Monday, March 28, 2016

TournamentCompare v0.004 - Player Pools

One slight deviation from the old Clarion application is that I have instituted player pools as a starting point for a tournament. All players for a tournament are entered into a sorted list (player pool) that the tournament template draws from. This allows the generation of the player pool to be independent from the tournament template and can therefore be used to test all sorts of varied starting positions and even seeding patterns.

For smaller tournament formats (8 players or less), every possible player pool combination can be generated and used in sequence to eliminate another possible bias. For 8 players this would be 40320 combinations (8 factorial). If a tournament is larger than 8, you could potentially take the top 8 combinations and interleave other random players to give 16, 32, 64, or 128 player pools. This would give a large control set (~40,000 samples), but at least give a somewhat even spread of higher candidates.Unfortunately it would also eliminate the random chance that the most skilled player meets the second most skilled player in round 1 of a 128 player tournament. Since generating a random player pool will be quite easy, I can also test the difference of these two formats and isolate a specific amount of bias due to potential seeding.