Ram8 Posted January 21, 2015 Share Posted January 21, 2015 And Bournemouth have just LOST AT LEEDS .. !! Link to comment Share on other sites More sharing options...
Albert Posted January 21, 2015 Author Share Posted January 21, 2015 Silly question - does it take into consideration where teams still have to play each other? Both teams can't winIt can be made to, makes effectively no difference until the very last games of the season, i.e. <5, even then, not a huge deal of difference.LOVE Albert's stats pages... However this doesn't take into account any improvements or dips in form moving forwards... An example like Birmingham coming back from looking in trouble this season, Palace in the Prem under Pulis last season and mid season drop offs like we did in the promotion season under the angry scottish sausage sausage sausage sausage sausage sausage sausage sausage Sorry got a bit stuck there... Anyway... It doesn't take into account any overall changes in form... Not small runs, but complete changes...The entire point of it is to present things on a statistical basis to give context to the league table. Maybe I didn't explain it particularly well, so I'll try again. Give me a bit.I don't mean to be rude because obviously you have put a lot of though and effort into this but, have you not spent hours basically coming up with a complicated way of presenting the league table? If everyone has played the same amount of matches and everyone has played each other once then surely the higher up the league table, the higher the probability of finishing higher up the league?!Basically, but what it does is quantifies the probabilities in a way that a league table alone can't do.I didn't do this recently by the way. This is a code I wrote last January testing out features in a programming language I needed to get my head around for my work. I actually posted quite a bit using it back then as well. I basically just put the current league table through it and wrote it up, was actually quite a quick one.Really interesting to read, that. Of course this is a very simplified model with lots of assumptions, like not taking into account possible injuries, dates that games will be played (clustered fixtures etc.), potential signings, effects of managerial changes, teams with more/less to play for nearer the end of the season, how the weather on the day may suit one team over another, etc. etc. So very much take the results with a pinch of salt, that's model analysis 101. But a very nice visualisation and interesting to think about.This one is actually independent of model. It's essentially (as others have also noted) a different way of presenting a league table. It gives context to the shape and form of the table, to the chances of teams and essentially a way of allowing context to rule over outright numbers.In that way it's not "results" anymore than the league table is really. The part of note is that simply allows context to the league table to go with our position, rather than "we're top today".However, as mentioned it can be made to incorporate which teams are played and such, but ultimately it has little impact unless there are a very small number left. The reason being that over time such things even out. Link to comment Share on other sites More sharing options...
Angry Ram Posted January 21, 2015 Share Posted January 21, 2015 I don't mean to be rude because obviously you have put a lot of though and effort into this but, have you not spent hours basically coming up with a complicated way of presenting the league table? If everyone has played the same amount of matches and everyone has played each other once then surely the higher up the league table, the higher the probability of finishing higher up the league?! That's stats for you.. Basically a long winded way, of telling you what is right in front of your face. Link to comment Share on other sites More sharing options...
Angry Ram Posted January 21, 2015 Share Posted January 21, 2015 Alberts a fun guy, that's I got from this thread. PPG nothing to do with tea then. Link to comment Share on other sites More sharing options...
davenportram Posted January 21, 2015 Share Posted January 21, 2015 Its almost all back in our hands 3 points off top with a game in hand on 1st and second. And we have to play them both again. Link to comment Share on other sites More sharing options...
eddie Posted January 21, 2015 Share Posted January 21, 2015 And Bournemouth have just LOST AT LEEDS .. !! So that's Leeds and Forest nailed on for automatic promotion now then. Link to comment Share on other sites More sharing options...
Albert Posted January 21, 2015 Author Share Posted January 21, 2015 There seems to be some confusion over exactly what was done, and what it's meant to represent, so I'll try and reword the explanation.This is not predictive, it is a method to quantify the chances teams have of finishing in certain areas of the table. Essentially points per game is taken from the league table, used to estimate a probability curve of a team getting different points per game over the rest of the season. These probability distributions are then compared to determine the chance of a team finishing above a number of teams, hence determining the chance of a team finishing in each place. It's essentially a fancy way of showing the league table.What do I mean by this though. I while back I gathered years worth of data on the Championship, and how teams performed from various points in the season, and from that attempted to figure out how the league table evolved. The key thing of note was that when comparing how teams performed throughout the rest of the season, to how their performed in the earlier part of the season, it roughly fit a standard distribution. This allowed calibration of the chances of a team to perform to a certain level in terms of points per game, based on their prior performance. That is, the most obvious of statement, teams that have performed well are more likely to perform well during the rest of the season.Now, some people seem to wonder "but what out this extraordinary circumstance". The thing is, that's entirely taken into account. When I say "Cardiff have a 1% chance of making the playoffs", that is literally what it says. For each team in Cardiff's current position, 1 in 100 will make the playoffs. If you have say, 5 teams around those odds each season, then you'd expect that one would make it once every 20 years. If you have 10 at that level, then one every 10 years, etc.This is the thing though, the model is independent of changes to teams. That is, teams becoming better, or becoming worse due to changes in playing staff, injuries and such is taken into account in the original calibration of the standard curve (which generates the probability). That is, it's not trying to cover that situation, as that within the model considered an unpredictable factor, and not something it's trying to do. We can sit around all day and discuss how signing John Somebody is going to change the way we move the ball between midfield and attack, and how that'll win us the league, but in reality he could equally disrupt the team and lead us to dropping off and going through the playoffs (Hi Billy).So basically, what it's saying is that there is a certain chance of ending up in particular positions based on current performance, and yes, the obvious result should and does reign (again, because it's not trying to be predictive, rather a different way of presenting the information). You expect the best performing teams should be the most likely to end up top, and of course that is the result. The point though is that it quantifies not only that, but also the chances of teams well out of the current reckoning.So to basically summarise, this is not "if we perform as we have to now, we'll end up where we are", it's "based on our and the rest of the league's performance, these are the odds or various positions". It takes into account the chances that teams can greatly improve, or drop off, hence why Cardiff, Forest and Bolton appear at around 1% on the playoff and relegation candidate lists. It's not about prediction, but rather giving a sense of the chances we're talking, as well as offering a tool to determine how much damage a poor result, or how much good a positive result is having on our position. Link to comment Share on other sites More sharing options...
Brammie Steve Posted January 21, 2015 Share Posted January 21, 2015 Ceteris paribus Albert! Ceteris paribus chum! All other things being equal. IF no weird happenings come! The bookies'd scream when they saw you, The football pools run out of dough. You'd pocket your weight in gold sovereigns, But life's never like that you know. Statistics are great in their place Al, They prove what you want them to prove. But nothing in life's ever equal, Variety's spices still move. So given the Rams' motivation, We'll go up this season for sure, We should've gone up in the last one But stats don't account for Zamora! Link to comment Share on other sites More sharing options...
Angry Ram Posted January 21, 2015 Share Posted January 21, 2015 There seems to be some confusion over exactly what was done, and what it's meant to represent, so I'll try and reword the explanation.This is not predictive, it is a method to quantify the chances teams have of finishing in certain areas of the table. Essentially points per game is taken from the league table, used to estimate a probability curve of a team getting different points per game over the rest of the season. These probability distributions are then compared to determine the chance of a team finishing above a number of teams, hence determining the chance of a team finishing in each place. It's essentially a fancy way of showing the league table.What do I mean by this though. I while back I gathered years worth of data on the Championship, and how teams performed from various points in the season, and from that attempted to figure out how the league table evolved. The key thing of note was that when comparing how teams performed throughout the rest of the season, to how their performed in the earlier part of the season, it roughly fit a standard distribution. This allowed calibration of the chances of a team to perform to a certain level in terms of points per game, based on their prior performance. That is, the most obvious of statement, teams that have performed well are more likely to perform well during the rest of the season.Now, some people seem to wonder "but what out this extraordinary circumstance". The thing is, that's entirely taken into account. When I say "Cardiff have a 1% chance of making the playoffs", that is literally what it says. For each team in Cardiff's current position, 1 in 100 will make the playoffs. If you have say, 5 teams around those odds each season, then you'd expect that one would make it once every 20 years. If you have 10 at that level, then one every 10 years, etc.This is the thing though, the model is independent of changes to teams. That is, teams becoming better, or becoming worse due to changes in playing staff, injuries and such is taken into account in the original calibration of the standard curve (which generates the probability). That is, it's not trying to cover that situation, as that within the model considered an unpredictable factor, and not something it's trying to do. We can sit around all day and discuss how signing John Somebody is going to change the way we move the ball between midfield and attack, and how that'll win us the league, but in reality he could equally disrupt the team and lead us to dropping off and going through the playoffs (Hi Billy).So basically, what it's saying is that there is a certain chance of ending up in particular positions based on current performance, and yes, the obvious result should and does reign (again, because it's not trying to be predictive, rather a different way of presenting the information). You expect the best performing teams should be the most likely to end up top, and of course that is the result. The point though is that it quantifies not only that, but also the chances of teams well out of the current reckoning.So to basically summarise, this is not "if we perform as we have to now, we'll end up where we are", it's "based on our and the rest of the league's performance, these are the odds or various positions". It takes into account the chances that teams can greatly improve, or drop off, hence why Cardiff, Forest and Bolton appear at around 1% on the playoff and relegation candidate lists. It's not about prediction, but rather giving a sense of the chances we're talking, as well as offering a tool to determine how much damage a poor result, or how much good a positive result is having on our position. So to sumarise, its all bollx? Link to comment Share on other sites More sharing options...
Whoneedshuls Posted January 21, 2015 Share Posted January 21, 2015 Its almost all back in our hands 3 points off top with a game in hand on 1st and second. And we have to play them both again. Puts into perspective how crucial that away game at Bournemouth will be Link to comment Share on other sites More sharing options...
davenportram Posted January 21, 2015 Share Posted January 21, 2015 Albert. In your original post you said that without the forest result we would have a 43% probability of going up. Just curious. Did you replace the loss with a win or just blank it and use one less game to work out PPG so far? Link to comment Share on other sites More sharing options...
i-Ram Posted January 21, 2015 Share Posted January 21, 2015 For ****'s sake Albert, you post all this fantastic ******** just to end with a summary of:- "Ultimately though we're doing well in a tight league which has shown anything can happen". At times like this I refer to Confucius. One of his more popular quotes is "life is really simple but we insist on making it complicated". I like one of his lesser know quotes though which is "if you want to get Repetitive Strain Injury much better to masturbate frantically than tap away tediously on a keyboard". Link to comment Share on other sites More sharing options...
RamNut Posted January 21, 2015 Share Posted January 21, 2015 If we forward project the mean values of the total results to date.............yes yes And factor in a variability in the base data......................yes Whilst assuming an underlying validity in the strategic projection............come on Then............what? what? Well..... Anything could happen. Alberts theory of general relativity 2015 Link to comment Share on other sites More sharing options...
Angry Ram Posted January 21, 2015 Share Posted January 21, 2015 I'm lost for words. Wish Albert was... Link to comment Share on other sites More sharing options...
Ram a lamb a ding dong Posted January 21, 2015 Share Posted January 21, 2015 Wtf??!??!! Link to comment Share on other sites More sharing options...
Albert Posted January 22, 2015 Author Share Posted January 22, 2015 Albert. In your original post you said that without the forest result we would have a 43% probability of going up. Just curious. Did you replace the loss with a win or just blank it and use one less game to work out PPG so far? For those parts the results are blanked (for both our and Forest's end). It's essentially, "if we had the chance to have it again". As mentioned elsewhere, in a straight swap for a "beat Forest, lose to Ipswich" or "beat Ipswich, lose to Forest", we were much better off with "beat Ipswich". So to sumarise, its all bollx? Why do you say that? If we forward project the mean values of the total results to date.............yes yes And factor in a variability in the base data......................yes Whilst assuming an underlying validity in the strategic projection............come on Then............what? what? Well..... Anything could happen. Alberts theory of general relativity 2015 Again, it's not to predict, it's not a projection, it's to make sure that the chances are put in perspective properly. It's more a case of: - Use PPG to estimate "true mean" of team and variance in this (that is, PPG for the case of an infinitely long league season in the same conditions as now) - Use prior data (10 years used to calibrate, 5 to Test) to determine how "wide" the standard distribution of later season to early season performances is (this was done at various stages of the season as well) - From this it generates a probability curve of end of season points Here is an example of such a probability curve, with us and Forest: As you can see, there exists the possibility of us finishing below Forest if they perform extremely well and we perform extremely badly. We of course know that such a chance exists, the aim here is to quantify those chances. From here the probability of the team finishing above others can be found (which place found from beating a certain number, that is, reordering of the table is entirely taken into account). This is actually quite large calculation. Things like Sunderland or Reading's amazing runs to promotion are taken into account in the data used to calibrate. The chances are also quite clearly stated in the way the data is presented. That is, when you see this list: 1. Ipswich - 46% 2. Bournemouth - 39% 3. Middlesbrough - 33% 4. Derby - 33% 5. Brentford - 21% 6. Watford - 12% 7. Norwich - 8% 8. Wolves - 8% It doesn't mean that Ipswich and Bournemouth are going up, it means that over ten "chances" from this position you'd expect: 1. Ipswich to go up 5 times 2. Bournemouth to go up 4 times 3. Middlesbrough to go up 3 times 4. Us to go up 3 times 5. Brentford to go up twice 6. Watford to go up once 7. Norwich to go up once 8. Wolves to go up once It might even be worth changing the data presentation to being in the same form as odds. That is: 1. Ipswich - 9/4 2. Bournemouth - 5/2 3. Middlesbrough - 3/1 4. Derby - 3/1 5. Brentford - 5/1 6. Watford - 8/1 7. Norwich - 12/1 8. Wolves - 12/1 People going around going "Bournemouth are going to go up" or "our Top 2 chase is faltering" or even "what if we fall off and miss the playoffs", the point is that it offers insight into how likely or unlikely that is. It does take into account enough data that the level of potential variation from changes to the playing staff (injuries, signings, etc.) into account. Link to comment Share on other sites More sharing options...
Albert Posted January 22, 2015 Author Share Posted January 22, 2015 And Bournemouth have just LOST AT LEEDS .. !!Also note that the table used was after this match. It's in the probabilities. Link to comment Share on other sites More sharing options...
Angry Ram Posted January 22, 2015 Share Posted January 22, 2015 account. Link to comment Share on other sites More sharing options...
Albert Posted January 22, 2015 Author Share Posted January 22, 2015 See post #49Does math scare you? Link to comment Share on other sites More sharing options...
Angry Ram Posted January 22, 2015 Share Posted January 22, 2015 Does math scare you? Maths dont scare me, do them all day long. People try and serve stats up to me all day long, mostly contradictary and bollx. Even got two people once to do me a little project, knowing full well they had differing views, the results were...... Predictable. Same set of data and completely different results. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.