my thoughts)

You dont know the wether it was a value bet after a single bet, the actual outcome isnt important.

:eyes :wall Then you don't know whether it is value or not so how can you say it is. Doesn't that register. It is a single bet never to be repeated (capiche), beforehand you didn't know its true chance of winning and you only guessed it. If it wins you can pretend you know about 'value' if it loses you throw out the old "if I did the same bet 100 times" routine. Either way you are wrong. At least with a coin you can say it has a 50% chance. Anyway enough of this, this is ELO's thread about elo ratings not the 'mythical value bet'

[Machine]

Re: Elo rating system (my thoughts)

This is where Linear Regression, the next piece of the puzzle goes. http://en.wikipedia.org/wiki/Linear_regression http://en.wikipedia.org/wiki/Regression_analysis Again this is not 'Bookie Bashing' rhetoric, but sound and established maths. A linear regression line has an equation of the form Y = a + bX, where X is the explanatory variable and Y is the dependent variable. The slope of the line is b, and a is the intercept (the value of y when x = 0). From analysis of 12,342 results we get: Home win: Slope = 8.28, Intercept = 31.92 Away win: Slope = -5.65, Intercept = 34.76 Draw: Slope = -2.63, Intercept = 33.31 X = Rating difference Y = Predicted outcome %

What sort of R² results are you getting for H/D/A

[gloryhunters]

Re: Elo rating system (my thoughts) I used a concept of ELO when I did my first system and it made a slight profit, but the edges are very small. The problem with the ELO system is that it misses so much when looking at football matchs as style of play come into effect. For example say Arsenal and Man U had the same rating, I would bet on Utd to win at Blackburn in a tough battle over Arsenal, whereas I think Arsenal have a better chance away at a more footballing side such as Chelsea and Liverpool than Man U. Another thing the system misses is weather, derby games, injuries etc. I assume that the bookies will be using formula and regression methods in picking their prices and I think the best value is to find the outliers rather than try to replicate a possible bookie system.

[ELO Evangelist]

Re: Elo rating system (my thoughts)

What sort of R² results are you getting for H/D/A

Home = 0.9429 Away = 0.9428 Draw = 0.6927 Makes your mouth water doesn't it. Even my maths mentors are impressed they told me to aim for 0.7 across the board but the Home/Away figures far surpassed this.

[ELO Evangelist]

Re: Elo rating system (my thoughts)

Who are the millions of internet users using this brilliant system every day?

I think hardly anyone is using ELO based models. And even if they are I wonder how many of them know how accurate their algorithm is compared to the others.

[Machine]

Re: Elo rating system (my thoughts)

Home = 0.9429 Away = 0.9428 Draw = 0.6927 Makes your mouth water doesn't it. Even my maths mentors are impressed they told me to aim for 0.7 across the board but the Home/Away figures far surpassed this.

Looks good for the H/A