If you want to balance the game...

[arnaudus]

Hello all,

As I see your brains smoking because you are trying to balance the live games, here are some useful informations.

I tried to summary the results of real leagues. Respectively : French 2nd league 2004, England 1st league 2004, Italian 2nd league 2004 and German 2nd league 2005.

Average points/game : 1.34 1.35 1.30 1.37 (average: 1.34)
Average goals/game : 1.13 1.33 1.23 1.36 (average: 1.27)
Stand dev points/game 0.30 0.40 0.27 0.33 (average: 0.33)

First team points/game: 1.92 2.36 1.87 1.96 (average: 2.02)
Last team points/game: 0.76 0.86 0.72 0.96 (average: 0.82)

If bygfoot's leagues results fit these data, I think the realism of the game will be quite good.

Good night!

Arnaud.

[vector]

yes its amazing how close it all is.. who would have thought last place Norwich would beat the Man UTD.. for eg.

[gyboth]

thanks for your work, arnaud. we're going to do the balancing after the feature freeze, though.
two remarks:
1. why did you take mostly 2nd leagues?
2. i'd also need the average goal difference. i've got the impression that currently there are too many matches with a team not scoring (1-0, 2-0).

gyözö

[arnaudus]

1. why did you take mostly 2nd leagues?
gyözö

Just because I did not found the results of the 1rst leagues But anyway, I was surprized by the fact that the results are almost identical, whatever the league and whatever the country.

2. i'd also need the average goal difference. i've got the impression that currently there are too many matches with a team not scoring (1-0, 2-0).
gyözö

For information, the average goal difference over the whole season is... 0 The better way might be to compute the frequencies of 0-0, 1-0, 0-1... but I haven't these data. A good workaround for the moment may be to suppose that the number of goals scored by a team follows a Poisson distribution of mean ~ 1,3, but I am realizing that it is certainely usefull to have this average number of goals at home and away.

Anyway, I hope this will help us to make bygfoot more realistic.

[gyboth]

For information, the average goal difference over the whole season is... 0

wow, you're a quick calculator of course i meant the absolute value of the goal difference, where e.g. 1 - 2 and 2 - 1 are the same.

gyözö

[cavehippo]

If it helps, the issue of statistical analysis of football games has already been the subject of some formal research.

http://arxiv.org/abs/cond-mat/0110605
gives a lot of very complex analysis of the problem, whereas

http://arxiv.org/abs/cond-mat/0110605
is less detailed but maybe also easier to understand

and
http://arxiv.org/abs/cond-mat/0409609
models the goal scoring records of individual players, as well as their movements between clubs.

[gyboth]

http://arxiv.org/abs/cond-mat/0110605
gives a lot of very complex analysis of the problem, whereas

a nice paper, thanks for the hint.

http://arxiv.org/abs/cond-mat/0110605
is less detailed but maybe also easier to understand

i'm not sure which one of the two i skipped over, the easy one or the hard one (i didn't look at the second link since it seems to be the same as the first link ). it wasn't too hard, anyway.

http://arxiv.org/abs/cond-mat/0409609
models the goal scoring records of individual players, as well as their movements between clubs.

well, this might be something for a later stage of Bygfoot.

i'd say the approach with the average goals is easier to follow than the article above, since in the article the authors try to find a distribution that fits the data (the number of goals in thousands of matches) best. but for quick and easy testing of the results the game delivers, it's simpler to calculate the average goals than some poisson or frechet distribution parameter.

nevertheless, thanks for your ideas, maybe you'd like to try your hand with the fine-tuning screws in the bygfoot_constants file to help balancing

gyözö

[Guest]

(i didn't look at the second link since it seems to be the same as the first link ). it wasn't too hard, anyway.

Sorry, yes, I mis-pasted, this is the second paper....

http://arxiv.org/abs/cond-mat/0002100

[gyboth]

here's a Bygfoot 1.9 season statistics for the english leagues.

"Goals/G HomeG/G AwayG/G Away % GD/G" mean goals, home goals, away goals per game, away goals in % of all goals, and goal difference per game.

"home win" and draw and loss should be clear.

Code: Select all

Premiership
Goals/G HomeG/G AwayG/G Away % GD/G 
2,41     1,47    0,93    0,39    1,38 
Home win %: 0,49 Loss %: 0,25 Draw %: 0,26 
------------------------------------ 
Football League Championship
Goals/G HomeG/G AwayG/G Away % GD/G 
2,33     1,41    0,93    0,40    1,32 
Home win %: 0,49 Loss %: 0,27 Draw %: 0,24 
------------------------------------ 
Football League One
Goals/G HomeG/G AwayG/G Away % GD/G 
2,46     1,39    1,06    0,43    1,26 
Home win %: 0,45 Loss %: 0,30 Draw %: 0,25 
------------------------------------ 
Football League Two
Goals/G HomeG/G AwayG/G Away % GD/G 
2,63     1,61    1,02    0,39    1,30 
Home win %: 0,49 Loss %: 0,23 Draw %: 0,28 
------------------------------------ 
Nationwide Conference
Goals/G HomeG/G AwayG/G Away % GD/G 
2,21     1,36    0,85    0,39    1,29 
Home win %: 0,47 Loss %: 0,25 Draw %: 0,28 

and here's attached a table i found on the web about goals and results in different leagues. the translation is a bit weird but should be understandable nonetheless.

the Bygfoot stats are fairly realistic, i'd say. and i didn't have to do a lot of tuning, either, i only reduced the home advantage a bit.

the only thing i didn't find yet (on the web) is average goal difference. but since that's some more or less deterministic function of goals/game and home win % (i think), it's not that grave i didn't find it.

gyözö

[vector]

excellent!!