Some statistics from the way to 5th Nova:
Total d20s rolled: 32510
Average roll: 10,474213
Total Natural 1s: 1638
Total Natural 20s: 1577
Combat:
Natural 1s in combat: 495
Natural 20s in combat: 529
Initiative rolls: 2288
Ranged attacks and melee attacks are closer than expected (5129 vs 5222) [no information on spells as they have too much variability].
1617 Reflex saves, 756 Fortitude saves, 696 Will saves
Skills:
Acrobatics: 567; Athletics: 725; Bluff: 343; Computers: 1059; Culture: 951; Diplomacy: 782; Disguise: 35; Engineering: 1588; Intimidation: 229; Life Science: 498; Medicine: 431; Mysticism: 617, Perception: 2714; Physical Science: 266; Piloting: 894; Professions: 463; Sense Motive: 564; Sleight of Hand: 65; Stealth: 461; Survival: 573
Other:
Games played as a player at 1st level or with pregens: A few. Games played at 2nd level or higher with characters that were not pregens: Zero.
Manually generated chronicles: Around 768. Chronicles made with RPG Chronicles: Around 54. Joy about not having to fill out new chronicles manually: 100 %.
Some notes:
Almost every scenario from the four seasons should be included at least once, some several times. This includes the Skitter one-shots and One-Shot #1. Not included are the newly released 4-09 and 4-10, as well as Starfinder: The Starfinder Four VS. The Hardlight Harlequin and One-Shot #2 for obvious reasons. Due to AP credits, the total amount of games is probably a bit under 150 games.
The vast majority of the rolls should be rolls from the PCs, though NPC rolls will also be included (no way to really filter the rolls without specifying each and every PC name). This should mostly affect attack rolls and saves, NPC skill checks do not come up very often. NPC initiatives are usually handled automatically and will therefore not affect the amount of initiative rolls too much.
Note that this will not be 100 % correct, so treat it with some caution. Some rolls are from test games, there are also test rolls made by players, random rolls, rerolls, … A few rolls have been lost due to crashes or parsing errors. It’s also entirely possible I accidently attributed some rolls incorrectly, I just quickly adapted some older code. Trick attack rolls can skew certain skills. In some cases, players manually rolled d20s, making them not feasible to categorize.
Also bear in mind that for the rarer checks there is a self-fulfilling cycle: Players are less likely to take them, which also causes them to be rolled less. E.g. if a check allows Disguise or Computers, the Disguise check is unlikely to actually be rolled, even if it actually would be a RAW option.