PrisonersDilemmaTournament A cheater got to first place (and got caught).

The raw quick and easy results are out, and a video of some of it should be uploaded at some point, but cary got caught up in the first place strategy cheating by loading the other strategy and finding out what moves it will make ahead of time. The person who cheated admitted to doing it when they were asked and that they were just trying to see if it would work. Cary's trying to figure out how much needs to be done in order to get the true results, since there were some other strategies that were potentially behaving weirdly.

Jun 18 '21 10:06 duckboycool

RIP 1059, you will be remembered

Jun 18 '21 10:06 CursedFlames

The video is out now, but it was recorded before this all happened. Everything else besides the first place and first place being an outlier in the graphs should still be good though.

Jun 18 '21 10:06 duckboycool

I think we can still learn something from this, while most people are trapped in a dilemma, some people can just game the system and gain an advantage. Much like real life lol

Jun 18 '21 13:06 arielsysdev

I honestly am interested in the number of (bad actor) submissions there was. I personally poked around with breaking it just to see what was possible and instantly figured out that you could manipulate random to mess with other strategies as well as override some of the numpy functions being used to get information you where not meant to have. I did look briefly into trying to load the code of the other player but didnt see anything off the bad without importing other libs. (fyi I did not submit anything to the comp just messed around locally)

Jun 18 '21 19:06 m10653

If all the data was saved from the 30 full rounds, no more simulations need to be run and ProfessorX can just be skipped when summing up the scores.

Jun 19 '21 01:06 Zac-OS

If all the data was saved from the 30 full rounds, no more simulations need to be run and ProfessorX can just be skipped when summing up the scores.

Yup, there are 30 numpy arrays with the full pairings results. Cheaters can be filtered out without needing to re-run the whole tournament.

Aside from cheaters who've already been caught, there was discussion of possible king maker strats influencing the meta. Our top strategies in the discord server scored 2.9 - 3.0 while none of the strats here scored above a 2.85 possibly indicating minions defecting against non-king strats.