FYI, I’ve set up the official event page.
I’ve also been running various simulated generations of the field to see what the real one might look like.
A logic mistake affects some of my estimates above: I added two numbers with different filtering criteria to get the figure of 10,700 games with fewer than 5 ratings. Using the most recent IFDB backup data, the correct value is around 11,500 games, i.e. a little over 75%.
I’m seeing 15,133 total games known. Of those, 12,324 are (according to license and genre data) non-commercial, non-adult, non-flagged games that meet the basic eligibility requirements. Of those, 9,007 have fewer than 5 ratings (i.e. a little less than 75%).
This leaves a total of 3,317 games in the initial eligibility pool. The top 10% of those by adjusted rating is a group of just 332 games, as opposed to the 454 used for the above sample picks. Because the contestants in the Free IF Playoffs were highly-rated popular games, the reduction in the pool size increases the frequency of previously-appearing games in the sample sets. In a set of ten sample runs for which I counted them up, the average fraction of repeat games was 23.33% (i.e. 12 or 13 games out of 54).
For comparison purposes, I did the same 10-run tests with eligibility pools of the top 12.5% (i.e. top 415 qualifying games) and top 15% (i.e. top 498 qualifying games). For those, the average fraction of repeats was almost identical, both rounding to 17% (i.e. 9 games out of 54). I think the “long tail” pattern of player picks from the People’s Champion Tournament contributes to that, but it could just be coincidence.
Note that those are just averages, and the actual result can vary quite a bit. The top 10% study samples ranged from 10 repeat games to 16 repeat games, and I know that at least one that I looked at before doing the study had 18 repeats – a full third of the field. Also, on occasion I have seen some significant clumping, e.g. a half dozen games from the Little Match Girl series, or an unusually large number of games from the same author.