Monday, 22 February 2010

Weekend Roundup

Fulham closed at 2.11 and Villa at 1.43 for another two winners. An incredible weekend with a big day on the sunday. Really cleaned up there.

Midweek numbers are being crunched and hopefully there will be a post later tonight with any bets.


  1. Right I'm back. Got away with it! All I wanted to ask really is about your ratings - from what I've gleaned it seems you produce your rating based on other people's ratings. Is that correct? Any rating system I guess would rate Manchester United a superior team to say Leeds. I'm sure your ratings though wouldn't then say just Manchester United are a bet. At 2/1 they would be, at 1/100 they wouldn't be, so at what point do you take the teams odds to win into consideration? I mean your list of teams to back and back x2 was great but surely you have a price at which a team 'should' be and compare that to what is actually available? Those with the biggest discrepancy making the list. Am I anywhere near and if so could you also put up the 'fair' price you are trying to beat? If I'm nowhere near could you explain in a little more detail please. Looking forward to seeing your picks for midweek, do you only cover England or Europe and Euro leagues? Anyway as always good luck, Mark

  2. Hi mark.

    You are on the right sort of lines - I am using my own ratings along with filtering them through a couple of other sets of ratings. Quite right about Man Utd vs Leeds. The basis of mine is the ELO system which you can find out more about on google (if you don't already know).

    The reasoning behind not providing the prices is fairly reasonable. I have tracked this season's results and found that the closing lines of the teams I am betting are not "suffering" as a rule - so I am not putting up minimum prices. The value bets were there all the way until the off.

    This is interesting since it also suggests that the market is not always as smart as markets with large liquidity are (e.g. stock markets). There's a long and boring post coming about that one day I'm sure.

    The difference in football and say horse racing is that the price movements as a rule are not as volatile. What I am doing here is identifying games where the pricing structure is significantly wrong in the first place. Any changes from this pricing structure aren't going to be serious enough to impact upon the returns (IMO) otherwise I don't put them up as bets.

    One of the major reasonings behind price movements is team news for example - but if a team is an even money shot in the first place and I think it should be much shorter (let's say 4/6) - and then they leave out 2 players and the price drifts to 6/5, its fair to say that my initial 4/6 quote would be bigger, but still would be nowhere near the 6/5.

    Also, some of the books I am placing my bets at, and betfair with a 2% account, are not prices that are achieveable to everyone. So its not really fair to put the prices up since most won't be able to achieve them. I also pick and choose my time to bet based on exactly what the asian market is doing (e.g. if I see a team on the drift, I will wait and bet later). At the weekend I noticed blackpool on a big drift, and it plateaued at about 1.84 DNB in asia. Some time later, I noticed that it had come back to 1.76 DNB and so I made the decision to take the price there and then. I got 2.51 (pinnacle). The eventual SP (if you will) was 2.35, but I don't have the time to update the blog after each price I have taken.

    There will be times when the prices I used to compile the sheets have moved significantly (i.e. to only make it fair value rather than a decent value bet), but they will be rare enough for me to not worry about their overall impact on things.

    Also, by betting before seeing team sheets etc, I don't think much of methodologies that give out the price - I don't have a degree of confidence than Man Utd should be 1.163 tonight for example, because there are too many variables to be that precise. I also don't use the kelly criterion (mostly because the stakes would be rather unmanageable on some of the prices I have in - let's say for example I did make a team 3.63 and they were 7/1 (not impossible), the correct Kelly stake would be such that I would start to move a price in any league below league one.

    I have the odd European pick but not with a degree of confidence to share them on the blog just yet.

    Thanks for your posts


  3. Very interesting. I have heard of Elo, if you mean they guy that produced the chess ratings? I understood they didn't fit football too well as it was basically designed for a simple 1X2 scenario, and whilst that is true for football as well it makes no allowance for whether a team wins 1-0 or 4-0. Maybe I understood wrong or maybe your ratings don't depend upon this allowance?

    How far back do your ratings go and how many games back do you include to create them (if you're combining I suppose it is variable between each rating)? What sort of scale are the ratings allocated to each team? Is each team individually rated and so at the the end of a given season (in general) the EPL champions will have the highest rating and those relegated from CC2 will have the lowest rating or are they league specific ratings?

    Sorry for all the questions but ratings are what I'm really trying to get into at the moment.

    Finally, for now, once you have your rating how do you convert these ratings into true odds?

    Again good luck with tonights selections, Mark

  4. Mark - I think if you do a lot more digging you'll find a lot more ELO-related (to football) rating stuff on google etc.

    I can't answer all those questions without giving away what has been my life's work for a good few years, on and off. But I can say a few things:

    i) If you have a wide knowledge of football, and something like the "notts County" or the "manchester City" effect is taking place, there's every reason to no bet those teams for as long as you feel comfortable. There's always other opportunities, other games.

    ii) "Average" rating systems like Rateform (if you have heard of it) suffer because they simply do not go as deeply into the underlying parts of the game. A 1-0 win or a 4-0 win is not the be all and end all of a game - but regularly, the possession, penetration, shots and corners that a team is enjoying/creating are a much better indicator of their true underlying ability.

    In an ELO-style system you can compare one team with another (from different leagues/countries). However matches like the Europa league ones (and a lot of cup matches) have other dynamics which are unique to those particular games and therefore need a much much more advanced model. If I was going to use ratings on those games, I'd want them to be the football equivalent of what Accuscore does for american sports - literally having a wide database down to the individual player. A serious undertaking to say the least!

    Conversion of ratings into odds - that's a thesis in itself. There's more than one way to skin that cat - but if you can create some reasonable "banding" e.g. a team with a rating of 1200 plays a team with a rating of 990 10000 times in your database, and the observed results have been x homes, y draws and z aways, and put confidence intervals etc on them, that's one way to see when the prices might be significantly out of line.

    Good luck and I hope those snippets have helped rather than served to confuse further. Its a huge undertaking and I wish you good luck.