That stats predictor site posted yesterday

Just had a chance to have a quick look at:

"The weighted method takes the opponents record and home field advantage into account when randomly picking scores, so the better team is more likely to win."

Be interesting to know how the weightings work. And whether the "opponents record" has been weighted according to credibility and relevance (the more recent past results being more relevant than the older past results). Perhaps some kind of blended Bayesian method could be used to improve the model, whereby an a-priori distribution (or estimated number of total points accumulated say) is blended with the past data (or "opponents record") weighted by how many games have been played to date? Also, the fact that teams coming into form now or are "on the beach" and have nothing to play for etc isn't taken into account somehow is a bit of a weakness.

The bit about home advantage is ok I suppose, unless it's known that certain teams tend to play better either at H or A and that could perhaps also be factored in somehow.

Overall, it's an interesting attempt at trying to stochastically model the league table though. Good effort. Does anyone have the code/exact algorithm he used though? Probably make more sense if that's accessible somehow.

Posted By: essexcanaryOTBC, Mar 22, 01:57:04