Calculating team's total goals in footbal

goalies_are_different · February 3, 2011

Hi! I am trying to find a decent method of calculating the expected goals of a team in football. After testing several approaches that were backed with mathematical reasoning at certain degree I am up to a very simple and straightforward method that some of you, I believe, are familiar with. Here is an example: Home teams in Seria A scored an average of 1.4 goals a game so far this season. Away teams scored 0.95 goals. Let's have the Bari-Inter game that is played tonight (well, it's actually on, being scoreless after ten minutes as I am writing) Bari score an average of 0.63 and allow an average of 1.36 when playing at home. The figures for Inter are 1.0 and 1.2 respectably (when playing away). If we divide Bari's 0.63 goals by the average goals a team scores at home (1.4) we get 0.45. Dividing Bari's 1.36 goals allowed on average by 0.95 (goals allowed by a home team on average) we get 1.43 We do the same for Inter, with the only difference being that their average goals scored are divided by 0.95 and goals allowed - by 1.4 (that's because Inter are the away team). The figures we get are 1.05 and 0.85 respectably. The final step is: Goals for the home team expected = 0.45*0.85*1.4 (That is the home team 'attacking ratio' by the away team 'defense ratio' by average goals scored by a home team in the league. The result in this example is 0.54 Doing the same thing for Inter we have 1.05*1.43*0.95 which is 1.431 So we have 1.431 goals expected by Inter against 0.54 goals for Bari. What we can do with these two figures is to plug them in to some modified Bessel functions of the first kind and we have the following odds for our game Bari 7.17 (14%) Draw 3.71 (27%) Inter 1.69 (59%) Applying 7.5% bookies margin we have roughly Bari 6.67 Draw 3.45 Inter 1.57 What we can also do is calculate the game total goals. That's easy because scoring in football is Poisson process with goals scored by the teams being independent events. In our case we have total goals expectation of 1.97 goals. Speaking of Under/Over 2.5 goals that translates into 66% chance of Under 2.5 and 34% of Over 2.5 (1.52/2.92) Unfortunately this method of estimating team totals has its flaws. First of all, in the beginning of the season there is no team stats. Another thing is that some teams may have their figures way off the averages (Chelsea scored 3.2 and allowed 0.0 goals at home in their first five games this season). Some teams have their schedule easier, hosting weak teams in early stages of the season (Chelsea had WBA, Stoke, Blackpool, Arsenal and Wolves that allowed 16 goals in these 5 games) So, with Bari and Inter still scoreless after 45 minutes, I would gladly ask, you guys, for your expert opinion on how to tweak the numbers before feeding them to the Bessel functions. I am sticking to this functions because they prove to be extremely accurate if you give them the averages for a complete season. Here are examples with low scoring, high scoring and 'normal' scoring leagues England Premiership 09/10 (2.77 goals/game) Actual results: 51% 1 25% X 24% 2 Bessel: 52% 1 24% X 23% 2 Holland 2nd Division 09/10 (3.2 goals/game) Actual results: 46 % 1 23 % X 31 % 2 Bessel: 47% 1 23% X 30% 2 France 2nd Division 08/09 (2.31 goals/game) Actual results: 47% 1 29% 2 24% X Bessel 47% 1 27% X 26% 2

brecki · February 4, 2011

Re: Calculating team's total goals in footbal Hi, first let me say that this idea of calculating goals by using statistics is an old shoe. It has been tried very very often and it will always bring your account to zero. The reason is quite simple: You use seasonÂ´s averages to calculate a single game, but as during a season many factors lose their importance because they get neutralized more and more when the number of games played increases, these factors hit you 100% in a single match. LetÂ´s create an example: You calculate a total goals expectation of 2,45 goals for a cup match. Now the game is inplay and you watch it on tv. The final result is 0-0 after hitting the post four times, two goals disallowed due to wrong offside calls and nearly three penalties not given by the referee because he misses to give the first one and then decides to not give any penalty so no team can say the referee ruled the game to a certain way. What is the result? You lost ! What are your emotions? You are frustated ! What do you thing? How can I add these factors to my formula? Here is the answer to question three: You canÂ´t !!!!! Because one week later, the rematch takes place and the following happens: You again watch the match on tv. The final result is 5-4 after hitting the post four times, but this time these all were goals as the balls hit the post on the inside. Two goals count although they were clearly offside. Two freekicks take a deflection and get in and finally the referee gives two penalties as he feels bad that day, canÂ´t run as usual and therefore sees the fouls from a bad angle. There are so many uncalculatable factors in a single game you canÂ´t give an expectation for, thatÂ´s why any calculation for a single match using seasonÂ´s averages is absolutely nonsense. To underline these statement, look at the spanish division. Using seasonÂ´s averages you never would be able to calculate a 2-0 defeat of the FC Barcelona by Hercules --- but it happened ! There were nearly 50 chances to score for Barca, but that day the ball just didnÂ´t want to get in. No formula in the world could predict such a result, because this single result doesnÂ´t depend on any past games. Every game has to be seen for itsself, and that is exacly the same to any goal prediction for a single game.

mister magoo · February 4, 2011

Re: Calculating team's total goals in footbal Gotta agree with Brecki on this one...if it was that easy we'd all be rich...:unsure

ChiChi92 · February 4, 2011

Re: Calculating team's total goals in footbal Have to disagree with you brecki. Imo you have more chance to take your account to zero by following goonersguide ratings as you do in one of your post than trying to predict goals with your own made model which you backtested, at least it's worth trying. Same for Hercules, the goal here is to find a system to identify value, not a system that will tell you Hercules WILL win. Anyway,g_a_d, let me call you gad, I understand that in your formulas your are doing a kind of geometric average: your formula simplifies to: Expected goals for home = (home scored * away allowed) / (average home scored) just a thought but how is it more logical than Expected goals for home = (home scored + away allowed) / 2 The problem with your formula is if you sum all predictions for one day in a league i'm not sure that the average nb of home goals you predict is equal to the average home goals scored historically, is it ? (whereas it works with the other formula) Regarding the points you made: -you can wait x matches for stats to settle before playing -you can add a factor to weight goals scored/allowed to take into consideration the quality of teams plaed so far

goalies_are_different · February 4, 2011

Re: Calculating team's total goals in footbal ChiChi, I have been playing the numbers for a while trying to figure out a way of calculating a weight or another 'inflation' factor. I will probably have to find a trained statistician to help me with this. The thing is that if after a complete season I can match the actual results with these Bessel functions, I should be able to do it for a single game. I am now doing kind of a reverse engineering. Take Pinnacle's odds for tomorrow Aston Villa Game, I found out that in order to match the game odds and the 2.5 goals under/over the team totals could only be 1.35 Villa, 0.85 Fulham. Using Bessel functions and Skellam distribution with this geometrical averages, it works surprisingly well for NBA and NHL games. Especially with NBA, the differences are very slight.

ChiChi92 · February 4, 2011

Re: Calculating team's total goals in footbal Do you mean that for NBA it works well for a single match ? Do you fit well to the market or get "better prices" than market meaning you have an edge ? My guess is that it works better with NBA because there is a higher number of goals, but if it works for NHL it should work for football games. But is your goal here to fit or find value ? Did you make a backtest to see if there is an edge playing every game where your price deviate significantly from the market ? Will have a look at it

goalies_are_different · February 4, 2011

Re: Calculating team's total goals in footbal Yes, it works very well for single games in NBA. Yes, the idea is to find a value bet and bet it. It works even better for hockey as there scoring is more independent and random process than in football. When I first tried it with NBA I thought that my game totals and spreads will be way off the market, but surprisingly the game totals are quite accurate and close to the market ones. There are however some games where the differences are huge, and it's probably for a reason. I don't take into account injured players in NBA (time consuming). I can back test it only with football as I have historical odds only for football. But I feel that there is still some work to be done before back testing. Back testing is the easy part, writing a program that will do it will take me two days. (I have some experience testing Joe Buchdahl's Last Six Games system which took me only a weekend)

ChiChi92 · February 4, 2011

Re: Calculating team's total goals in footbal of course still some work to do before having a good system, but no harm in seeing if it gives a 0+ ROI using just this

Kumquat Tree · February 4, 2011