The Probability of a Monkey Defeating Yi Chang-ho

[tchan001]

http://tchan001.wordpress.com/2012/08/1 ... -in-a-tie/
Read the second part of this post on a new Chinese go book enjoy

[daniel_the_smith]

I think actual random play is more like 50k or 150k than 20k. But still an amusing piece.

[tchan001]

But the monkey isn't just playing randomly, he's playing random moves which are legal moves.

[topazg]

But the monkey isn't just playing randomly, he's playing random moves which are legal moves.

Even within legal moves, I'd also estimate randomly is at best 50k

[hyperpape]

Absolute beginners really are almost 30k on KGS, and they're playing only legal moves, most of which are better than random.

[tchan001]

Is there actually a way to play illegal moves on KGS?

Regardless of whether the monkey is 21k or 50k, the point is P1 is a very very small probability but will still be much bigger than P0 and the rest of the arguments will still be very usable.

Anyways, I'm not the author of the piece. I'm only presenting it in English to the Western audience

[topazg]

Absolute beginners really are almost 30k on KGS, and they're playing only legal moves, most of which are better than random.

I've played a lot of absolute beginners, and they don't compare to idiotbot for badness, even when they're playing their first game having had the rules just explained to them.

[HermanHiddema]

A "random legal move monkey" is not stronger than 300k

[palapiku]

If you accept the assumptions given in the article about the relative strength between ranks, then P0 corresponds to a rank of 330 kyu.

The actual strength of a random player is probably much lower than even that (what would stop it from killing all of its living groups?)

[HermanHiddema]

(what would stop it from killing all of its living groups?)

Presumably there is a chance that it will pass in a situation where white has no legal moves and must also pass, so the game is then counted?

E.g Suppose I give it 280 handicap, putting a black stone on every coordinate where either the row or the column is odd (so only the point that have both coordinates even remain open). If it passes before filling 80 of its 81 eyes, it then wins.

[palapiku]

That's a good point... I didn't think of the actual logistics of playing with huge handicap

Of course, the assumption in the article is wrong anyway - the probability of losing to someone one stone weaker is not a static 25%. It actually approaches 50% as strength of the players declines, i.e. komi becomes less of a factor. (This happens because variance of final scores increases. High dans often end games within a few points of each other. A game between two nearly-random players could easily finish with a difference of thousands of points).

[hyperpape]

Absolute beginners really are almost 30k on KGS, and they're playing only legal moves, most of which are better than random.

I've played a lot of absolute beginners, and they don't compare to idiotbot for badness, even when they're playing their first game having had the rules just explained to them.

And idiotbot isn't really 30k. People resign against it: http://eidogo.com/#url:http://files.gok ... t-Brau.sgf.

[Koosh]

I think the most interesting part of this article is the attempt to show the gap between ranks mathematically. A simple range of numbers like this really puts the difficulty of rising to the top into perspective. Much more than the (x books to 1k, y books to shodan, z books to 2dan, ect…) theory which doesn’t seem to have any basis whatsoever.

[Bill Spight]

OK. Random Monkey sees eight legal moves on the board, and he also has a pass move available. He has four live groups, each with two single point eyes. Eight to one Random Monkey fills an eye.

[Txewì]

We can assume Random But Not Stupid Monkey who will pass if and only if all legal moves fill an eye.