[lmm]

> After extensive experimentation, they realize this: 50% of the time, the white player wins and 50% of the time, the black player wins (we'll ignore draws and any first-move advantage for the example).

But this is empirically false. It fails to explain that, for example, some players consistently beat other players, or at least have a much higher than 50% winrate.

If chess really did consistently just have a 50% outcome, then it really would be equivalent to an overcomplicated coinflip. So I don't find this parable at all convincing, and remain a logical positivist.

[bruce511]

It's not _players_ here that matter, just the color. If I play a grand master (or indeed just any moderately competent player) 10 times, I will lose all 10. But assuming we share who starts, white will win 5 times and black will win 5 times.

Further, for some random chess tournament, I'm guessing the win ratio for black/white is close to 50/50 (perhaps white is a bit ahead for having first mover advantage - which is stipulated as ignored in the article)

It turns out the result of the game has lots to do with the player, and Little to do with the color, which makes the alien first model simple, accurate, and completely wrong.

[lmm]

> It turns out the result of the game has lots to do with the player, and Little to do with the color, which makes the alien first model simple, accurate, and completely wrong.

And discovering that relationship between players and winning is important and valuable! But trying to understand how the pieces move before you've even understood that this is a game of skill is putting the cart before the horse, and gives you a model that really is less useful than the coin-flip model.

[jjk166]

But how can you propose that chess is a game of skill unless you first propose that there are players doing something at which they could be skilled?

[lmm]

You'd observe that the players treat it as a contest, and care about the results. It's like if you were trying to understand, say, a water pump, that you couldn't understand see internals of - speculating blindly about what kind of gears and motors might be inside would be completely futile, the path to understanding it would be to first understand what it's being used for, and then you might be able to start to reason about how.

[jjk166]

No, you would observe blobs of water and carbon jiggling and perhaps making some noises. You'd have to develop quite a sophisticated model of human behavior to infer that they are treating it as a contest and care about the results. Even then, if they've previously seen humans at a casino, they might reasonably deduce that chess is no more a game of skill than craps is.

[lmm]

> You'd have to develop quite a sophisticated model of human behavior to infer that they are treating it as a contest and care about the results.

Sure. But again, trying to understand chess without having that understanding would be putting the cart before the horse.

> Even then, if they've previously seen humans at a casino, they might reasonably deduce that chess is no more a game of skill than craps is.

That would be a reasonable starting assumption, but they'd eventually notice contradictions: the fact that some players consistently had advantages over others, more experienced players generally beat less experienced players, commentary and analysis of board positions is considered worthwhile...

[raldi]

The aliens can’t see the players, other than that one is playing as “white” and the other “black”. They don’t have access to information about a particular player going from game to game.

[lmm]

That's not how I read the article's introduction. But under that interpretation, it would be meaningless for the aliens to try to understand the game in any more detail until they understood that kind of basic information.

[simonh]

That's the question the article is addressing. I think you're putting too much weight on the rough and sketchily defined example and missing the actual topic.

[lmm]

But the claim lives or dies by its example. The author's whole point is that he's discovered or imagined a situation in which doing science "badly" would be better than doing it "well". If his imaginary situation doesn't actually hold up then his whole argument is nonsense.

[simonh]

Analogies are simply tools to facilitate discussion, they can't themselves say anything definitive about the underlying claim. That's why it's useful to interpret them generously. If I came up with a flawed analogy for time dilation would that show that time dilation is nonsense?

I think a better approach on encountering a flawed analogy is to attempt to improve it, come up with a better one, or address the underlying claim directly, but saying "this analogy is unclear to me the end" isn't going to get you very far in life.

[lmm]

> Analogies are simply tools to facilitate discussion, they can't themselves say anything definitive about the underlying claim. That's why it's useful to interpret them generously. If I came up with a flawed analogy for time dilation would that show that time dilation is nonsense?

That's a bizarre attitude. Thought experiments were a major and important tool in developing the theory of relativity, precisely because they were taken seriously and done rigorously; contradictions weren't waved away as "just a tool to facilitate discussion".

> I think a better approach on encountering a flawed analogy is to attempt to improve it, come up with a better one, or address the underlying claim directly, but saying "this analogy is unclear to me the end" isn't going to get you very far in life.

On the contrary, being willing to call out nonsense has served me very well. There is no "underlying claim" here; the story is what's meant to carry the claim, and if the story doesn't work (and it doesn't) the whole thing falls apart (and it does). It's nonsense, top-to-bottom.

[arrrg]

I mean, who is which color actually is a coin flip (it‘s random or as good as random and there are repeat games to make sure the first mover advantage doesn’t play a role), so the model seems to be correct in that regard. They modeled that part of reality correctly, just missed that that’s not what the game is about.

It‘s an interesting way of being “wrong”.

[TameAntelope]

The author did himself a disservice by attracting chess players to criticize his example that clearly could have been about literally any game more complex than coin flipping that has roughly 50/50 parity.

He could have referred to the NFL or something, maybe that would have helped. The point was to set up the aliens to have a gross approximation of outcome determination that is both accurate and wrong.

The aliens could have decided to use home vs. away teams as their "coin flip" and stopped investigating once they found out that home teams win a bit more, for example, and the "crazy" alien in the parable could suggest that examining only the first play of the game is a better answer. The "crazy" alien would be told to stop because the first play isn't more predictive, but the fact that the alien is looking at gameplay is more "right".

[xscott]

>The author did himself a disservice by attracting chess players to criticize his example

Heh, I think the author's example is fine if the reader gives it a charitable interpretation. The disservice is from all the overly pedantic people who want to argue about chess instead of the philosophy of science.

[vecter]

There are so many unreasonable assumptions baked into this parable that make it difficult to take seriously. Any amount of meaningful investigation by the aliens would yield more insights into the fact that the game is in fact not a random coin flip. Even recognizing that white has a slightly higher win rate [0] would demonstrate some asymmetry that suggests the existence of a deeper structure or symmetry-breaking.

[0] https://en.wikipedia.org/wiki/First-move_advantage_in_chess

[ALittleLight]

The article is not a treatise on how aliens might investigate chess though. That's just a symbolic example to illustrate the idea that you could have a simple model with good predictive power that is somehow less correct than a complicated model with worse predictive power.

[lmm]

But the whole point of the article is to claim that its example are possible, and yet it fails to demonstrate that. We can posit a world in which God declares that logical positivists may not enter heaven, and in such a world logical positivism would be a poor choice, but there's no evidence that that's actually the world we live in.

[8note]

The pieces that it does predict some things better:

Is one player happier than the other right now?

How long will it take to make the next move?

How long will the game last?

There's lots and lots of observable things other than who wins that the coin flip model does a worse job at. Thus, the example makes just as good an argument that making better predictions is still what makes a model better.

The missing question is "what are you using the model for?"

Are you trying to learn chess? Understand human psychology? Predict the outcomes of chess games? The same model isn't going to be the best for all of them. Like, the super GM level stockfish is going to suck at predicting beginner chess games

Whether one model is better than another is all about the purpose it's trying to serve, and how well it does that.

[est31]

Have they had individuals play the game enough times so that they can see this first mover advantage with a good p value?

Maybe they only know that the win chance for white is between 0.4 and 0.6.