Jump to content

A football rating system, discussions, ideas


Recommended Posts

Re: A football rating system, discussions, ideas

Very interesting thread, been following from the start and was going to wait for more details , but can't wait . What about " new manager syndrome " and local derbies which last season I would have avoided. After joining P.L. and getting a basic understanding of value I follow these with interest and even considered a strategy Keep it coming Mr O
Well I would say here, that even if these factors exist and are significant, they are of secondary importance to the metric that you choose - and are another case of fine tuning the model which I tend to ignore completely - one of the problems I think with a lot of systems is they get unnecessarily complex in terms of the number of variables used at a very early stage. It's then difficult to see if the real metric has any value or not. I'm sure there will be some thinking that what I'm presenting here is a very complex model - well, mathematically speaking it couldn't be simpler - just one very basic metric, and an elegant least-squares procedure that takes care to equalise teams that have played different numbers of games, had different schedules etc. One thing - at the end of the day when you have your model, done all your testing, and are ready to stick some cash on - now thats when the subjective side can really kicks in - so you've read all about the game, there are injuries, suspensions, local derby?...well, it didn't factor into the testing, and didn't make a problem then, so why should it now? Having the discipline to really follow even a well tested system is incredibly difficult - I have a problem with that part myself....
Link to comment
Share on other sites

  • Replies 265
  • Created
  • Last Reply

Top Posters In This Topic

Re: A football rating system, discussions, ideas

Mr O, My ratings are all in a database designed by that good chap, Jeffers. All ratings threads have been updated in the Britain forum if you need to have a peek. ;)
Ah Paul - yes, would like to put your ratings in. I'm an absolute database novice I'm afraid - know absolutely nothing about them. Perhaps you could get Jeffers to put the data in a spreadsheet - just need columns with Home Team, Home Peformance, Away Team, Away Performance...for every match That would go for anyone who wants to test a metric out here by the way...very easy for me to plug it in, in that format...
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

Ah Paul - yes, would like to put your ratings in. I'm an absolute database novice I'm afraid - know absolutely nothing about them. Perhaps you could get Jeffers to put the data in a spreadsheet - just need columns with Home Team, Home Peformance, Away Team, Away Performance...for every match That would go for anyone who wants to test a metric out here by the way...very easy for me to plug it in, in that format...
You're not alone Mr O, I can't even design a webpage, let alone a database. :) You can ask Jeffers, just give him a PM, just don't say I sent you. :lol
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Hi Muppet, Been trying to post up a spreadsheet here for the last 2 days - there's a bit of a problem that I know Datapunter is looking into. In the meantime - if you want to send me a pm with your email address - I'll send over the spreadsheet that I used to make the tables above.

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Thanks Swooperman, No lack of enthusiasm here - just taking a little time out...looking at where this is going. Hope not to put anyone off, but there about another 7 chapters to come :o (will post a list soon...) I've been looking through some old notes - remembering stuff I'd forgotten (worrying...) so I'm probably getting as much out of this as anybody! Anyway - glad there's some enjoyment there - I like this stuff - I've been messing around on-and-off with this for over 20 years now - as some strange bloke once said, 'It's what I do!':D

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Here's the spreadsheet I used to do the matrix calculations for the tables above - Thanks Datapunter :) for help to attach file - Please let me know here if it doesn't load / work!! I hope it's easy to use and understand with the notes above. You can change the metric by clicking on the page marked 'g', but you have to have your own data to put in. (Excel will only invert a 256 x 256 matrix, so you can only use a max of 256 matches with this) It might be good to experiment with anyway - knocking it down to just 4 or 5 teams is an excellent way to get a good feel for the kinds of ratings it comes up with. Perhaps any 'form fanatics' might like to produce a LS solution with the same metric and post up the table, by changing the sizes of the matrices, for the last 60 games say... This method has it's advantages and disadvantages - I actually use another method of solution for my own work now which is coming up next...but I've also done a lot or work with this and can recommend having a go. Anyway - it's just a bit of fun - only generating ratings at the moment...

conv_10.xls

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas one question: for the goal expectation home vs away, you included in a value to accommodate the home advantage. surely this has to be done in both the HT goals and also shots on target metrics? (simply the difference of 'average' home and away values?)

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Hi Muppet, You've an eagle eye there!! Actually, when I put up the table for 'half-time' ratings, I did subtract 0.2 from the score difference to compensate for home advantage (it's about 0.4 for full time, so thought that would be reasonable for now...). I didn't bother doing anything with the shots on target one - though if you wanted to pursue that I'd say you were right, there is perhaps a home advantage factor there too that you could put in. btw - A while back I looked at large samples of data, and found the correlations between margin of victory, and shots on target, shots, corners etc. It's not hard to do so could easily post up the results again shortly - but seem to remember that they had a bearing on the match result in the order I just posted. Corners had almost no correlation at all!! ie - If you just knew who had the most corners in the match you would have a hard time predicting who the winner was. The best correlation with margin of victory, was margin of victory:lol , which is why I use that as the basic metric... I've only ever used this type of LS solution to try and predict final scores - in fact I've been looking at a completely different approach for trying to deal with 'in-running' style predictions. But I think you would be right - if not looking at final scores, it would be best to alter the metric in someway depending on what you're looking at.

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Thanks for all the PM's!! Glad you are finding this interesting:) I'm going to put one I had here as it concerns using the spreadsheet given above...

hi there mr O. what a great worksheet. thanks alot for sharing it. is it possible to extend the sheet so that it copes with the maximum 256 games (which allows you to miss the first 124 games or 31% of the season- which may be a good thing - it equates to a team's first 12 games) ? at the mo it is difficult with the arrays to extend them - i am trying to look at a full season from 03/04 if you don't mind - to see what it looks like. this is well interesting!
Hi Muppet, You should be able to extend to 256 games very easily - just re-enter the matrix calculations. Excel is not very user friendly for matrix work - you have to know the order (size) of the solution matrix you are going to get, then paste it in all the cells. Read entering 'array formulas' help if not sure. It;s the 'transpose matrix' function that limits it's us to 256 games here - as the spreadsheet is 256 columns wide - it doesn't like it after that!! There are some quite convoluted ways of dealing with that - but better is to use the simplified LS solution which I'll show next in the thread. I'm going to put this answer on thread - hope you don't mind - enjoy the spreadsheet anyway!! Jon
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas I was looking for a quote on 'planning' to throw in here, lighten up the thread a bit. Unfortunately this one's a bit :eek !?

Men often oppose a thing merely because they have had no agency in planning it, or because it may have been planned by those whom they dislike. Alexander Hamilton Anyway, hope this doesn't attract too much opposition - as here's how I see the rest of the thread at the moment...I suppose if you don't like it, Mr Hamilton would suggest you get involved in the planning and tell me! Intro - Ratings Lists, Metrics 1. Generating a rating list - Least Squares solution 2. Generating a rating list - A simplified LS solution + another solution that approximates a LS solution method. 3. Creating data for backtesting 4. Using a rating list as a prediction model 5. Assessing the Model - How good is it? 6. Improving the Model - Fine tuning, combining different metrics, starting again with a new metric (that's what I normally do!!) 7. Finally, the model versus the Bookie! Not out of chapter 1 yet though - still a few points to discuss...
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas Planning, i think you're in for a long run here Mr Onemore :) People tend to follow threads like this in their own time, you'll only get a few following as you go along and a lot will come afterwards. Speaking for myself, i'll need some time to 'digest' all the info. And i'll probably have a good look in a few weeks when i have a week off work. So just keep things coming...

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas looking forward to Chapter 2 - but if it works and is simpler than chapter 1 ,why Chapter 1 ? Sorry for asking a silly question , but it the way my brain works. What about Chapter 7 now then explain why - i.e. Chapter 1 to 6 Cheers

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

Not sure that is a good thing, but we'll soon see, my ratings system is much simpler.
I think the matrix method puts it on a bit more of a mathematical footing and makes it more versatile and open to modifications further down the line. It's not complicated if you know a bit of basic matrix maths, you should find plenty of easy stuff on the internet.
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas To be honest Paul, I love arithmetic and am very fast with calculations, but hate maths beyond that, it just bores me silly. :( My point was more that I don't think a ratings system should be complex, I think it's asking for trouble and tweaking galore. However, I'm always pleased to be proved wrong as it means I've learned something new. ;)

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas i don't understand why transposing more than 256 is a problem, surely by rearranging the data 90 degrees the 256 columns will be rows - obviously i' grossly misunderstanding what's going on really, but can someone explain?

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

i don't understand why transposing more than 256 is a problem' date=' surely by rearranging the data 90 degrees the 256 columns will be rows - obviously i' grossly misunderstanding what's going on really, but can someone explain?[/quote'] There's no problem storing the data like that on the spreadsheet. However you won't be able to use the matrix multiplication function to work out 'A transpose' times 'A'. As I mentioned to you in the PM, there are some quite convoluted ways of getting around this - I spent a long time working with these myself - you could write your own function or a macro for starters - but I'd recommend holding fire until you read the next part of this thread...
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

looking forward to Chapter 2 - but if it works and is simpler than chapter 1 ,why Chapter 1 ? Sorry for asking a silly question , but it the way my brain works. What about Chapter 7 now then explain why - i.e. Chapter 1 to 6 Cheers
Brookes - couldn't have put it better than PaulM03. Just by the way, perhaps you should read post #1 in thread again - you'll not find a successful system here - but you will find a method to develop your own ideas and test them in a systematic way. To perhaps save you, and others, their time - the last sentences by me in this thread will be something like this; So, as I stated right at the beginning, this metric is not quite good enough to sustain a long term profit. However, another metric may well prove to pass the tests above - good luck - hope you find it!!
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

To be honest Paul, I love arithmetic and am very fast with calculations, but hate maths beyond that, it just bores me silly. :( My point was more that I don't think a ratings system should be complex, I think it's asking for trouble and tweaking galore. However, I'm always pleased to be proved wrong as it means I've learned something new. ;)
Paul - you've made a couple of these comments so far. As I previously replied - the rating list I use as an example here is simplicity itself. The metric I use here is far simpler than the performance ratings you provide. The LS solution (and moreover - it's simplified versions) is a known Mathematical way of dealing with the data produced. I think it's a very elegant solution too. Also, 'tweaking' a system is something I just don't do - countless times on this thread I have advocated looking at longer term trends to avoid what I term 'fine tuning'. I abandon poor metrics, and find new ones rather than embark on pointless efforts of tweaking a bad system. If we both agree that a system should be simple, then there is nothing I will present here to 'prove you wrong' on that point.
Blimey Mr O, has to eb the most complex ratings system I've ever seen! :eek Not sure that is a good thing, but we'll soon see, my ratings system is much simpler.
As to these comments - I have to ask which part of
Intro - Ratings Lists, Metrics 1. Generating a rating list - Least Squares solution 2. Generating a rating list - A simplified LS solution + another solution that approximates a LS solution method. 3. Creating data for backtesting 4. Using a rating list as a prediction model 5. Assessing the Model - How good is it? 6. Improving the Model - Fine tuning, combining different metrics, starting again with a new metric (that's what I normally do!!) 7. Finally, the model versus the Bookie!
do you think I should skip to make it simpler? I know that I won't commit money on a 'system' unless I've satisified myself on all the above. I think it's a bit reckless to provide a ratings list/system without testing it beforehand to see if it works. I think it's a standard procedure that lots of readers here will agree with (It's what I do anyway!!) Still, I'd be interested if you'd like to expound your views if you feel different - I'm also looking to learn from the thread.
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas This thread is great Mr Onemore- looking forward to the next chapters. Would you just please explain why you needed to pre-multiply by the tranpose matrix of A during the matrix calculations rather than just pre-multiply by the inverse Matrix of A to give you . r = inv A . g I'm guessing it is easier to calculate the inv. of A.At? Seeing as I didn't even know what a transpose matrix was though before I just looked at Google I don't know why? Keep up the good work with this thread!

Link to comment
Share on other sites

Re: A football rating system, discussions, ideas

This thread is great Mr Onemore- looking forward to the next chapters. Would you just please explain why you needed to pre-multiply by the tranpose matrix of A during the matrix calculations rather than just pre-multiply by the inverse Matrix of A to give you . r = inv A . g I'm guessing it is easier to calculate the inv. of A.At? Seeing as I didn't even know what a transpose matrix was though before I just looked at Google I don't know why? Keep up the good work with this thread!
Hi bcrazy Fair idea - in essence it's right. But the short answer as to why that's not possible is that A is no longer a square matrix after we put the last teams rating equal to zero - the last column is 'deleted'. (You can only find the inverse of a square matrix). In effect, we have too few variables and too many equations - over determined. (If you remember the kind of simultaneous equations from school - in that case there were 2 equations and 2 unknowns, and that can be solved exactly. But over determined systems generally have no exact solutions, that's why we choose something like a LS solution to approximate) btw: At.A gives a square matrix that can be inverted. As to why is the LS solution of A.r=g is inv(At.A).At.g ?? Well, that's a 'text-book' proof which I've no intention of delving into here - horrible maths!! - but I'm sure there must be info on the internet. Glad you're enjoying the thread:)
Link to comment
Share on other sites

Re: A football rating system, discussions, ideas I'll be honest with you Mr O, I was out with Osesame tonight and we talked about you and this thread. I admitted that it's a little over my head simply because I don't like maths to this extent, which doesn't embarrass me to admit that, so I would say just carry on as Osesame understands and thinks you are doing a good job. :)

Link to comment
Share on other sites

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.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...