This reminds me of a theory about how oracle bones (throwing bones in a fire and interpreting patterns in the cracks to tell the future) actually gave advantages to hunters. If we assume the patterns were as good as random, then the influence on the hunting pattern would be to increase the chaos and unpredictability. Therefore a group of hunters who used bones to tell them when and where to hunt would be less predictable, and harder to avoid. (The animals couldn't learn your pattern, because by definition there is no pattern in randomness.)
Sirlin on Design's video titled "Solvability and Donkey Space (Game Theory)" comes to mind. In there, he walks about how this aspect of gameplay at greater lengths, although Jon does a great job at summarizing the principle.
There's a card game called Bullsht where you have a 52 card deck, and you play cards face down and declare what they are. You win when you're out of cards. Nosey people try to get into everyone's heads so they can dominate the game because they can call you on lying, and then you have to pick up all the cards placed down, and the next player gets to play whatever they want. If you decide your goal is to punish people for trying to mindread too much, you can bait them into losing to the player next to you by always obviously lying. You can make their worst nature play against them trivially, and they will not be able to win the game until they understand they are being punished for playing too aggressively by players who feel just as much victory in choosing the winner as they feel when they are winning. What you want to do in the game matters far more than what the rules suggest you should want. Letting rules dominate your goals is a trap. Part of the point of the game is to provide an environment where people who won't mind their own business can have a chance to see the consequences of their actions in low stakes environment.
This reminds me of a space game whereas there are players who go farming/mining, and pirates who farm them. The problem here is simply that as the initiating one you have a massive advantage (fully offense equipped ships, any group size, enemy is usually alone) and disrupt the gameplay of someone else who just wanted to do something calm for a while. Sure, you have the satisfaction of ruining someone's time... and I have the issue of possibly losing customers for having them being punished for not harassing other players.
I think the analogies here are conflated and only tangential, but there is a kernel of truth to what he's saying. "Outside the box thinking" vs. "Tried and true".
You can only lose by deviating from a GTO-strategy, because math, unless the opponent is also deviating, then there are exploitative strategies. The analogy to studios blows, predictability is not an issue, you play each option in rock paper scissors 1/3 of the time and break even (in complex games GTO strategies crush and are hard to approximate). What big studios do is not going to be nearly optimal, human estimations of what is optimal in complex systems suck even under experts, we just find good enough heuristics. The dead players essay seems like a pop sciency way to say that not being able to adapt is a problem. Big studios just attempt to copy what made money before which at least in terms of making good products is extremely flawed.
>you play each option in rock paper scissors 1/3 of the time and break even The point is that humans are not rational automatons, and if you pay attention, you might notice that the distribution is uneven - influenced by biases people associate with the concepts of "rock", "paper" and "scissors", and by biases against themselves. Please google it, I'm not sure if I can post links these days, it's going to be on the first page
Exactly. When Jonathan is taking about “game theory optimally poker.” The strategy he’s describing is not game theory optimal poker - he’s describing a poker strategy that is good, but doesn’t adapt or incorporate weighted randomness. But, a truly game theory optimal poker strategy would be unpredictable and account for all possible other strategies. Poker is actually a game mathematicians still can’t “solve” the optimal strategy for. What Jonathan is talking about here is like if someone built a poker bot that plays poker in some good but non-optimal way. If you can always predict the behavior of the bot then you can take advantage of the bot. Ex: if the bot always and only goes all in with a pair of aces. If you’re a “live player” and pick up on that pattern, then whenever you see the bot go all in, you know it has a pair of aces, so you never lose to it and you always fold. In a truly GTO poker bot, an optimal poker bot would bluff some percentage of the time with any hand, and there would never be a way to be sure what the bot has in its hand. Over the long run, a game theory optimal “dead player” bot would never have to change its strategy, and would always come out on top with a large enough sample size, whether the other players are live or dead. It’s just hard to conceptualize that in poker. In rock paper scissors, optimal play is just 1/3rd chance of each option. In the long run, no strategy can beat or take advantage of a “1/3 of each option” strategy in rock paper scissors
@@DrinktheLatteThere's an approximate (as in, if they spent more computational effort they can make the approximation even closer but it's long past the point where it doesn't matter in practice) solution for Heads Up (ie two players) Limit (ie you don't need to decide how much to bet) Texas Hold 'Em (today's most popular poker variant) named Cepheus. Dead players running Cepheus can't be beaten, another player with the same strategy will break even, and the same for a hypothetical different GTO strategy since there's no reason to believe there can't be more than one possible optimality. Any "live players" trying to exploit this just lose money. For example if Cepheus looks down at two of diamonds and four of spades as first to act, it will always just fold. You might suppose you could benefit here, surely you can open such hands occasionally and get value Cepheus does not? Nope, if you do that the Cepheus strategy is going to keep taking more of your money than you win. Of course a human can't memorise the strategy, it's too hard for anything other than a machine.
Poker is not a good example here. In poker, the game theory actually dictates that there's an optimal strategy, that does not lose to any other strategy. So, if the "dead players" are just playing the game-theory-optimal strategy, then no one can come in and eat their lunch.
Is that really true? Surely even an optimal strategy relies on some statistical norm, right? That doesn't mean someone else could invent a better one, but you could hypothetically take greater risks to win in a case where you otherwise might not have?
False. Maybe in a a1v1 scenario, but one vs multiple players the game theory falls apart. If someone goes all-in, there is no right way to behave, either you think they are bluffing or you don't
"Tragedy of the commons and why i hated tournament poker" is a video from someone who used to live off poker. I highly recommend it because it was interesting. It touches on at least one of the incentives that might land you outside the solved/optimal strategy. But I'm just gonna spoil part of it. The point of the video is that this thinking isn't done in poker and it's kinda depressing.
Poker is not solved, and if you listen again, Jonathan never claims that it is. And it wouldn’t be relevant if he did. He is precisely talking about non-optimal strategies and is not wrong about non-optimal strategies
This reminds me of a theory about how oracle bones (throwing bones in a fire and interpreting patterns in the cracks to tell the future) actually gave advantages to hunters. If we assume the patterns were as good as random, then the influence on the hunting pattern would be to increase the chaos and unpredictability. Therefore a group of hunters who used bones to tell them when and where to hunt would be less predictable, and harder to avoid. (The animals couldn't learn your pattern, because by definition there is no pattern in randomness.)
God have superior sense of humor
Making dumb people trusting oracles look like fools, but also making smart who don't trust oracles look like fools
No good answer lmao
Sirlin on Design's video titled "Solvability and Donkey Space (Game Theory)" comes to mind. In there, he walks about how this aspect of gameplay at greater lengths, although Jon does a great job at summarizing the principle.
There's a card game called Bullsht where you have a 52 card deck, and you play cards face down and declare what they are. You win when you're out of cards. Nosey people try to get into everyone's heads so they can dominate the game because they can call you on lying, and then you have to pick up all the cards placed down, and the next player gets to play whatever they want. If you decide your goal is to punish people for trying to mindread too much, you can bait them into losing to the player next to you by always obviously lying. You can make their worst nature play against them trivially, and they will not be able to win the game until they understand they are being punished for playing too aggressively by players who feel just as much victory in choosing the winner as they feel when they are winning. What you want to do in the game matters far more than what the rules suggest you should want. Letting rules dominate your goals is a trap. Part of the point of the game is to provide an environment where people who won't mind their own business can have a chance to see the consequences of their actions in low stakes environment.
That card game is so fun, thanks for reminding me of its existance!❤
This reminds me of a space game whereas there are players who go farming/mining, and pirates who farm them. The problem here is simply that as the initiating one you have a massive advantage (fully offense equipped ships, any group size, enemy is usually alone) and disrupt the gameplay of someone else who just wanted to do something calm for a while. Sure, you have the satisfaction of ruining someone's time... and I have the issue of possibly losing customers for having them being punished for not harassing other players.
I think the analogies here are conflated and only tangential, but there is a kernel of truth to what he's saying. "Outside the box thinking" vs. "Tried and true".
should put NPC char in the thumb
Mang0 vs. mew2king
I had the same thought
You can only lose by deviating from a GTO-strategy, because math, unless the opponent is also deviating, then there are exploitative strategies. The analogy to studios blows, predictability is not an issue, you play each option in rock paper scissors 1/3 of the time and break even (in complex games GTO strategies crush and are hard to approximate). What big studios do is not going to be nearly optimal, human estimations of what is optimal in complex systems suck even under experts, we just find good enough heuristics. The dead players essay seems like a pop sciency way to say that not being able to adapt is a problem. Big studios just attempt to copy what made money before which at least in terms of making good products is extremely flawed.
I wanted to understand this, but I find your writing difficult to read.
>you play each option in rock paper scissors 1/3 of the time and break even
The point is that humans are not rational automatons, and if you pay attention, you might notice that the distribution is uneven - influenced by biases people associate with the concepts of "rock", "paper" and "scissors", and by biases against themselves. Please google it, I'm not sure if I can post links these days, it's going to be on the first page
Exactly. When Jonathan is taking about “game theory optimally poker.” The strategy he’s describing is not game theory optimal poker - he’s describing a poker strategy that is good, but doesn’t adapt or incorporate weighted randomness. But, a truly game theory optimal poker strategy would be unpredictable and account for all possible other strategies. Poker is actually a game mathematicians still can’t “solve” the optimal strategy for. What Jonathan is talking about here is like if someone built a poker bot that plays poker in some good but non-optimal way. If you can always predict the behavior of the bot then you can take advantage of the bot. Ex: if the bot always and only goes all in with a pair of aces. If you’re a “live player” and pick up on that pattern, then whenever you see the bot go all in, you know it has a pair of aces, so you never lose to it and you always fold. In a truly GTO poker bot, an optimal poker bot would bluff some percentage of the time with any hand, and there would never be a way to be sure what the bot has in its hand. Over the long run, a game theory optimal “dead player” bot would never have to change its strategy, and would always come out on top with a large enough sample size, whether the other players are live or dead. It’s just hard to conceptualize that in poker. In rock paper scissors, optimal play is just 1/3rd chance of each option. In the long run, no strategy can beat or take advantage of a “1/3 of each option” strategy in rock paper scissors
@@DrinktheLatteThere's an approximate (as in, if they spent more computational effort they can make the approximation even closer but it's long past the point where it doesn't matter in practice) solution for Heads Up (ie two players) Limit (ie you don't need to decide how much to bet) Texas Hold 'Em (today's most popular poker variant) named Cepheus.
Dead players running Cepheus can't be beaten, another player with the same strategy will break even, and the same for a hypothetical different GTO strategy since there's no reason to believe there can't be more than one possible optimality. Any "live players" trying to exploit this just lose money. For example if Cepheus looks down at two of diamonds and four of spades as first to act, it will always just fold. You might suppose you could benefit here, surely you can open such hands occasionally and get value Cepheus does not? Nope, if you do that the Cepheus strategy is going to keep taking more of your money than you win.
Of course a human can't memorise the strategy, it's too hard for anything other than a machine.
I call the dead players - metaslaves.
😬
opposed to the metabarons. Hm!
Do you really need to read this essay? Or can I just watch Data bust-up Kolrami in Stratagema?
Poker is not a good example here. In poker, the game theory actually dictates that there's an optimal strategy, that does not lose to any other strategy. So, if the "dead players" are just playing the game-theory-optimal strategy, then no one can come in and eat their lunch.
Is that really true? Surely even an optimal strategy relies on some statistical norm, right? That doesn't mean someone else could invent a better one, but you could hypothetically take greater risks to win in a case where you otherwise might not have?
False. Maybe in a a1v1 scenario, but one vs multiple players the game theory falls apart. If someone goes all-in, there is no right way to behave, either you think they are bluffing or you don't
No, poker hasn't been solved.
"Tragedy of the commons and why i hated tournament poker" is a video from someone who used to live off poker. I highly recommend it because it was interesting.
It touches on at least one of the incentives that might land you outside the solved/optimal strategy.
But I'm just gonna spoil part of it. The point of the video is that this thinking isn't done in poker and it's kinda depressing.
Poker is not solved, and if you listen again, Jonathan never claims that it is. And it wouldn’t be relevant if he did. He is precisely talking about non-optimal strategies and is not wrong about non-optimal strategies