An approach to identify which factors most closely correlate to victory. A few possible approaches (taken from conversations on bgg)
a)
What I'm thinking about is a matrix, which basically quadruples all histograms and gives a separate histogram per place achieved. E.g. network place for #1, network place for #2, ...
The data points per histogram will roughly get quartered, so significance will be lower, but the number of games is large enough. Only with histograms per place achieved you can make meaningful statements about success factors. E.g. (made up example) if the #1 Swarmlings almost all picked FAV10 while #3/#4 Swarmlings only picked it in 40% of their games, you can conclude "to win as Swarmlings, you better have FAV10", and if the distribution of pick order across places achieved is pretty similar, you can conclude "the seat from which you pick Swarmlings is not much of a factor for winning".
b)
A histogram showing win % based on scoring in a given range (e.g. Scoring 12-15 cult points as giants leads to winning 4% of games).
That might be useful. You'd need the reference value too (in your example: "On average, Giants win 16% of games in which they participate")
c)
Identifying the number of points needed in a category for a 75%, and 50% win rate. (e.g. "when darklings score at least 27 cult points, they are likely to win (75% change); they have better than a 50% change to win when they score 20 points in cults)".
That could be interesting as well. Also when it comes to available bonus tiles and round scoring.
An approach to identify which factors most closely correlate to victory. A few possible approaches (taken from conversations on bgg)
a)
The data points per histogram will roughly get quartered, so significance will be lower, but the number of games is large enough. Only with histograms per place achieved you can make meaningful statements about success factors. E.g. (made up example) if the #1 Swarmlings almost all picked FAV10 while #3/#4 Swarmlings only picked it in 40% of their games, you can conclude "to win as Swarmlings, you better have FAV10", and if the distribution of pick order across places achieved is pretty similar, you can conclude "the seat from which you pick Swarmlings is not much of a factor for winning".
b)
c)