Timeline for Analysis of komi values for increasing Go board sizes and agents strength
Current License: CC BY-SA 3.0
21 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 12, 2022 at 9:17 | comment | added | Stef | Based on my previous comment, you would need to proceed by iteration. If on some board size with 0 komi you find that Black wins by n points in average, then set the komi to n then play many self-play games again; now you might find that Black still wins, but by a smaller margin, so increase the komi by that margin; then start again. Continue to repeat until komi converges. | |
| Aug 12, 2022 at 9:16 | comment | added | Stef | If I understand your method, it goes like this: 1) Agent plays many self-play games with 0 komi. 2) Note by how much Black wins on average. 3) Use this average value as komi. However, this method assumes that Black would play the same with 0 komi as with komi. This is certainly not true, as we tend to play "safer" moves when we're ahead in points, and "more risky" moves when we're behind. That is, most agents, human or computer, consider that the "best" move is not necessarily the move that maximises our expected win margin, but rather the move that maximises our confidence in winning. | |
| Dec 6, 2017 at 20:12 | vote | accept | Adam Stelmaszczyk | ||
| Nov 16, 2017 at 8:45 | answer | added | kubus | timeline score: 3 | |
| Oct 26, 2017 at 15:08 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 218 characters in body
|
| Oct 26, 2017 at 15:02 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
deleted 29 characters in body
|
| Oct 26, 2017 at 14:47 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 316 characters in body
|
| Sep 19, 2017 at 13:49 | history | tweeted | twitter.com/StackCompSci/status/910138678780895233 | ||
| Sep 19, 2017 at 12:57 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 191 characters in body
|
| Sep 19, 2017 at 12:45 | comment | added | Adam Stelmaszczyk | Edited, now I use the word "fair", but linked the definition. Thanks for the comments, it probably made the post clearer. | |
| Sep 19, 2017 at 12:44 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 16 characters in body
|
| Sep 19, 2017 at 12:38 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 82 characters in body
|
| Sep 19, 2017 at 12:15 | comment | added | Adam Stelmaszczyk | @Evil Related to what I wrote above. I didn't use word "fairness" in my question, since humans are out of that question. It's just like one is given a game as in game theory, strictly defined as a set of states and actions and you have to e.g. find Nash equilibrium. No fairness, if people like to use it etc. So, the board games stack exchange is the opposite where I see this question belonging too. Thanks for the first link. They used optimal agents, that's why size of the boards was so small. I did some googling before, I proposed something a bit different, computationally tractable. | |
| Sep 19, 2017 at 12:01 | comment | added | Adam Stelmaszczyk | @DavidRicherby It indeed is a question about Go, I haven't found such tag. In this question I view Go as an abstract strategy game and game theory is under mathematics, which seems to overlap with theoretical computer science domain, e.g. because of the methods used (game trees, neural nets, reinforcement learning etc). I think I see what you mean that an answer would have "bad Go" - as in komi from artificial agents (especially optimal ones) might be bad for humans (make them unhappy). I agree. i just thought about such analysis of abstract game. Doesn't have to do anything with humans play. | |
| Sep 19, 2017 at 11:43 | comment | added | Adam Stelmaszczyk | @YuvalFilmus Yes, one can take human data, subtract used komis and see what were the average scores for first player. However this seems much more difficult setup to draw any conclusions from. More noisy. Because every agent (human) was different. Humans also learn over time, get tired etc. Also the number of human games is limited. With agents, one can have exactly the same players and let them play as many games as desired. | |
| Sep 19, 2017 at 10:47 | comment | added | David Richerby | To me, this seems like more of a question about Go than about computer science. For example, it seems much more likely that an answer would be wrong because it contained bad Go rather than bad computer science. | |
| Sep 19, 2017 at 10:45 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 51 characters in body
|
| Sep 19, 2017 at 10:36 | history | edited | Adam Stelmaszczyk | CC BY-SA 3.0 |
added 95 characters in body
|
| Sep 19, 2017 at 10:24 | comment | added | Yuval Filmus | This is roughly how Komi values have been determined, with computer players replaced by human players. | |
| Sep 19, 2017 at 10:07 | review | First posts | |||
| Sep 19, 2017 at 10:24 | |||||
| Sep 19, 2017 at 10:06 | history | asked | Adam Stelmaszczyk | CC BY-SA 3.0 |