Guest Dosheroonie Posted October 10, 2003 Share Posted October 10, 2003 I am intrigued by this poisson thingy, but totally unable to work it out. Please can someone explain so an idiot like me can understand, how to make a poisson calculation in excel to get 1x2 percentage chances, using as an example the following estimates: home team predicted goals 1.8 away team predicted goals 1.1 Thanks in advance for any help. Quote Link to comment Share on other sites More sharing options...
Guest madmick Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Poisson is a numerical distribution function, like the normal (or Gaussian) function The 'easiest' way to do what you are asking is to calculate the chance of the home team scoring any number of goals up to some maximum (e.g. 10 or 20). Use the Poisson function in Excel. It takes the form Poisson(X,lambda,false) where X is the number of goals you are interested in (e.g. 0 to 20) and lambda is the predicted goal average (1.8 and 1.1 in your example). Now do the same for each away score. You will have two lists of 10 or 20 probabilities of the given team scoring that many goals. Now you have to combine them by multiplication for each possible scoreline. If you do from 0-10 (11 scores) then you will have an 11 x 11 matrix. This is the probability of each of these scorelines. The 1X2 can be derived for this by adding together all those scorelines that result in a home win, draw or away win respectively. good luck with it - the key of course will be in how you derive your average goal figures Quote Link to comment Share on other sites More sharing options...
Guest muppet Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please hiya . i used poisson to look at under/over last season. as mick explained i worked out the prob of each score, and from this % chance under 2.5, and over 2.5 goals. got about 6% yield off 180 games. never considered it to do 1x2 good luck anyhow. Quote Link to comment Share on other sites More sharing options...
Guest Dosheroonie Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Thanks Mick I am grateful for your quick response and the info is superb. I've just got one more question (and apologies because this probably seems obvious to you). I am trying to sort it out in my head which numbers go where. Probably easiest if I put an example. POISSON(0,1.8,1.1) - is this the correct way to calculate the home team's probability of scoring 0? And then the 0 is replaced by 1, 2, 3 etc? Then for the away team is it POISSON(0,1.1,1.8) with the 0 changing in the same way? If so, the probability seems to be increasing for the obviously unlikely scores, eg 10, 11, 12. If you can show me what I am doing wrong I will be extremely grateful. Thanks again Quote Link to comment Share on other sites More sharing options...
Guest Dosheroonie Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Cheers Muppet I'm just looking into this at the mo. I am gonna try to apply it to the 1x2. (I hope). I will then be posting up a new system. What's that groaning noise ;) Quote Link to comment Share on other sites More sharing options...
tipstar Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Dosh "The probability of a team scoring X goals" where Lambda (l) is the average number of goals scored by the team, e is a mathematical constant (2,718 and X! is X faculty (X multiplied with all real numbers between 1 and X - e.g. 5 faculty = 5*4*3*2*1 or 7 faculty = 7*6*5*4*3*2*1) Example: Let's say that the average number of goals scored by Barcelona is 3,1, and we want to calculate the probability of them scoring 2 goals: p(2) = 3,1^2 * e^-3,1 / 2! = 0,216 The probability is then 21,6%. The fact that we are able to calculate probabilities of a team scoring x goals gives us the opportunity to predict the exact goal scoring for each team. Let's do another example: Arsenal meet Liverpool at Highbury: Arsenal lambda (average goals per game) = 1,8 Liverpool lambda = 1,6 We want to calculate the probability of the result 1-1: First we calculate the probability of Arsenal scoring 1 goal using the poisson formula showed above = 29,8 % We do the same with Liverpool, and find 32,3 % We now multiply these two numbers = 9,6 % In order to predict for instance the home win probability, you simply add up the percentage of all probable home win results (1-0, 2-0, 2-1 etc.). Lambda The main problem is to calculate the lambda value, i.e. the average number of goals scored by the team. It must be an average value of goals scored by the team and goals conceded by the opponent. Quote Link to comment Share on other sites More sharing options...
Guest muppet Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please cheers tipstar - way quicker than me! you still on the old poisson tipstar?? Quote Link to comment Share on other sites More sharing options...
tipstar Posted October 10, 2003 Share Posted October 10, 2003 Quote Link to comment Share on other sites More sharing options...
Guest OddsAgainst Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Tipstar, you have Rangers at 1.59 true chance and Celtic at 6. :eek :eek I would suggest that these odds are way, way out. Celtic are never a 5/1 true chance at Ibrox. Any bookmaker offering anything like this would surely be killed in the stampede. Even Rangers' fans would be queueing for odds like that! Quote Link to comment Share on other sites More sharing options...
Guest madmick Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please would also make the point tipstar that your draw %'s look very low. Quote Link to comment Share on other sites More sharing options...
Guest Dosheroonie Posted October 10, 2003 Share Posted October 10, 2003 Re: Help with Poisson calculation, please Thanks Tipstar, All this help is very much appreciated. I will be trying to get my head around all this mathematical gubbins over the weekend ;) Hopefully will be able to make some calculations before EPL resumes next week Quote Link to comment Share on other sites More sharing options...
tipstar Posted October 11, 2003 Share Posted October 11, 2003 Re: Help with Poisson calculation, please Odds ,Madmick The poisson calcs are correct, it is the number of games this season which are distorting the figures. Rangers home played 4 GF 16 GA 3 Celtic away played 4 GF 5 GA 2 Rangers ave = Rangers GF 16 + Celtic GA 2 = 18/tot no of games 4h+4a = 2.25 Celtic ave = Celtic GF 5 + Rangers GA 3 = 8/tot no of games 4h+4a= 1 Rangers 4-3 = 0.69% 4-2 = 2.07% 4-1 = 4.14% 4-0 = 4.14% 3-2 = 3.68% 3-1 = 7.36% 3-0 = 7.36% 2-1 = 9.81% 2-0 = 9.81% 1-0 = 8.72%.....R/total = 57.78 DRAW 0-0 = 3.88% 1-1 = 8.72% 2-2 = 4.91% 3-3 = 1.23% 4-4 = 0.17%....D/total = 18.91 CELTIC 3-4 = 0.31% 2-4 = 0.41% 1-4 = 0.36% 0-4 = 0.16% 2-3 = 1.64% 1-3 = 1.45% 0-3 = 0.65% 1-2 = 4.36% 0-2 = 1.94% 0-1 = 3.88%.....C/total = 15.16 Rangers win % = (57.78/91.85)*100 = 62.90% Draw % = (18.91/91.85)*100 = 20.58% Celtic win % = (15.16/91.85)*100 = 16.50% The %age may be slightly out to my s/sheet as this rounds up the no's to 2 decimal places. also the total divisor (is that the correct name?) will not be 100 as I stop each calc at 4-3 for home and away and 4-4 for draws which causes a slight error . Quote Link to comment Share on other sites More sharing options...
tipstar Posted October 11, 2003 Share Posted October 11, 2003 Re: Help with Poisson calculation, please Dosh you may want to play around with this for a while until you get your own sheet set up Poisson Quote Link to comment Share on other sites More sharing options...
Guest OddsAgainst Posted October 11, 2003 Share Posted October 11, 2003 Re: Help with Poisson calculation, please Tipstar, it seems to me this method does not take into account the gulf in class between the Old Firm and the rest of the league. Let's forget about the number of games played for the moment. Rangers are bringing to the game a goal expectation earned mainly against teams of a lower class. The maximum possible number of games involving Celtic in the average is one. (if this game is the second in the season). Therefore Rangers are coming into the game with a goal expectation gained against teams of a lower class than their opponents. This can only be increased by adding in their opponent's average goals conceded. So we now have a goal expectation against Celtic which is higher that what they have actually achieved against opposition of a lower class. Whereas I would expect Rangers to score on average 3 goals at home against Dundee for example, I would only expect them to score on average 1.31 goals against Celtic at this time. Have you thought about using a moving average instead of the games played this season? That way you don't have a problem with too little data at the beginning of a season. Instead of using games played this season you could use the last 228 games (for example, choose what you think will work best) played to compute your averages. Few teams change radically from season to season and those that might are obvious (Chelsea) and you can keep a special eye on them and also the promoted / relegated teams who will be bringing averages from another league. Quote Link to comment Share on other sites More sharing options...
Guest Dosheroonie Posted October 11, 2003 Share Posted October 11, 2003 Re: Help with Poisson calculation, please Cheers tipstar, That looks extremely useful. I will certainly be using that for a little while, as it is taking me a while to get my own spreadsheet up and running. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.