For doubles ladders, the 'pairs performance' tab shows an analysis of games played organised into player pairings. The goal is to enable club captains (etc.) to work out which players are most effective together and so get the best from everyone on the team.
The "synergy score" itself is calculated using a statistical model. This estimates the performance of each pairing, whilst adjusting for the quality of their opponents.
Behind the scenes, we use a 'random effects model' implemented using the
python statsmodels package. The advantage of this type of model is that it accounts for a number of potential pitfals when looking at the performance of pairings in leagues, for example that: Certain pairings may tend to play specific other pairs. For example, when paired with a 'good' player I might tend to play other good pairings. A win or loss to these high quality opponents should not be interpreted the same as a win or loss to lower quality opponents. Also, players may play different numbers of games in specific pairings, and this needs to be taken into account before judging a pairing as good or bad (we can't be confident until a certain number of games have been played).
The analysis is not foolproof, but it should tend to identify which combinations of players will be most effective.