Pl is some Joke. Five games to go. West ham Fourth from Bottom. Awful season and 15 points clear of relegation. Hopefully 24 or 25 would be safe next season. We might manage that
A mate of mine has come up with hospitality tickets for the last home game. Two course meal and then over to east stand to those seats in the middle. Apparently he bought them pre season thinking this would be the promotion decider. The sly bugger only told me yesterday i would have not kept quiet that long. Really good price too so could not turn it down. Jinxy is back
I looked into this a while back and it made my head hurt. I had thought they would simply give win draw loss points based on xg. Ie if the xg is close they call it a draw if it’s by a clear enough margin they would call a win…. So 3,0,1 points for each team. What I believe actually happens though is they (usually) run (thousands of) simulations to work out the probabilities of different result happening (win lose draw for each game based on the xg’s)… and then dish out the points according to the likelihood/probability of each result happening. If anyone other than Aski reads this and is still awake, xp points, like xg, is a form of regression analysis. Everyone gets that the sides that have the best chances don’t always win - xp works out how often they ‘should’ win. Expected points awarded will be between 0 and 3. So for example if the simulations from the xgs for a particular game show Leeds would win the match 60% of the time, draw it 30% of the time and lose 10% of the time… they then do .6x3 (pts) + .3x1(pt) + .1x0 pts which equals 2.1 points. It would be reversed for the opposition. As a point of note according to expected points we should have topped the league and therefore been promoted last year. And we should not have been relegated the year before. And people wonder why I’ve been saying we need to change keeper for 3 years
That makes sense. As yourself, I always considered that XPoints would be based on XG v XGA, whereby if one team had an XG of 0.6 and their opponents had an XG of 1.4, then the expected result would be a draw, based on roundings, where as an XG of 0.6 Vs 1.5 would result in an away win. However I could never make it work, no matter which of the stat sites I used, nor could I see a method which could be easily replicated to test. Just did a search on expected points calculations and see that the Google AI response matches your explanation. Also found an article regarding a python script that i will have a look at, as that seems more based on probability than simulating and thus from my perspective, would be simpler to test the accuracy of. Appreciate the response Milky, as it gave me the incentive to look into it deeper, instead of pushing it to one side like I usually do
was this the article? If not it might help you if you want to create your own model. https://medium.com/@nashrougvie/exp...nd-how-to-program-your-own-model-d229ef452aa5 Your commitment and curiosity on these matters is commendable fella- I’m happy to just look up tables done by others I think the python script would be the holy grail for you
Think I looked about 5 or 6 sites after reading your response, a couple of which were on medium.com, but will check that one out too. https://mckayjohns.substack.com/p/how-to-calculate-expected-points Is the one I looked at, it's more related to programming, but also links to https://fivethirtyeight.com/methodology/how-our-club-soccer-predictions-work/ which is more about the actual methodology Been on the internet for so long I never trust anything I read on there, and my natural curiosity for numbers is enough to keep me out off trouble
Best of luck with it mate. I’m Well aware of the limitations of xg… it doesn’t account for quality of strikers and keepers, it only measures shots, so dangerous crosses etc aren’t accounted for. But as long as you accept what it does and doesnt do - it does give a reasonable indication of underlying team performance. There’s a reason why betting companies use it. It’s why they always had us as clear favourites to go up. From my own experience I’ve used xg tables to spot underperforming sides. For example Coventry struggled under robins this season and he was sacked and replaced by lampard. But they’ve been well placed on xg all season - they just were making mistakes and not getting the breaks. They’d have very likeky risen whether they sacked their manager or not.
