The piechart below shows a breakdown of the success rate in predicting
game results, from the perspective of the home team. The fraction of games
in which the home team won as predicted is shown in green. The fraction
of home games that were lost as predicted is shown in light brown.
The fraction of home wins
that had been predicted as losses is shown in dark green, and the fraction
of home losses that had been predicted to be wins is shown in dark brown.
The plot below shows the success rate for predicting game winners
for games each week during the season. Error bars are calculated
based on the number of games played that week. When few games are played,
it takes only a few games going against the statistical expectation
for the success rate to change by a relatively large amount. The fewest
games played were the first week of the season.
The plot below shows the success rate for predicting game winners
depending on how close the winning margin that was predicted by the
strength ratings. The success rate for games that were predicted
as wins by 0-0.5 goals is shown at the far left, followed by games where the
winning margin was predicted to be 1, 2, 3, 4, ... goals, etc. We rarely
miss predicting the winner when the winning margin is
expected to be 3 goals or more.