A Test for Board Positions

judicata · #1

I've been thinking of an interesting-sounding experiment, and I suspect it has been done before (perhaps they do this in go schools to analyze positions?): Start with a board position anywhere from 20-100 moves (or whatever) into a game that is thought to be roughly equal and have several pairs of players of similar strength play it out. Pro games might be best, but not necessarily.

It would be interesting to see the results of a pro-pro experiment (i.e., pros playing out pro games) and compare them to amateurs' results of the same game (e.g., what if black wins most of the time if amateurs play it out, but white wins more if pros play it out?). Alas, this will likely never happen because of sample size problems, etc. But, oh well.

Perhaps playing styles would affect the outcomes (maybe especially among amateurs): e.g., if black opens with the low chinese and that is the substitute's favorite opening, black may be more likely to win, etc.

A pre-emptive response to naysayers: No, the experiment is not perfect--it could even be misleading. But it should still be interesting.

Again, I have to suspect that this has been done before. Anyone know?

Kirby · #2

It seems that the challenge with this type of experiment is in determining board positions that are "thought to be roughly equal".

What is thought to be roughly equal may differ from person to person, and could have a correlation with game results involving said people.

TMark · #3

One of the Japanese magazines used to do this, getting strong amateurs to play one side and a pro to do the same and compare the results. Mostly, I think it was experiments in yose, to see how much difference there would be in the results.

Best wishes.

freegame · #4

why not do this experiment here on the forum. :tmbup:

Just select a finnished pro game and post the position at move 100.
from the finished game you already know how two pro's played out the game.
next ask people on the forum to play out the game in pairs of two (with each pair approx same rank)

pairs can just set a date and time to play the game out on KGS and post the result on the forum.
I would be willing to participate.

daal · #5

Although this focuses on the last few moves, Bill Spight posted a fairly similar experiment just this morning:

viewtopic.php?p=50322#p50322

judicata · #6

Kirby - c'mon, I didn't post enough caveats?

Seriously though, it isn't hard to find a pro game that are close enough to equal after 50 moves. It really doesn't matter how close to equal the game actually is--the experiment would just be less interesting if one side had obviously goofed already.

Tmark - Neat. Because of the nature of the endgame, I bet such an experiment can net concrete results.

On freegame's suggestion, I think I'll post a game later with a sign-up to see if there is any interest. If not, no loss.

daniel_the_smith · #7

I might be interested in helping that experiment...

Bill Spight · #8

OK, Here is a game between top pros, that ended in a half point win.

Click Here To Show Diagram Code: [go]$$B Black to play, move 103. Japanese rules. 5.5 komi. $$ --------------------------------------- $$ | . . . . . . . . . . . . . . X X O O . | $$ | . . X X O . . . . . O X X . X O O X X | $$ | . . X O O X . . . . O O X X X X O . . | $$ | . . X O . . . X . O . . . . O X O O O | $$ | . X . . O . O O O . . O O X O X X X . | $$ | . X O . . . X X X X . . . O . O O X . | $$ | . O . . O . . . . . X . . . . . O O . | $$ | . . . . . . . O O O X . . . . . . X . | $$ | . . . O . . . X O X X O . . O . O . . | $$ | . . X O . . . O X , X . . . . , . . . | $$ | . . . X O O . . X . . . . . . . X . . | $$ | . . X . X . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . X . . . . . . . . . . X . . . | $$ | . . X . . . . . . . O . . . . . . . . | $$ | . . X O . . . . O . . . . O . , X . . | $$ | . O O . . . O . . . X . . . . X . . . | $$ | . . . . . . . . . . . . . X . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Loons · #9

So... Judicata's conceptual experiment aims loosely at finding if there are types of advantageous or even positions that amateurs have trouble using one side of ? (Like, a large sample of pros will reach about 50-50 win rate playing a game out themselves or amoungst themselves starting from a thickness vs profit early game position, but a large sample of amateurs might win significantly more with the territory (or thickness) ?

We could have a series of games starting from the endpoints of very complicated thickness vs profit joseki (in correct fuseki contexts) between players of same rank (or varying rank with adjusted komi) to reach similar findings, if we accept as given that a situation described as 'even' by pros is really even by pro standards.

Edits various: Post slightly changed and added to.

judicata · **#10**

Let's see if this works

.

Guidelines:
(1) Sign-up as you would a Malkovich game, using this thread (maybe I should've created a new one?). I suspect that this would work best playing a live game (e.g. on KGS), so it may be time-zone dependent.

(2) Players should be fairly close in strength. This is up to the players, really. All strengths are welcome.

(3) No looking up the game, of course.

(4) Post the sgf in this thread.

Your starting position is below-a pro game up to move 50. I had some trouble picking a game and starting point. For example, I didn't want to start the game in the middle of a complicated fight. I managed to find one with pro commentary that the position was equal.

EDIT: Or we could use Bill's which has more moves...

judicata · **#11**

Loons wrote:

So... Judicata's conceptual experiment aims loosely at finding if there are types of advantageous or even positions that amateurs have trouble using one side of ? (Like, a large sample of pros will reach about 50-50 win rate playing a game out themselves starting from a thickness vs profit early game position, but a large sample of amateurs might win significantly more with the territory (or thickness) ?

That would be interesting, and is a suspicion I have. Of course, I doubt we'd ever get enough players (pro and amateur) to have any confidence in this. My aim is a broader "let's see what we get" approach.

Bill Spight · **#12**

judicata wrote:

EDIT: Or we could use Bill's which has more moves...

Sorry, I overlooked your note that said you were choosing a game. Go ahead with the one you picked.

daniel_the_smith · **#13**

Your SGF file goes to 60, and that seems like a more logical place to start

judicata · **#14**

Bill Spight wrote:

judicata wrote:

EDIT: Or we could use Bill's which has more moves...

Sorry, I overlooked your note that said you were choosing a game. Go ahead with the one you picked.

No worries - I actually started typing my post before yours, but then I changed my mind on the game 4 times, so by the time I posted it, your post was up. My fault for being slow

.

freegame · **#15**

to get this going and provide an example the fist played out game (blitz)

B+5.5

A Test for Board Positions

Who is online