Team ratings - spotting a peak

[YorkshireFootyFan]

[YorkshireFootyFan]

Re: Team ratings - spotting a peak I ran a test with one season's matches. For League 1 2009/2010, I calculated the average rating for each team home and away and then substituted them for the game by game figures - the end profit improved dramatically, so the initial signs are good and there is some mileage in this. The above was to be expected, the average stops you giving too much weight to a team at it's peak, and too little weight to a team at the bottom of a trough. My next step again with this one season's data was to try and calculate the average as the season progressed so for instance in the data above for the second game Brentford's rating would be (40+52)/2=46. However when done for the full season the improvement over the original method was only small. Using the median rating for all teams again didn't have much effect on the original results. I'll test the end of season average rating for a few more leagues when i get chance, but if that provides the same degree of improvement for all seasons tested I will have to have a rethink. If that is the case, I will then look to assess how much teams change from season to season, to see if there is any benefit in calculating my ratings over a longer period. This will then introduce the problem of assessing teams who have been playing in a higher or lower league. Again if anyone has any ideas please contribute them, my methods above are rudimentary to say the least, any input would be appreciated.