Darn, the clones are contesting the early pool against me well in part because they put in code to exploit 0-bot and 1-bot and I didn’t. My plans for the early game focused more on dealing with attackers.
I’m curious which of the silly/chaos army bots passed my simulation test and got simulated.
Some clones doing significantly better than others is a bit confusing since for now they’re all supposed to be doing the same thing. I guess some got really lucky/unlucky with other bots’ random rolls?
It’s worth noting that the clones aren’t even being significantly aggressive against outsiders yet. This huge advantage is just from the perfect self-cooperation. I was kind of expecting a midgame where the clones fought a bloody struggle to clear out the non-clone cooperators while I profited off both sides, but the outsiders might be wiped out too fast for that to happen.
Also worth noting that on the next round my fallback behavior changes from a fold-ish EquityBot to DefenseBot. Most attackers seem to be gone or marginal at this point, so I’m not sure that changes much.
All clones behave exactly the same until round 90. Even the seed for the random number generator is the same.
All I can imagine is that a tiny difference in score due to facing different bots snowballs into a significant different pie share due to the multiplicative effect that simon noted. There was a Silly 0 Bot. Any clone that was lucky enough to face it on round 1 gorged itself with score. Same thing with Silly 1 Bot and a few others. Since they disappeared fast, it’s a one-time bump in score that cannot be averaged over time.
Ah, I had misunderstood how the system works. I had not read carefully and assumed some kind of weighted round robin. Random pairings allow for a lot more random variation.
All clones should act equally against non-clones until the showdown round. I guess some outsider bots could be adjusting behavior depending on finding certain patterns in the code in order to respond to those patterns, and the relevant patterns occur in the payloads of some clones?
FWIW, doing better or worse in any given round has a multiplicative effect between rounds, not additive. So that might affect the level of randomness, though even with 100 it seems really big to be random.
Eyeballing the graphs it looks to me that CliqueZviBot is outperforming (multiplicatively) the average performance of the other cliquebots in every single round.
This is super odd if this Bot is indeed acting in exactly the same manner as the other clique bots.
ETA: Genuinely curious how this got downvoted even before it turned out to be correct.
Multicore gained some favor with me when he did an enormous amount of tagging during the tagging sprint. Figured I would use my entry for the good, even if I didn’t have time to write my own thing.
Darn, the clones are contesting the early pool against me well in part because they put in code to exploit 0-bot and 1-bot and I didn’t. My plans for the early game focused more on dealing with attackers.
I’m curious which of the silly/chaos army bots passed my simulation test and got simulated.
Some clones doing significantly better than others is a bit confusing since for now they’re all supposed to be doing the same thing. I guess some got really lucky/unlucky with other bots’ random rolls?
It’s worth noting that the clones aren’t even being significantly aggressive against outsiders yet. This huge advantage is just from the perfect self-cooperation. I was kind of expecting a midgame where the clones fought a bloody struggle to clear out the non-clone cooperators while I profited off both sides, but the outsiders might be wiped out too fast for that to happen.
Also worth noting that on the next round my fallback behavior changes from a fold-ish EquityBot to DefenseBot. Most attackers seem to be gone or marginal at this point, so I’m not sure that changes much.
No, 10 rounds of 100 turns is a decently large sample size—I think some are actually doing badly against outsiders.
All clones behave exactly the same until round 90. Even the seed for the random number generator is the same.
All I can imagine is that a tiny difference in score due to facing different bots snowballs into a significant different pie share due to the multiplicative effect that simon noted. There was a Silly 0 Bot. Any clone that was lucky enough to face it on round 1 gorged itself with score. Same thing with Silly 1 Bot and a few others. Since they disappeared fast, it’s a one-time bump in score that cannot be averaged over time.
Ah, I had misunderstood how the system works. I had not read carefully and assumed some kind of weighted round robin. Random pairings allow for a lot more random variation.
All clones should act equally against non-clones until the showdown round. I guess some outsider bots could be adjusting behavior depending on finding certain patterns in the code in order to respond to those patterns, and the relevant patterns occur in the payloads of some clones?
FWIW, doing better or worse in any given round has a multiplicative effect between rounds, not additive. So that might affect the level of randomness, though even with 100 it seems really big to be random.
Eyeballing the graphs it looks to me that CliqueZviBot is outperforming (multiplicatively) the average performance of the other cliquebots in every single round.
This is super odd if this Bot is indeed acting in exactly the same manner as the other clique bots.
ETA: Genuinely curious how this got downvoted even before it turned out to be correct.
What are the names of your 2 vassal PasswordBots?
PasswordBot and DefinitelyNotCollusionBot. They were submitted by Ruby and habryka, who responded to my request on the LW Tagger Slack.
Multicore gained some favor with me when he did an enormous amount of tagging during the tagging sprint. Figured I would use my entry for the good, even if I didn’t have time to write my own thing.
I see, they’re lumped with your bot in the red portion of the pie, and still running after 10 rounds.