|
Overwined posted:Out of curiosity, how often do you high sdk and dan players review your own games? I check every game and try to find something I did wrong and how to fix it, then try to apply that in the next game that gives me the opportunity to do so. If it's just one little thing that's still some opportunity to improve but I mostly just want to play.
|
#
?
Jun 26, 2017 13:53
|
|
|
|
| # ? May 23, 2024 21:34 |
|
|
Does anyone play on tygem? I'm thinking of getting on there, because A) you can play for money and B) it seems to be where people such as students from Myongji University's department of baduk studies plays and I am just sort of into that. Also check out this department of baduk studies where you can get a literal masters or doctoral degree in loving go. I'm so jealous. http://www.mjubadukstudies.com/en/
|
|
#
?
Jul 4, 2017 21:03
|
|
|
empty whippet box posted:Looking at game 9 that they just released, move 52-56 - why? this seems like a genuine boneheaded mistake on black alphago's part. It just made me
|
|
#
?
Jul 4, 2017 22:22
|
|
and I can't find a way to make it not seem like a bad series of moves.
|
Bisse posted:Definitely a mistake, question is why... maybe the two AI's are putting so much pressure on each other that they are more likely to make mistakes? AlphaGo does seem to have a tendency to take risks when it can't see a clear path to victory. So many hanes.
|
|
#
?
Jul 4, 2017 23:07
|
|
|
I don't think a human being will ever defeat alphago ever again. However, I don't think that means that mankind has been 'defeated' because I don't believe a human being could play like alphago. I don't think it's possible. So, while we'll learn a lot from alphago, I don't think it's more than a great study tool for those that have reached the outer limits of human go technique to go further.
|
|
#
?
Jul 5, 2017 00:49
|
|
|
Bisse posted:Definitely a mistake, question is why... maybe the two AI's are putting so much pressure on each other that they are more likely to make mistakes? AlphaGo does seem to have a tendency to take risks when it can't see a clear path to victory.
|
|
#
?
Jul 5, 2017 01:58
|
|
|
Shadonra posted:Look at the result of that sequence. Black's left group started dead, is still dead, but White spent an extra move killing it - Black played 2 moves there and White answered 3 times. So Black got an extra move in the bottom right, at the cost of making White thicker. It's an exchange, not at all a clear mistake. 
|
|
#
?
Jul 6, 2017 04:29
|
|
|
empty whippet box posted:Does anyone play on tygem? I'm thinking of getting on there, because A) you can play for money and B) it seems to be where people such as students from Myongji University's department of baduk studies plays and I am just sort of into that. Also check out this department of baduk studies where you can get a literal masters or doctoral degree in loving go. I'm so jealous. http://www.mjubadukstudies.com/en/ let me know if you figure out how to get coins or bet with them. it has never worked for me.
|
|
#
?
Jul 6, 2017 23:34
|
|
|
empty whippet box posted:I don't think a human being will ever defeat alphago ever again. However, I don't think that means that mankind has been 'defeated' because I don't believe a human being could play like alphago. I don't think it's possible. So, while we'll learn a lot from alphago, I don't think it's more than a great study tool for those that have reached the outer limits of human go technique to go further. A human being will never defeat alphago again because Google Team dismantled and put it out of service.
|
|
#
?
Jul 8, 2017 03:15
|
|
|
Reached 8k on pandanet  I have about 85 more wins than losses, too....movin' on up!! Fuckin' love this game.
|
|
#
?
Jul 10, 2017 16:59
|
|
|
AdorableStar posted:A human being will never defeat alphago again because Google Team dismantled and put it out of service. Humans win again. 
|
|
#
?
Jul 10, 2017 21:15
|
|
|
I mean that's the AI equivalent of flipping the board and beheading your opponent because he beat you so soundly.
|
|
#
?
Jul 10, 2017 21:46
|
|
|
Overwined posted:I mean that's the AI equivalent of flipping the board and beheading your opponent because he beat you so soundly. Sounds extremely historically accurate to me.
|
|
#
?
Jul 10, 2017 21:48
|
|
|
silvergoose posted:Sounds extremely historically accurate to me. It's perfect. But then again, maybe after a mutli-million self-game session, AlpaGo put up on one of its human interface terminals "Shall we play a game?" And one of the options was "Global Thermo-Nuclear Go" so someone at DeepMind said "gently caress No" and just unplugged it.
|
|
#
?
Jul 10, 2017 22:08
|
|
|
Considering how much better than us alphago really is, I can only assume that this is part of its plan.
|
|
#
?
Jul 11, 2017 19:35
|
|
|
I’ve been programming an iPhone board game for the past week and it really made me want to get into Go for some reason. Oh well, time to download KGS. I’m also on Chapter 4 of Integrated Chinese so I’m practically a scholar-gentleman already.
|
|
#
?
Jul 14, 2017 08:13
|
|
|
Luigi Thirty posted:I’ve been programming an iPhone board game for the past week and it really made me want to get into Go for some reason. Oh well, time to download KGS. I’m also on Chapter 4 of Integrated Chinese so I’m practically a scholar-gentleman already. Aw hell yes thread overlap. This guy has been programming random poo poo on the most ridiculous of platforms.
|
|
#
?
Jul 14, 2017 15:12
|
|
|
There's something strangely insightful about trying to code the rules of Go, I recommend this to every programmer who also happens to play the game. spoiler: it's not as simple as it looks like
|
|
#
?
Jul 14, 2017 18:32
|
|
|
Symbolic Butt posted:There's something strangely insightful about trying to code the rules of Go, I recommend this to every programmer who also happens to play the game. Yeah I have no idea how I'd program in the rules for determining whether something is alive, without getting false positives for seki or false eyes. I guess you could maybe say "surround the 'outside' of this shape with enemy stones--is there then any way to fill the interior with enemy stones that kills it?" or something along those lines. I dunno, I'm always very impressed by OGS's ability to pick out living / dead shapes. My guess is that they use some sort of neural network, which gets feedback based on how players mark stones at the end of a game. (Using a neural network is what I'd actually do if I had to program this).
|
|
#
?
Jul 14, 2017 18:58
|
|
|
VikingofRock posted:Yeah I have no idea how I'd program in the rules for determining whether something is alive, without getting false positives for seki or false eyes. I guess you could maybe say "surround the 'outside' of this shape with enemy stones--is there then any way to fill the interior with enemy stones that kills it?" or something along those lines. I dunno, I'm always very impressed by OGS's ability to pick out living / dead shapes. My guess is that they use some sort of neural network, which gets feedback based on how players mark stones at the end of a game. (Using a neural network is what I'd actually do if I had to program this). This seems too fancy. If I remember right, there's a living-group detection algorithm on Sensei's Library somewhere.
|
|
#
?
Jul 14, 2017 23:47
|
|
|
Symbolic Butt posted:There's something strangely insightful about trying to code the rules of Go, I recommend this to every programmer who also happens to play the game.
|
|
#
?
Jul 15, 2017 00:20
|
|
|
Doubleposting because I forgot about this great example featured in the book Mathematical Go: "It is a rare example of a Go position which is apparently not yet resolved, but in which either player does better if [their] opponent plays first." Some Japanese judges once judged this to be 3 points for black, and that agrees with every other ruleset. But then in the 1989 Japanese rules (which I think still apply for professional games in Japan), they say it's seki and worth no points at all. dirby fucked around with this message at 00:49 on Jul 15, 2017 |
|
#
?
Jul 15, 2017 00:39
|
|
|
silvergoose posted:Aw hell yes thread overlap. This guy has been programming random poo poo on the most ridiculous of platforms. Just because I can implement Reversi on the Amiga doesn't mean I'm any good at it  e: my KGS username is Luigi30 if someone wants to add me to ITGO! Luigi Thirty fucked around with this message at 02:25 on Jul 15, 2017 |
|
#
?
Jul 15, 2017 02:21
|
|
|
Shadonra posted:This seems too fancy. If I remember right, there's a living-group detection algorithm on Sensei's Library somewhere. Following this post's advice, I looked on sensei's library and found this page. Benson's algorithm is interesting, but seems fairly limited since it can only determine if a player's stones are alive unconditionally, even if that player always passes. It seems like you would want a weaker condition for a real go program. So up next I guess I'll have to read some of these papers. Anyways, thanks for the recommendation!
|
|
#
?
Jul 15, 2017 08:05
|
|
|
You can absolutely code the Chinese ruleset. By extension, AGA shouldn't be hard. But oh boy good luck on the japanese set and addendums. The last time I looked at it they didn't pass the 'conforms to basic rules of logic' muster.
|
|
#
?
Jul 15, 2017 17:10
|
|
|
Hoo boy played some games on IGS and now I know what you mean by losing 100 games quickly
|
|
#
?
Jul 15, 2017 21:11
|
|
|
Don't give up! If you can handle losing a ridiculous number of games before winning one, and you can handle losing games in general after that, then Go is such a rewarding game to play. I find it's very difficult for people who get into it as adults to get over the initial learning curve, though, especially if they don't know anyone IRL to play. I feel like I am only able to improve constantly because I play my dad a couple times a day. I'm 8k, he's 13k, so the rank difference means our games aren't rated because he refuses to use handicaps against me(and he still wins sometimes too,  ). This lack of pressure from the game not being ranked combined with it being my dad makes it easy for me to improvise and try new things. If you find a regular opponent you feel comfortable losing to, you can improve really fast, basically.I have no good explanation for why i lose to him at all though, I assume he is weaponizing his fatherly wisdom against me somehow.
|
|
#
?
Jul 16, 2017 20:04
|
|
|
I don't plan on giving up, I'm just playing against SmartGo until I can actually beat it on a 9x9 board with 4 handicap stones before trying IGS again I'm using IGS since it has an iPhone client.
|
|
#
?
Jul 16, 2017 21:59
|
|
|
Shadonra posted:This seems too fancy. If I remember right, there's a living-group detection algorithm on Sensei's Library somewhere. Just have the players decide it; you're not coding a bot to play.
|
|
#
?
Jul 18, 2017 03:32
|
|
|
I've managed to soundly defeat some 8k+ players now....if I can manage to keep my current pace of advancement I could be sportin' a Dan-level ranking sometime next year! That's honestly on my bucket list, and if I get that far then there will be no prying me away from this game. I'll just obsess until I'm 9D or I die. I wonder if a really skilled player could peg the exact rating of players just by watching a game. Are there particular moves only 8k players make, and particular moves only 7k players make? It's such an abstract game, I feel like someone who is wicked good at one aspect of the game could wind up progressing, but it seems like you can't let any aspect of your playing really lag behind because it's all so intricately connected. It's always amazed me how well the ranking system seems to predict who will win and lose games - a 5k player against a 7k player with two handicaps will win as reliably as a 5k player against a 5k player with no handicaps. empty whippet box fucked around with this message at 10:09 on Jul 25, 2017 |
|
#
?
Jul 25, 2017 10:07
|
|
|
Isn't it like that because it's defined to be that way?
|
|
#
?
Jul 28, 2017 05:08
|
|
|
AdorableStar posted:Isn't it like that because it's defined to be that way? I think they're exclaiming at how accurate the system works. Like, sure, it's designed to work that way, but it can frequently predict how badly you'll lose, not just that you'll lose.
|
|
#
?
Jul 28, 2017 14:59
|
|
|
silvergoose has the main idea, but the details of rating systems can get pretty hairy. I'm simplifying a lot, but: For KGS, the system basically tries to make sure that this is true for an even game: P(A wins) = 1 / ( 1 + exp(RankB-RankA) ) and simply assumes that one stone of handicap should be roughly like one point of rank (and as silvergoose points out, it's surprising how well that happens to work). And for AGA, it's similar in spirit but with even more complicated math.
|
|
#
?
Jul 29, 2017 15:56
|
|
|
It's been a good few days (I'm littleprof on IGS)
|
|
#
?
Jul 29, 2017 20:56
|
|
|
empty whippet box posted:It's been a good few days you probably need to change your rank to ~2 stones stronger if you're winning by 20+ points that consistantly
|
|
#
?
Jul 30, 2017 02:18
|
|
|
derp posted:you probably need to change your rank to ~2 stones stronger if you're winning by 20+ points that consistantly I just got promoted to 7k, in fact. I'm not sure what my ceiling is right now but I've been diligently studying and working to improve for a while now. When I first started playing 8k players I got loving shithoused about 5 times in a row while almost always winning against any 9k, and suddenly in the last few days I seem to have made a bit of a breakthrough and started dunking on 8k players. I've never played a 7k or above player in a non-handicap match before, so I'm expecting to do some losing now, and thus some more leanin'!
|
|
#
?
Jul 30, 2017 03:21
|
|
|
congrats tho feels good to win so much
|
|
#
?
Jul 30, 2017 03:36
|
|
|
I lost by over 100 points to the first 7k player I played...so I don't think I'm gonna try to skip any ranks. e: but it's ok I beat the next one by 1.5 because my performances are hilariously inconsistent. empty whippet box fucked around with this message at 21:00 on Jul 30, 2017 |
|
#
?
Jul 30, 2017 09:36
|
|
|
i got to 3k on ogs, huzzah. i havent even been trying to improve or really thinking much about go but i got better i guess somehow?
|
|
#
?
Jul 31, 2017 06:07
|
|
|
|
| # ? May 23, 2024 21:34 |
|
|
Of course, I lost six in a row after winning ten in a row. My pride. 
|
|
#
?
Jul 31, 2017 06:32
|
|