Predicting who wins after the opening

[Uberdude]

From another thread the issue of the effect of an opening advantage on who wins the game came up. User 'often' was of the view that it had no effect on win rate, whereas I thought it had a small advantage, my guesstimate was a 10 point lead for KGS 5ks would lead to a perhaps 55-60% win rate, and maybe 65% for KGS 3ds. This thread is to test that by picking games from KGS or elsewhere, making a judgement of the position after the opening (and interesting exercise in itself) and then seeing who wins.

So to kick things of I picked a random KGS 3d game:

And this game is near enough to even after the opening and once the middlegame fighting starts that it can't contribute to this experiment.

Next game! (I suspect most games will be within 10 points after the opening so hunting for suitable games will take a while; maybe we should use weaker players).

Again even after opening.

Table of results so far:

Code: Select all

KGS 3d
<10 lead: 3
>=10 leader win: 1
>=10 leader lose: 0

[Jujube]

Might be interesting to have a strong engine evaluate (yeah not ideal) the opening (when does it know the opening is finished?) so that games can be done in batch? See what correlation there is between evaluation function and result?

[Uberdude]

Might be interesting to have a strong engine evaluate (yeah not ideal) the opening (when does it know the opening is finished?) so that games can be done in batch? See what correlation there is between evaluation function and result?

Yup, I suggested that on the original thread, this is the manual version. Anyone have crazystone/fuego/pachi/zen etc and fancies writing a script?

[emeraldemon]

How about my game? Who was ahead in the opening?

[gowan]

I don't think many people (if any) can accurately estimate the score after the opening and the beginning of the middle game. If pros could do this we'd see many games resigned after the opening. Another point is that there is no clear way to specify when the opening ends. Middlegame fights happen during the opening sometimes. And what about those games with no fuseki? Of course fuseki is not synonymous with opening.

[Kirby]

I don't think many people (if any) can accurately estimate the score after the opening and the beginning of the middle game. If pros could do this we'd see many games resigned after the opening. Another point is that there is no clear way to specify when the opening ends. Middlegame fights happen during the opening sometimes. And what about those games with no fuseki? Of course fuseki is not synonymous with opening.

I think there's a difference between estimating the score, and knowing the result. I'm sure there are board positions where pros would unanimously agree that one side, black or white, is ahead. But it doesn't mean that it's the end of the game - pros make mistakes, too.

In one of the other study journals, someone recently brought up an old game from 2010 where Lee Sedol died pretty big early in the game. He happened to go on and win it.

Around that time, Myungwan was giving a workshop in the bay area, which I attended. I remembered him commenting about that game. He said he took a look in the early middle game, and it looked like Lee Sedol had lost, so he stopped watching, and went on to do something else. When he came back to see that Lee Sedol had won, he was surprised.

I doubt that Myungwan's estimation that Lee Sedol was behind after dying was inaccurate. Probably Lee's chances of winning were pretty slim. Comebacks can still happen, though, which is why I think there's a difference between estimating the score and knowing the result.

[gowan]

I don't think many people (if any) can accurately estimate the score after the opening and the beginning of the middle game. If pros could do this we'd see many games resigned after the opening. Another point is that there is no clear way to specify when the opening ends. Middlegame fights happen during the opening sometimes. And what about those games with no fuseki? Of course fuseki is not synonymous with opening.

I think there's a difference between estimating the score, and knowing the result. I'm sure there are board positions where pros would unanimously agree that one side, black or white, is ahead. But it doesn't mean that it's the end of the game - pros make mistakes, too.

In one of the other study journals, someone recently brought up an old game from 2010 where Lee Sedol died pretty big early in the game. He happened to go on and win it.

Around that time, Myungwan was giving a workshop in the bay area, which I attended. I remembered him commenting about that game. He said he took a look in the early middle game, and it looked like Lee Sedol had lost, so he stopped watching, and went on to do something else. When he came back to see that Lee Sedol had won, he was surprised.

I doubt that Myungwan's estimation that Lee Sedol was behind after dying was inaccurate. Probably Lee's chances of winning were pretty slim. Comebacks can still happen, though, which is why I think there's a difference between estimating the score and knowing the result.

I should have said I doubt many people can predict who will win after the opening, rather than estimate the score. My comment about seeing resignations after the opening still applies.

[yoyoma]

Uberdude how are you picking where the opening ends?

gowan you don't think we can say it's more likely for one player to win the game when he has an opening advantage? Uberdude's examples talked about 55% or 65% chance to win. I don't see how you jumped from that to the disadvangated player should resign?

[Kirby]

I should have said I doubt many people can predict who will win after the opening, rather than estimate the score. My comment about seeing resignations after the opening still applies.

Then, I agree with you.

[mitsun]

How about my game? Who was ahead in the opening?

OK, I will bite. If we call the first 18 moves the opening, I would say W was slightly ahead, mostly because the exchange in the lower right corner was good for W. But the difference is not enough to significantly affect the result at any amateur level. Maybe 52% chance for W to win?

I am not strong enough to say much about the middle game fighting through move 45. However, I am reasonably confident that both sides made mistakes which were large enough to overcome any small opening advantage or disadvantage

The invasion at move 47 looks likely to get B into deep trouble. I think I would have looked for a lighter way of reducing, maybe at Q11. Or possibly a deep invasion with more aji like S15, making miai of R17 and R12. The bad result B got here certainly affected the game outcome more than all previous moves combined. After W60, I would bet 80% on W winning. By W76, that goes to 90%.

[often]

Game 1
nomatter v. thechaboot
at the point of white 26 it's a relatively simple opening with b having a smaller advantage, but black chose a fight and white made his group heavy. because of that decision the opening is over and it's a middle game battle.

Game 2
nomatter v. thechaboot
Opening is pretty much done after move 19. Seems even with it leaning a little towards white because of his moyo. However, white chooses to invade and jumps into a middle game situation afterwards with a running group.


Game 3
greendemon v. kykkk
the middle game started around move 46, at that point white had the advantage but there was a way for B to live with his invasion that could've changed the board situation.


The fact that people are handpicking these exposes this to some sort of bias

Perhaps you should also provide insight as to when you think the opening/middle game transition happens and what the standing is.

[Uberdude]

Please don't get hung up on the precise meaning of 10 points. +/-50% is fine. "A bit but not a lot" is the meaning. Or "a bit more than komi".

[Uberdude]

The fact that people are handpicking these exposes this to some sort of bias

My game selection algorithm was:
1. try to remember the KGS 3d I played a few days ago.
2. look in his game history for the first even game against a 3d

I don't see that adding much bias. That both games were within ten points after the opening was as I expected.

[Bill Spight]

How about my game? Who was ahead in the opening?

My rule of thumb for the end of the opening is around move 50. In this case, I would say move 46. (The fuseki ended, as mitsun indicates, at move 18.)

I think that White is clearly ahead, as Black has made two small groups on the top side, and the one on the left is plainly insecure, and the one on the right has too many stones. White's group is also insecure, so maybe White should have continued the fight a bit longer.

[ez4u]

Consider how many games we will need in order to test the hypothesis if we are talking about an advantage of 60%-40%. Add in the fact that there is no agreement on when the opening ends or what the advantage is. This shapes up to be a rather long thread!