Life In 19x19 http://prod.lifein19x19.com/ |
|
Testing Theories Against GnuGo http://prod.lifein19x19.com/viewtopic.php?f=10&t=3429 |
Page 1 of 1 |
Author: | hailthorn011 [ Mon Mar 14, 2011 2:21 pm ] |
Post subject: | Testing Theories Against GnuGo |
Hello, it's me again with all of my Kyu-ish ignorance. ![]() I like testing new things that I learn, so I've decided to start practicing newly learned techniques against Gobots before utilizing them in games against humans. It's something I plan to start rather than something I already do. Do y'all think there is any value in this, or is it simply a waste of time? ![]() |
Author: | BaghwanB [ Mon Mar 14, 2011 2:25 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
Eh... Without having tried it myself, I'd say it might be OK for a few concepts/aspects but horrible for others. The trick (which I can't do) is tell which is which. I'd guesstimate that it could be OK on close in 3rd/4th line fighting maneuvers but the big strategic ones could have no real relevance to playing humans. I'll gladly defer though to anyone who has actually tried it out though. A while back I tried playing more 9x9 against computers to get more L&D and fighting experience and I think it helped a bit, but that's a pretty limited sample set. Bruce "Single Digit Board" Young |
Author: | topazg [ Mon Mar 14, 2011 2:28 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
I would say practicing techniques against bots is probably the worst practice you can use bots for. Bots are reasonably well known for having all sorts of blind spots, and techniques may consistently work 100% against a bot even if they are technically not correct or well played out. Playing bots I consider fine, practicing techniques requires humans so that you can handle various spurious responses to your sequences. Just my 2p/2c ![]() |
Author: | hailthorn011 [ Mon Mar 14, 2011 3:10 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
BaghwanB wrote: Eh... Without having tried it myself, I'd say it might be OK for a few concepts/aspects but horrible for others. The trick (which I can't do) is tell which is which. I'd guesstimate that it could be OK on close in 3rd/4th line fighting maneuvers but the big strategic ones could have no real relevance to playing humans. I'll gladly defer though to anyone who has actually tried it out though. A while back I tried playing more 9x9 against computers to get more L&D and fighting experience and I think it helped a bit, but that's a pretty limited sample set. Bruce "Single Digit Board" Young Yeah, I used to do up to around 50 9x9 games a day with Gobots. I think it helped somewhat. And I can agree with you, there are some areas where it'd probably be weak. One thing I've noticed is that humans don't respond like Gobots....lol |
Author: | hailthorn011 [ Mon Mar 14, 2011 3:12 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
topazg wrote: I would say practicing techniques against bots is probably the worst practice you can use bots for. Bots are reasonably well known for having all sorts of blind spots, and techniques may consistently work 100% against a bot even if they are technically not correct or well played out. Playing bots I consider fine, practicing techniques requires humans so that you can handle various spurious responses to your sequences. Just my 2p/2c ![]() I understand that entirely. I played 9x9 Go with a Gobot so much that I could predict with high accuracy how the bot would respond to my moves. But really what I do is test out fuseki strategies that I learn mainly because I feel like that's my weakest area by far. Or that and everything else considering I'm only 11k.... |
Author: | Chew Terr [ Mon Mar 14, 2011 3:23 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
One thing I tried once is to agree with a friend to play the first 30 (or whatever) moves of the opening, and then review together, over and over. That kind of let us mess around with different ideas, and see if any obviously failed or worked well. While it can be intimidating to practice with people, working with a friend can take some of the pressure off. |
Author: | hailthorn011 [ Mon Mar 14, 2011 3:48 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
Chew Terr wrote: One thing I tried once is to agree with a friend to play the first 30 (or whatever) moves of the opening, and then review together, over and over. That kind of let us mess around with different ideas, and see if any obviously failed or worked well. While it can be intimidating to practice with people, working with a friend can take some of the pressure off. That would be great if I had any friends who played Go. ![]() |
Author: | LocoRon [ Mon Mar 14, 2011 4:25 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
Have you tried visiting the KGS Teaching Ladder? I'm sure you could get some help in there, if you have a game where you were experimenting with an idea, and would like someone else's perspective.... (Or, you know, just get some good ol' teaching games) |
Author: | Dusk Eagle [ Mon Mar 14, 2011 7:25 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
I have to agree with topazg - doing this is more likely to reinforce bad habits than good ones. Have you seen this thread on how to beat GNUGo? It's a stupid strategy that works perfectly against that bot. I think having a teacher is the easiest way to improve. But if you don't know anyone stronger and don't want to pay for lessons, just frequent studying and trying out what you learn against humans is a good way to learn to. Also, don't give up on some concept you've learned if you can't make it work right away. It's easy to be told something like "Don't attach to weak stones," or "Use thickness to attack," but it takes a while before you can see first-hand the truth of these concepts. |
Author: | hailthorn011 [ Mon Mar 14, 2011 7:40 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
Dusk Eagle wrote: I have to agree with topazg - doing this is more likely to reinforce bad habits than good ones. Have you seen this thread on how to beat GNUGo? It's a stupid strategy that works perfectly against that bot. I think having a teacher is the easiest way to improve. But if you don't know anyone stronger and don't want to pay for lessons, just frequent studying and trying out what you learn against humans is a good way to learn to. Also, don't give up on some concept you've learned if you can't make it work right away. It's easy to be told something like "Don't attach to weak stones," or "Use thickness to attack," but it takes a while before you can see first-hand the truth of these concepts. I had no idea about that link. << *plots torture for the Gobots* (jk) But in all seriousness, I know. See, it helped at one point because I had a friend I could bounce off of. With his help I went from 16k to 12k in just a week. That's the fastest development I've ever experienced, so I know having a teacher/friend helps. Unfortunately, he stopped coming on as much. << |
Author: | hailthorn011 [ Mon Mar 14, 2011 7:41 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
LocoRon wrote: Have you tried visiting the KGS Teaching Ladder? I'm sure you could get some help in there, if you have a game where you were experimenting with an idea, and would like someone else's perspective.... (Or, you know, just get some good ol' teaching games) Not yet, but I'll have to look into that. It's been mentioned to me several times now. |
Author: | Chew Terr [ Tue Mar 15, 2011 8:49 pm ] |
Post subject: | Re: Testing Theories Against GnuGo |
hailthorn011 wrote: Chew Terr wrote: One thing I tried once is to agree with a friend to play the first 30 (or whatever) moves of the opening, and then review together, over and over. That kind of let us mess around with different ideas, and see if any obviously failed or worked well. While it can be intimidating to practice with people, working with a friend can take some of the pressure off. That would be great if I had any friends who played Go. ![]() Worth asking here for possible 10-15k practice partners? Wasn't someone (besides topazg) asking for something similar not too long ago? Sorry, I try to help, but my memory is abyssmal. |
Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |