servecold Posted March 24, 2009 Share Posted March 24, 2009 Re: Elo Access Function cant believe it was so simple extended AB3000:nana:$ Quote Link to comment Share on other sites More sharing options...
sjuesju Posted April 20, 2009 Share Posted April 20, 2009 Re: Elo Access Function There is something I can't understand. The question is: Is there a software that is being used for ELO ratings and the punter simply put the results or the punter do everything by his own on an excel sheet - the results, the tables and the predictions. Thank you in advance :) Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted April 20, 2009 Author Share Posted April 20, 2009 Re: Elo Access Function There is something I can't understand. The question is: Is there a software that is being used for ELO ratings and the punter simply put the results or the punter do everything by his own on an excel sheet - the results, the tables and the predictions. Thank you in advance :) Hi sjuesju - its a bit of both!! Asc and myself use Access to import all the historic football match data into a database with various routines, functions and queries to generate the ratings. On the other hand, there is an excel spreadsheet by MetallicIan further up the thread which should help you get a good start on maybe a single league. It really depends on what you feel most comfortable using to generate your ratings, but bear in mind that they have to be updated each week to enable you to produce a prediction so the easier you can make it the better and less time consuming it is for you. Quote Link to comment Share on other sites More sharing options...
Fast Eddie Posted April 23, 2009 Share Posted April 23, 2009 Re: Elo Access Function Good stuff, I'm interested in having a go at a rating system but don't really know where to start so I'll have a look into this elo thing and take a gander at MatellicIan's sheet, would be cool to create a macro to do the downloading and updating from footballdata so I migh try and think about that although my excel skills are moderate at best and as for VBA, well, I generally just copy and paste stuff from the various help sites across the web which I suppose is a good way of getting what you want if you've got the time and application. edit. I've got the elo file and unzipped it but it's an xlsm file and it won't open, is it not for excel (I use 2002 version) or am I being thick? Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted April 23, 2009 Author Share Posted April 23, 2009 Re: Elo Access Function Good stuff, I'm interested in having a go at a rating system but don't really know where to start so I'll have a look into this elo thing and take a gander at MatellicIan's sheet, would be cool to create a macro to do the downloading and updating from footballdata so I migh try and think about that although my excel skills are moderate at best and as for VBA, well, I generally just copy and paste stuff from the various help sites across the web which I suppose is a good way of getting what you want if you've got the time and application. edit. I've got the elo file and unzipped it but it's an xlsm file and it won't open, is it not for excel (I use 2002 version) or am I being thick? Oh aye mate - one click button for me! Downloads zip file with all the data, then fixtures, unzips the file, merges all the csv files into a single one, few delete/append/update queries and I'm done. All takes around 2 minutes. The important thing to understand is knowing what you want to do and that if you can do it manually by clicking downloading etc. it CAN be automated!! Quote Link to comment Share on other sites More sharing options...
MetallicIan Posted April 23, 2009 Share Posted April 23, 2009 Re: Elo Access Function Good stuff, I'm interested in having a go at a rating system but don't really know where to start so I'll have a look into this elo thing and take a gander at MatellicIan's sheet, would be cool to create a macro to do the downloading and updating from footballdata so I migh try and think about that although my excel skills are moderate at best and as for VBA, well, I generally just copy and paste stuff from the various help sites across the web which I suppose is a good way of getting what you want if you've got the time and application. edit. I've got the elo file and unzipped it but it's an xlsm file and it won't open, is it not for excel (I use 2002 version) or am I being thick? Hiya Eddie, yes you're correct, it is an Excel 2007 with macro file. I think you may be able to convert it for use with your Excel 2002 by downloading the Microsoft Office Compatibility Pack for Word, Excel, and PowerPoint 2007 File Formats here's the link... http://www.microsoft.com/downloads/details.aspx?FamilyId=941b3470-3ae9-4aee-8f43-c6bb74cd1466&displaylang=en Give it a go anyway, any problems with converting or using the file just PM me and we'll sort it. Ian. :ok Quote Link to comment Share on other sites More sharing options...
Fast Eddie Posted April 25, 2009 Share Posted April 25, 2009 Re: Elo Access Function Ian, Thanks for that link, it all works a treat now, very kind of you to take the trouble to help me out, and post up the sheet too. I'll get to grips with it in the next few days when I've got some time. Ed Quote Link to comment Share on other sites More sharing options...
daniel_5231314 Posted August 1, 2009 Share Posted August 1, 2009 Re: Elo Access Function Ian, thanks for that link first, but when i click on macro link, excel file stop all ,that why? what can i do? sorry for my english.. Dan Quote Link to comment Share on other sites More sharing options...
landheha Posted August 12, 2009 Share Posted August 12, 2009 Re: Elo Access Function I think the goal difference part of this ELO like rating system smells a bit. Football is not like chess. Man United is happy to win 1-0 from Telford if that saves energy and injuries for the next match. If Man United goes all out against Telford for the full 90 minutes and wins 20-0, it should not be rated very spectacular as well. At least the start rating difference should be taken into account for goal difference. Even better the home/away data should be taken into account as well. Quote Link to comment Share on other sites More sharing options...
MetallicIan Posted August 12, 2009 Share Posted August 12, 2009 Re: Elo Access Function I think the goal difference part of this ELO like rating system smells a bit. Football is not like chess. Man United is happy to win 1-0 from Telford if that saves energy and injuries for the next match. If Man United goes all out against Telford for the full 90 minutes and wins 20-0, it should not be rated very spectacular as well. At least the start rating difference should be taken into account for goal difference. Even better the home/away data should be taken into account as well. If you can make any improvements to the spreadsheet by all means do so and paste it up for us. cheers :ok Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 12, 2009 Author Share Posted August 12, 2009 Re: Elo Access Function I think the goal difference part of this ELO like rating system smells a bit. Football is not like chess. Man United is happy to win 1-0 from Telford if that saves energy and injuries for the next match. If Man United goes all out against Telford for the full 90 minutes and wins 20-0, it should not be rated very spectacular as well. At least the start rating difference should be taken into account for goal difference. Even better the home/away data should be taken into account as well. You rarely get a GD of 8 let alone 20 and IMO the adjustment adds a fair reflection on a games outcome and how the points should be divi'ed up. In my example from a few posts above we'd get a range difference of 41.85-10.8 = 31.05 points for scoring up to 19 more goals. Thats presuming the teams started on 1000 points. GD adjustment table: GD Adjustment 1 1 2 1.5 3 1.75 4 1.875 5 2 6 2.125 7 2.25 8 2.375 9 2.5 10 2.625 20 3.875 Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 23, 2009 Share Posted August 23, 2009 Re: Elo Access Function a1ehouse, Have you tried to modify the system by changing the ammount of points that the home team gets for the home field advantage? (100 points if I am not wrong). I am working on the same ELO calculations for the Premiership and I believe that the formula can be tweaked a bit by changing the ammount of points given for HFA. I would suggest replacing 100 with 85. Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 23, 2009 Author Share Posted August 23, 2009 Re: Elo Access Function a1ehouse' date=' Have you tried to modify the system by changing the ammount of points that the home team gets for the home field advantage? (100 points if I am not wrong). I am working on the same ELO calculations for the Premiership and I believe that the formula can be tweaked a bit by changing the ammount of points given for HFA. I would suggest replacing 100 with 85.[/quote'] Not looked at it myself, stuck with the 100pts for the hometeam and 30pts as the weight constant. Is this something you have already applied to your own calculations? Any observations? Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 23, 2009 Share Posted August 23, 2009 Re: Elo Access Function I am still going to backtest the calculations, but there is very good reason that makes me think 85 is more accurate. 85 is a result of calculation that provides a ratio between the points per game won by home team and away team. I saw it in a paper by an american guy who tries to adapt ELO system for MLS. It really makes sense. I can post the link to that paper if you like. Quote Link to comment Share on other sites More sharing options...
Lennon Posted August 24, 2009 Share Posted August 24, 2009 Re: Elo Access Function I am still going to backtest the calculations' date=' but there is very good reason that makes me think 85 is more accurate. 85 is a result of calculation that provides a ratio between the points per game won by home team and away team. I saw it in a paper by an american guy who tries to adapt ELO system for MLS. It really makes sense. I can post the link to that paper if you like.[/quote'] Yep that american guy has it right ... I'm using his formulas to create my own ELO ratings for as many leagues as I can. That 85 number needs to be adjusted though since not all leagues have the same home field advantage. A little OT but does anyone know how I could make ELO ratings for tennis? Has this been done by someone already? My only problem really is figureing out a way to adust the ratings for the amount of games won by a certain play ... so games instead of goals ... Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 24, 2009 Author Share Posted August 24, 2009 Re: Elo Access Function Both interesting posts - please post the url goalies, and adjusting the weighting for different leagues may also be a useful refinement given the variations between the teams between the leagues, may do some more work on this! e.g. currently I have the Liverpool game tonight rated at a 90% chance of them winning with a 100pt hometeam advantage. If I adjust this to 85pts they slip to 89% - only one percent but could be significant. Quote Link to comment Share on other sites More sharing options...
Lennon Posted August 24, 2009 Share Posted August 24, 2009 Re: Elo Access Function Both interesting posts - please post the url goalies, and adjusting the weighting for different leagues may also be a useful refinement given the variations between the teams between the leagues, may do some more work on this! e.g. currently I have the Liverpool game tonight rated at a 90% chance of them winning with a 100pt hometeam advantage. If I adjust this to 85pts they slip to 89% - only one percent but could be significant. Here's what golies was talking about ... http://mlselo.f2f2s.com/MLS_Elo.pdf Quote Link to comment Share on other sites More sharing options...
Fat Jack Posted August 25, 2009 Share Posted August 25, 2009 Re: Elo Access Function A little OT but does anyone know how I could make ELO ratings for tennis? Has this been done by someone already? My only problem really is figureing out a way to adust the ratings for the amount of games won by a certain play ... so games instead of goals ... Lennon Started looking at this last month. Have some rough and ready figures that need checking. Drop me a PM to discuss further. Maybe start another thread... ;) FJ Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function Sorry for being off, guys. Yep, that's the link I was taliking about. Tonight am going to run the calculations for the Premiership, but I have bad feeling about the regression, I hop I won't be right about it. Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Re: Elo Access Function I ran the changes on some of the smaller leagues cos they were quickest to do and found that the number of selections decreased, but the number of winning selections and ROI was not improved that much. Will look at some of the bigger leagues tho to see if it's the same across the piece, but ATM using 100 pts still gives a bigger ROI. Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function a1ehouse, how many unique rating differences do you have for 6250 Premiership games? Do you keep the figures up to 2 decimal places or round it up? Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Re: Elo Access Function 6235! Stored and calculated as a Double data type and not rounded at all. Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function So I was right about it, they are all unique. I am trying to imagine what the regression would look like. On the X axis you would have 6235 unique values and only 0 and 1 corresponding to them. Am I right? Wouldn't that affect the fitting quoefficient? Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function So, that's what you used as a criteria for the regression and not the simple rating difference? If so, that makes perfect sense, I will quickly sort my table by winning expectancy. And 5/95 cut off is probably the borders of the observation - ignoring everything below 5% and 95 percent% winning expectancy due to the small number of such games? Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Re: Elo Access Function Aye, otherwise you'd get an x-axis the size of the great wall! The cut off I initially used was =95% (home team likely to win) ignoring everything else in the middle. However look at the Liverpool rating the other night 90% to win yet they still got tonked!! :sad Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function I understand, but looking at the win expectancy formula I see that it is actually derrived by using the rating difference. Having 6235 unique difference means the same number of wining expectancies, no? (Sorry for all these questions, but this has been bugging me for weeks already) Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Re: Elo Access Function That's what a forum is about mate! Not really sure what you mean. I grouped them by Win Expectancy and counted the number of occurrences of each full time result: e.g. out of the 24 games where I predicted the We to be 7%: 0 ended in a home win 20 ended in an away win 4 ended in a draw I then expressed these as a percentage: H 0/24 = 0% A 20/24=83% D 4/24=17% and plotted these against the We. x-axis is basically We 0-100%, y-axis is the calculated %. Quote Link to comment Share on other sites More sharing options...
a1ehouse Posted August 26, 2009 Author Share Posted August 26, 2009 Quote Link to comment Share on other sites More sharing options...
goalies_are_different Posted August 26, 2009 Share Posted August 26, 2009 Re: Elo Access Function The crucial point I don't get is how exactly you group them if we assume that by having 6235 unique RD we should have 6235 unique winning expectancies. In order to get 24 games with WE of 7% you should have 24 games with the same rating difference, am I correct? Everything else I understand perfectly, the chart etc. 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.