encratite / dota2analysis Goto Github PK

This might be a humorous stat to figure out. I suspect there's not much deviation from 50% there, though.

It would be useful to have a new category in which the numbers of ranged and melee heroes are compared.

Consider groups of two or three heroes (beyond those there probably won't be enough samples) to obtain win ratios for combined picks. Even three is probably beyond what can be currently achieved, though.

User plopyk demands that confidence intervals be added to the statistics. Can be done.

This is not easy because it would require pulling far more data from the Dotabuff servers. I really don't want to hammer them too hard because I appreciate the service.

The web scripts should provide numbers of samples available per composition or rather their relative amount. Once again the bar width could be normalised according to the greatest number of samples found.