Life In 19x19

hyperpape wrote:Actually, 100 EGF is 20 kyu. AGA ranks go past 30 kyu. The systems aren't perfectly calibrated, but they're close enough that it's clear that there's a range of 20 kyus that the EGF lumps together in a single category of "players with rating 100".

An AGA 30 kyu is roughly an EGF -1000.

You can quibble about the exact numbers, but it does seem that Go has slightly more range.

Sure. Where I think it makes sense is that I don't think player strength is anywhere stable enough to talk about GoR ratings under 100, or even around 100 really. Talking about ranks under 20k is usually fairly pointless. I mean, if I grabbed an AGA 22k and a 25k would you be happy betting money on the 22k winning a 2 stone game? How about a 5 dan and a 2 dan?

Boidhre wrote:Sure. Where I think it makes sense is that I don't think player strength is anywhere stable enough to talk about GoR ratings under 100, or even around 100 really. Talking about ranks under 20k is usually fairly pointless. I mean, if I grabbed an AGA 22k and a 25k would you be happy betting money on the 22k winning a 2 stone game? How about a 5 dan and a 2 dan?

Yeah, but the same is true of chess ratings under 1000.

HermanHiddema wrote:

Boidhre wrote:Sure. Where I think it makes sense is that I don't think player strength is anywhere stable enough to talk about GoR ratings under 100, or even around 100 really. Talking about ranks under 20k is usually fairly pointless. I mean, if I grabbed an AGA 22k and a 25k would you be happy betting money on the 22k winning a 2 stone game? How about a 5 dan and a 2 dan?

Yeah, but the same is true of chess ratings under 1000.

Exactly. This is why I think it's pretty pointless to go Higher Bound - Lower Bound = Skill Gap.

Boidhre wrote:

HermanHiddema wrote:Yeah, but the same is true of chess ratings under 1000.

Exactly. This is why I think it's pretty pointless to go Higher Bound - Lower Bound = Skill Gap.

Just because you can't identify a clear lower bound for either of them, doesn't mean you can't be fairly sure that there's a difference in overall bound difference or magnitude.

Not that this measure necessarily means very much, but I think chess and go probably really are quantifiably different under it, with go having a few more levels - I've thought about it before (with some of the same objections), and I think Go can be reasonably said to have at least a small number more of these statistical grading levels, though the original example (2900 -> -1000?) seems to make rather too much of the poorly defined beginner levels that may not fit well in the model.

I'll stress that I'm not saying that this makes Go harder than chess in any useful sense. Even if we wanted to apply it to real life, another vital number for judging difficulty levels is how quickly people tend to move between them. It doesn't matter how many of these levels statistically exist if Go players tend to move up them much more easily, for instance.

Edit: Oh, and another interesting observation is that chess ais have pushed the grading system up by (I think) a few hundred points! Depending on what you make of that, it would rather close the gap with Go, though of course we don't know how many stones go bots will be giving top professionals in 20 years

amnal wrote:

Boidhre wrote:

HermanHiddema wrote:Yeah, but the same is true of chess ratings under 1000.

Exactly. This is why I think it's pretty pointless to go Higher Bound - Lower Bound = Skill Gap.

Just because you can't identify a clear lower bound for either of them, doesn't mean you can't be fairly sure that there's a difference in overall bound difference or magnitude.

Not that this measure necessarily means very much, but I think chess and go probably really are quantifiably different under it, with go having a few more levels - I've thought about it before (with some of the same objections), and I think Go can be reasonably said to have at least a small number more of these statistical grading levels, though the original example (2900 -> -1000?) seems to make rather too much of the poorly defined beginner levels that may not fit well in the model.

I'll stress that I'm not saying that this makes Go harder than chess in any useful sense. Even if we wanted to apply it to real life, another vital number for judging difficulty levels is how quickly people tend to move between them. It doesn't matter how many of these levels statistically exist if Go players tend to move up them much more easily, for instance.

Edit: Oh, and another interesting observation is that chess ais have pushed the grading system up by (I think) a few hundred points! Depending on what you make of that, it would rather close the gap with Go, though of course we don't know how many stones go bots will be giving top professionals in 20 years

I think the issue is (as you mention), say there are X grades of winning level in a game (each rank is 60% apart of whatever) and X-1 grades in a different game. If it takes much longer to gain a grade in the latter game (i.e. more effort, knowledge, training required etc) do we consider there to be more or less of a skill gap between beginner and supreme expert in it over the other? Is skill gap measured by winning percentages or by the amount of effort required to become an elite player? I'm not convinced a bigger "ELO space" tells us much useful information about how hard a game is to master at all.

Boidhre wrote:I think the issue is (as you mention), say there are X grades of winning level in a game (each rank is 60% apart of whatever) and X-1 grades in a different game. If it takes much longer to gain a grade in the latter game (i.e. more effort, knowledge, training required etc) do we consider there to be more or less of a skill gap between beginner and supreme expert in it over the other? Is skill gap measured by winning percentages or by the amount of effort required to become an elite player? I'm not convinced a bigger "ELO space" tells us much useful information about how hard a game is to master at all.

Yes, we can easily simulate this, actually.

I just invented a new game called 3-go. To win the game, you play 3 games of go against someone, and have to win at least 2 of them. Simple probabilistic calculation shows that if you have a 60% chance of winning a game of go against someone, then in 3-go you have a 65% chance of winning against that person. This new game of 3-go therefore has a wider Elo range. But of course, 3-go is not harder to learn than go.

Boidhre wrote:

hyperpape wrote:Actually, 100 EGF is 20 kyu. AGA ranks go past 30 kyu. The systems aren't perfectly calibrated, but they're close enough that it's clear that there's a range of 20 kyus that the EGF lumps together in a single category of "players with rating 100".

An AGA 30 kyu is roughly an EGF -1000.

You can quibble about the exact numbers, but it does seem that Go has slightly more range.

Sure. Where I think it makes sense is that I don't think player strength is anywhere stable enough to talk about GoR ratings under 100, or even around 100 really. Talking about ranks under 20k is usually fairly pointless. I mean, if I grabbed an AGA 22k and a 25k would you be happy betting money on the 22k winning a 2 stone game? How about a 5 dan and a 2 dan?

Yes, I would be, so long as the betting amount was small enough that I would not have an unacceptable risk--I believe the bet has a slight positive expectation in both cases, though I do realize it has a higher one with the 5dan and 2dan. I would bet $20 without much thought, because I can lose that several times without severe inconvenience. I would not bet $500.

HermanHiddema wrote:

Boidhre wrote:I think the issue is (as you mention), say there are X grades of winning level in a game (each rank is 60% apart of whatever) and X-1 grades in a different game. If it takes much longer to gain a grade in the latter game (i.e. more effort, knowledge, training required etc) do we consider there to be more or less of a skill gap between beginner and supreme expert in it over the other? Is skill gap measured by winning percentages or by the amount of effort required to become an elite player? I'm not convinced a bigger "ELO space" tells us much useful information about how hard a game is to master at all.

I just caught this: I'm not concerned to argue that go is harder to master. Indeed, the empirical evidence suggests that go and chess are very close on that metric.

I should state that I consider both chess and go more than deep enough for humans--though Chess does have a problem with opening books. I do not regard the depth that I'm arguing exists in go as much of an argument for it being a better game.

hyperpape wrote:I just caught this: I'm not concerned to argue that go is harder to master. Indeed, the empirical evidence suggests that go and chess are very close on that metric.

I should state that I consider both chess and go more than deep enough for humans--though Chess does have a problem with opening books. I do not regard the depth that I'm arguing exists in go as much of an argument for it being a better game.

Mostly my concern is the talk about the ELO ranges being different meaning something.

Maybe the different skill gaps can be pinpointed by way of the game mechanics: Chess and go are both essentially "race games". Meaning the player who makes the most efficient moves wins. Meaning the player who makes less mistakes wins.

A normal game of go lasts about 100 moves longer than a game of chess. Therefor 100 opportunities more to make mistakes. Therefor the gap between the worst possible play and the best possible play has to be greater.

Does that make any sense?

At the risk of un-derailing this thread. Has anyone played any of the GIPF Project series of games?

Three of the six games available are listed above go in the Board Game Geek rankings of Abstract Games.

Yinsh, Tzaar, Dvonn are the three highly-rated GIPF games, while Zertz, GIPF, Punct, Tamsk (replaced by Tzaar) round out the remaining GIPF games.

msgreg wrote:At the risk of un-derailing this thread. Has anyone played any of the GIPF Project series of games?

Three of the six games available are listed above go in the Board Game Geek rankings of Abstract Games.

Yinsh, Tzaar, Dvonn are the three highly-rated GIPF games, while Zertz, GIPF, Punct, Tamsk (replaced by Tzaar) round out the remaining GIPF games.

Yes, several of them. None of them appealed to me much but to be fair I didn't give them much of a go and it was a good number of years ago.

HermanHiddema wrote:I just invented a new game called 3-go ...
3-go is not harder to learn than go.

A propos 3 and Go: Some days ago I invented some Go variation. At first it was just a funny idea.
To test it, I tried it out on the smallest non-trivial board size 3x3.
Some players on KGS enjoyed to participate in my test.

I call it "3x3 Inverse-Go" or (rhyming with 3x3 tic-tac-toe) GigaGo.

You play with normal Go rules, but passing not allowed.
The first player unable to make a legal move ... WINS!

(To try this seems natural for any reader of "Winning Ways" who has thought along the lines of The last one who is able to play is the winner.
So I wonder who invented this before me. And how many )

KGS-player Alast proposed to restrict the number of moves allowed, in order to keep the players from playing forever by avoiding any risk.
This is an interesting addon which I refined a bit so as to balance the burden:
every time the board is "flipped" by taking the last liberty on the board, the number of moves is increased by one. Worked quite well with inital number of 30 moves.

Playing on 4x4 and larger is likewise possible, but it is astonishing how interesting 3x3 can be. Some people loved it and others despised it

Cheers,
Rainer
(GoChild GoRo with 1891645 points)

I love go but backgammon is far more enjoyable. IMO. Every week I have a game night. We usually play 1 game of go and 4-5 games of backgammon. There are dozens of variants for endless fun. Chess used to be on the agenda until I discovered go.

http://aagenielsen.dk/index_nef2.html Hnefatafl is another very old game that I recently learned about. It looks interesting. One thing I like about it is the two sides begin with different strength 'armies'.

msgreg wrote:At the risk of un-derailing this thread. Has anyone played any of the GIPF Project series of games?

Three of the six games available are listed above go in the Board Game Geek rankings of Abstract Games.

Yinsh, Tzaar, Dvonn are the three highly-rated GIPF games, while Zertz, GIPF, Punct, Tamsk (replaced by Tzaar) round out the remaining GIPF games.

So I double-checked this because I had seen that Go was hanging on in the top 50 of BGG (at #48!) whereas the other games aren't. And what msgreg says is true - in the abstract games list, the gipf games are ranked higher, whereas in the games list, go is ranked higher. Looking more closely, the average user rating for go is, 7.78/10, which is higher than any of the gipf games ("tzaar" comes closest, at 7.69). However, apparently this rating is not the rating used to rank games;

The BGG Rating is based on the Average Rating, but the number is altered. BoardGameGeek's ranking charts are ordered using the BGG Rating. To prevent games with relatively few votes climbing to the top of the BGG Ranks, artificial "dummy" votes are added to the User Ratings. These votes are currently thought to be 100 votes equal to the mid range of the voting scale: 5.5, but the actual algorithm is kept secret to avoid manipulation. The effect of adding these dummy votes is to pull BGG Ratings toward the mid range. Games with a large number of votes see their BGG Rating alter very little from their Average Rating, but games with relatively few user ratings will see their BGG Rating move considerably toward 5.5. This is known as "Bayesian averaging" and a quick search of both BGG and/or the Web will reveal much discussion on the topic.
In effect, usually the games with many votes will Rank higher than those games with the same Average Rating but fewer votes.

This doesn't really explain anything, though, because Go (obviously) has far more ratings than the other abstract games. (It also has more ratings than some of the games that beat it due to re-adjustment on the main chart.) Further, this doesn't explain why on the abstract list the rank is based on a "BGG rating" of 7.370 and on the general list the rank is based on a "BGG rating" of 7.544.

Anyway, there are few things that could be less interesting than the arcana of how a website rejiggers its rankings to keep them interesting and the resulting discrepancies, but I looked into it anyway, so there's your answer.