[subroutine]

The statement, "A machine that allows his alien eyes to see the first move." seems too information-rich for what the author then describes as the alternative model. Perhaps not the point of this thought experiment, but let's assume the aliens can only crudely see the first move: they know how many pieces there are on each side of the board, their color, but otherwise all the pieces look the same. They cannot see the surface of the board (they don't know it is checkered), but they can determine the exact center of each piece (so during the first move they can compute a precise coordinate of the piece that moves). Promptly after the first move, the image goes away. Also assume they trust the humans who've told them: "there are actually pieces on the board with functions". They have collected data on 1000s of matches.

What can the aliens know or deduce?

- The number of pieces

- The pieces are divided into 2 different colors (white and black)

- The pieces are split by color and placed at opposite (vertical) ends of the board.

- There is the same number of pieces in each color

- The pieces are arranged into 2 vertical rows and 8 horizontal columns

- Opening moves cluster at 2 different lengths, towards the opposite side of the board.

- The longer opening move length is typically ~2x the short opening move length.

- The opening move vector is always towards the opposite end of the board (not horizontal)

- The opening move is always white

- White wins slightly more often

- They are given the duration and outcome of each game (win, loss, or tie)

- ? When games end quickly, first-move advantage shrinks

- ? Sometimes a series of games end more quicker than usual (they observe a blitz tourney)

- ? Win P() is slightly different for each of the 8 (x2) possible opening moves

- ? Everything here: https://old.chesstempo.com/images/en/openingExp.png

What else?

Also, a note regarding the statement in the article that the alternative model "is less predictive over iterated games than the coin flip model". I don't think this is technically possible. If the model was selecting the winner at a statistically worse rate than 50% chance, you could just flip the sign of the prediction, and do that much better than chance.

[slx26]

Observing the movement of pieces quickly reveals that the board is divided in 64 squares. Squares being black and white or any other set of colors is actually irrelevant. They can basically fully understand the game from this, no need to make things complicated.

[subroutine]

I'm not sure they would be able to infer the board has 64 squares. They can't see the board, and can only see the locations of the starting pieces and location of the first move. The center of the board could be anything - 300 hexagons, a spin-wheel, series of mouse traps, etc.

Here's my interpretation of what they see pre- and post- opening move...

PRE: https://i.ibb.co/s6WnPNV/Chess.png

POST: https://i.ibb.co/7pg7GHW/Chess1.png

[tiborsaas]

This is a fun one, thanks for the illustration.

If you analyse 5000 board configurations you can deduce with a high certainty that there are 64 fields the pieces can be on.

The pressing question is then why can some pieces move only diagonally or only in one direction or any.

In the case of chess, going after the why question is pointless, but obviously that wouldn't satisfy any scientist: It works because someone decided that it should work like this.

Could it work with different rules? The alien scientist then probably opt for a positive answer calling for a multiverse of chess rules :)

[subroutine]

In keeping with the original authors description, the aliens can only see the opening move. I'm confused - how could they determine there are 64 fields (nb: I only know the very basics of chess)?

If they could view entire matches, it seems a trivial task to acquire a complete understanding of the underlying game, no?

[tiborsaas]

Oh sorry, I missed that part. But according to the rules of chess a random valid opening should cover all fields on the table so at least you'd know the layout. Pawns can move one or two fields.

Knowing only the first move would limit understanding heavily, that's true.