Discussion: What’s the lowest scoring alg?

[nobody5050]

I got to thinking, and it’s really hard to get a low scoring alg. Even playing always cooperate will still net you a pretty high score in the grand scheme of things because playing against TFT and always cooperate will raise your average.

[jherndon8]

You could make a reverse detective. Only cooperate if you're sure it's always cooperate, grimTrigger, or random, or any other non-tit-for-tat strat. Tit-for-tats you always defect.

[jherndon8]

But that'd be a whole different ball game if the pool of strategies was actually going for the lowest score

[Equbuxu]

I believe minimizing the score is basically the same problem as maximizing the score. It's always beneficial for you to cooperate even though if both parties defected the total score would be lower. The same dilemma but reversed

[Quadrapod]

I did some playing around and it seems to be fairly different. You want a few things to happen.

Against the static strategies you want to cooperate. So against always cooperate you want to cooperate so you don't exploit them and against always defect you want to cooperate so they exploit you.

You want to defect as soon as possible against grimTrigger. That will make it identical to always defect.

You want detective to think you're always cooperate so they exploit you. This can be done by cooperating 4 times in a row.

You want to cooperate with random so it's +1.5 compared to defecting which is +3.

You want to defect always against TFT strategies including Joss.

Combining that all together you want your first six moves will be [C, C, C, C, D, D]. You need to cooperate four times to get detective to exploit you, and you need to defect once to activate grim trigger and again to activate FTFT. From that point it's all about categorizing each of the strategies and choosing the worst response to each.

Also @JesseB0rn a score below 1 is impossible. The minimum outcome per round is 1.

[carykh]

omg this is a funny concept! Maybe i should shoutout the worst-scoring player in my analysis video, too 😆

I do agree with @Equbuxu that the game should be similar when reversed.

In this version, the rewards-breakdown is 5/3/1/0, but if it were symmetrical like 3/2/1/0, then I think scoring the lowest score would be exactly identical. Except, you'd know that most opponents have an opposite incentive than you, which would change things a bit

[JesseB0rn]

I did some playing around and it seems to be fairly different. You want a few things to happen.

Against the static strategies you want to cooperate. So against always cooperate you want to cooperate so you don't exploit them and against always defect you want to cooperate so they exploit you.

You want to defect as soon as possible against grimTrigger. That will make it identical to always defect.

You want detective to think you're always cooperate so they exploit you. This can be done by cooperating 4 times in a row.

You want to cooperate with random so it's +1.5 compared to defecting which is +3.

You want to defect always against TFT strategies including Joss.

Combining that all together you want your first six moves will be [C, C, C, C, D, D]. You need to cooperate four times to get detective to exploit you, and you need to defect once to activate grim trigger and again to activate FTFT. From that point it's all about categorizing each of the strategies and choosing the worst response to each.

Also @JesseB0rn a score below 1 is impossible. The minimum outcome per round is 1.

Typo, meant to type 1,115 , the font i changed my termial to is really badly readable

[redtachyon2098]

Fun concept indeed!