Use of SGF editors and Computer Go programs during games

    Keywords: Question, Culture & History

Quicksilvre: At play-by-email servers (Dragon, etc.) games can last for a very long time; even when they are moving swiftly there can be several minutes between moves. It is quite possible for a player to download the SGF file of a game and loading it onto an SGF editor (like Go Write) or a computer go program that can play out moves. In this manner, someone could play out a sequence of moves, like determining the life-or-death potential of a group.

The ethics of this has been discussed on Dragon:

It's probably been talked about elsewhere as well.

The questions are:

  1. Should players be allowed to analyze a game in progress on an editor (on their own)?
  2. Should players be allowed to analyze a game with a computer program (with the assistance of the computer)?

My feeling is, respectively, yes and no. An SGF editor really only provides a visual reference, while a go playing program adds a third wheel which undermines the game. I'd love to hear what everyone else thinks.

Cheyenne: Personally I believe it is very unfair on two counts.

  • It is unfair to your opponent. You are not representing yourself at your true strength. How would you feel if your opponent pulled out a joseki dictionary in a face to face game with you? Your opponent want's to play against you, not against a computer. Or carried to the extreme what if your opponent asks a strong dan over and starts asking for suggestions?

DrStraw: I'm not sure about this. The opponent is playing an anonymous person on the other end who has some rank. If that rank is achieved by consisteny using a tool as you say then then so what? Their rank may be false, but at least they are playing under the same conditions as the rank was achieved, so what is the big deal? The person is only hurting himself, not you.

  • It is unfair to yourself. What have you gained? Why not just review the game with your opponent afterwards? You are relying on a crutch. It is best to learn from one's mistakes. Go ahead and lose then study the game afterwards to determine how you could have played out the sequence.

As for using a computer to analyze the game. Figure that the computer is only going to be about 8 or 9 kyu in over all strength (basing my assumption on GnuGo's strength on NNGS). In addition From what I understand of feeding pro games into GnuGo is that GnuGo doesn't find the best plays anyways.

So.. to summarize my feelings ... I don't think it is fair. If you do decide to analyze the game (assisted or otherwise) then at least inform your opponent of the fact.

mgoetze: Well, I don't know about DGS, but I often use my laptop to record games at real-life tournaments, and none of my opponents have even bothered to look at my screen so far... I could easily look up joseki with Kombilo, but... it's just not ethical, I wouldn't do it even if I were desperate. OTOH, the temptation to use the score estimator is hard to resist sometimes. ;)


Bill: A couple of comments: First, computer analysis is in general poor, with the exception of the life and death program, Go Tools, which is very strong. However, it may be that moves may be compared by having the computer play itself thousands of times after each move in question. The results might be at a reasonable amateur level. (Later note: Indeed, while early programs that took that Monte Carlo approach were weak, recent advances have made them quite strong, even high dan level on the 9x9.)

Second, go has a history of adjourned games at high levels. Nobody thinks that the players did not analyze the positions during the breaks in play, even playing out variations on the board. They might even enlist the help of other players. In a famous game between Honinbo Shusai and Go Seigen Shusai's winning tesuji is rumored to have been found by Shusai's pupil, Maeda Nobuaki. Even if you do not get help, over the board analysis will up the level of your play.

Many years ago, when I was 3-dan, I played go by mail against Ted Drange, who was then shodan, on several small boards at once. I knew that Ted had played those boards often against strong players and had analyzed them thoroughly. Essentially, I was playing against a book, if not The Book. I was a bit overconfident, and wrote him that I would not analyze any game over the board. He was quite surprised.

I remember lying in bed at night, staring at the ceiling, working through all reasonable variations after move 5 on the 5x7. For that one I showed Ted a new winning line. I lost on the 6x6, however.

If I were playing by mail or email, I would expect my opponent to look up joseki, although I would not do so myself. Not that I am so confident, but it really is not much of an advantage. For one thing, you have to choose the right joseki. For another, it is often wrong to play joseki. I certainly would expect him to play out variations, and I would do so myself.


Dieter: There are even arguments in favour of getting help during a game. You can remove the ethical question by explicitly permitting this. Here's an excerpt of an interview I took from one of Europe's best players:

How to become stronger? (...) Secondly, you can’t do it alone. Only together you can become stronger. In the middle of the eighties we had a two weeks tournament in our country. The games were adjourned. During the break, you could analyse the games and it was not forbidden to receive help from other (stronger) players. Once I played such a game with Z., a 5 dan from P., and it was a tough one. The adjourn?ed position was pretty equal. I knew that Z., L. and B. used to analyze in team so I had to put up some tough resistance. A joint analysis with D. and S. showed that I would win by half a point, because in the last ko fight I’d have one ko threat more. I have come to understand yose much better after that game !


axd: About question #1: asking whether the use of editors (or even any other material) during turn-based playing is allowed or not makes no sense, as it cannot be enforced anyway. Edit added: if someone systematically uses an SGF editor, his/her rating will reflect the combined man-software system. If he/she sticks to systematically using the SGF editor, this will be totally transparent to online opponents.

Concerning question #2, I believe the next (inevitable) step is software that will assist during on-line game playing and effectively act as an extension to the human mind (computerassistedgo?). Separate tournaments will be organised to fight such cybergames. Welcome to a new form of playing.


Chris Hayashida: Here are my thoughts:

Should players be allowed to analyze a game in progress on an editor (on their own)?

I really don't think it matters. It's a bad habit, that's to be sure, since it will limit the growth of one's reading. I think it is always better to visualize the moves and work things out in your head. Even if you lose the game, you will get stronger. If you rely on "playing it out" you may end up winning more, but may be handicapped by weak reading.

I do record my game on my Palm, but I find that I don't play as strong because of the distraction of recording. However, I also play slower (because of having to input the moves) so it also keeps me from playing too quickly. I think the benefit of having the record outweighs the cost of the distraction of recording it. I don't, however, play out any variations on the Palm.

Should players be allowed to analyze a game with a computer program (with the assistance of the computer)?

No. In a practical sense, right now it doesn't matter much, as computers are not very strong. However, that doesn't mean that it should be allowed. I believe that computers and algorithms will continue to be stronger, and I believe that it could make a difference.

axd: I'm not sure if there is a consensus on what's understood by "analysing using a computer program", but by this I - personally - understand the assistance a computer can bring by supporting variations, performing counting, keep track of group statii; the human still has to generate and explore the variations, assign values to positions, decide where to go deeper or discard, decide on group status, etc. But the problem is more fundamental: there is no way to avoid computer assistance on Go servers (especially DGS), so does it make sense to interdict it? I want to go even a step further: the time will come when players will actively use computer assistance on live servers such as KGS.


Gul : My thoughts.

About question 1: I think in non real time games it should be allowed to review the game yourself. Doesn't really matter if you are using an editor or a board. You will gain a lot of strength studying certain positions.

About question 2: Doesn't really matter at the moment, but i think it limits your strength. Why are you playing at all if the computer plays for you?


GoJaC: Just to add my two cents', having lately started thinking about this because an opponent on DGS encouraged me to explore variations with an sgf editor... If we were to play anonymously on DGS under some nickname, then it does not matter who you are, if you are a group of 5 players playing together, so be it. Together you form the virtual entity that is being played against. Similarly if you happen to be part of a human-computer duo, what difference does it make?

In my case, I prefer my online identity on Go servers to be directly representing me, if someone challenges me, I consider it a personal challenge and want to beat them (or lose against them) with my own strength. I think you can compare this with Hikaru's situation in Hikaru No Go - he could choose to have Sai play, or Sai help him out, if he wanted to. Would there be anything wrong with him deciding he will be Sai's representative throughout his life?

(Sure, I prefer honesty, but things get a little technical there, if Sai is inhabitting him, he's a part of him, isn't he? Who's brain is doing the thinking? Who's neurons fire? Taking this discussion further might require a Use of Sai during games? humour page... <grin>)


Bildstein: I'm over this whole debate. I can sort of understand people who are below the level of computer play having strong feelings about it, but, because I know there is always someone stronger tham me, and there's always someone stronger than you, and because there is literally nothing riding on the game, none of this seems to matter.

For example, in DragonTourney2005, all games are even games, so when someone enters at an average rank, but ends up being much stronger and winning all their games, that's great. It just means their rating is wrong. And if they're a team of 5 players playing in collusion and analysing their games in great depth, so what? And if they win the tournament, so what? And even if it were a handicap tournament, so what? The only reason I could possibly have to be unhappy with them in that case is that they implicitly agreed to have a close game with me by entering a handicap tournament, and I will not get that because they will handily beat me by being stronger than they originally said. But that's still not a big deal.

In my opinion, the only valid reasons for people being unhappy about someone cheating in these ways are that they are concerned enough with their egos that they feel ripped off that they've lost an opportunity to make themselvses look good, or when the players have wagered something on the game.

As for whether or not these sorts of cheating are good for your game, it's a debatable point. My opinion is that it can be, and that it has for me, but I think it's up to everyone to make up their own mind.

GoJaC: Well said, you were much better at getting the point across. ^_^ (Bildstein: Thanks, GoJaC)


RiffRaff: Adding my two cents here, since I've been playing a bit on DGS lately. I've essentially used the same [ext] rule that the [ext] U.S. Chess Federation has for Correspondence Chess:

You may consult chess books and periodicals but not other players. You cannot use a computer or computer program to evaluate a game but you may use computers for record keeping.

I've personally been using these games as an opportunity to study particular Fuseki and Joseki in more detail than I would otherwise, since what I'm studying has immediate relevance in the game I'm playing. When my opponent deployed the low Chinese opening for instance, I pulled out The Chinese Opening by Kata Masao and read a bit on common themes and strategies in this opening. Since I've been playing Sanren Sei a lot with black, I've pulled out The Power of the Star-Point by Takagawa Shukaku several times to research some of the more common responses to various approach moves. I've also referred to Kogo's Joseki Dictionary for ideas, as well as Sensei's Library itself regarding oddities like the 3-4 7-3 Enclosure.

Basically I don't think there's anything wrong with doing all the research and analysis you like (with or without a SGF editor). I'm about as likely to make mistakes doing this (e.g. choosing the wrong Joseki for the global context) as I would be otherwise. Using a Go playing program to suggest moves though just feels wrong. Even if it would probably result in an easier game for me if my opponents did so.

DanSchmidt: I come from a chess background so for me it is natural to use the same rules as correspondence chess. I was surprised to find that go players do not necessarily approach turn-based games the same way.

Johnny5: I'll admit, there are times when playing on KGS that I'll save the game to disk in the middle of the game and try to play out a difficult situation. The other thing that I have found though is that I tend to lose those games anyway. It surprises me, but it seems like the majority of the time, if I'll actually force myself to read out the situation on the board, I end up making a good play or even discovering a great move I hadn't even noticed at first. I have the same jinx with using the SE during games, so I've been trying harder to get strong at estimating "score" at various points during the game.

Masamune: I'd compare it to pro games that have been played over several months times in the past. The players studied the board position and tried different paths. My conclusion is that it's ok in turn based games like on Dragon Go to use reference material as well as trying to play out paths. In quick games like on KGS I consider it cheating to play out sequences, use the score estimator or a joseki book. Counting and reading can be exhausting and cost time. Discipline, imagination and endurance are all part of the game.

Jedd: Although I've never used an SGF editor to play out a game in progress, when I was a beginner I would sometimes use a score estimator mid-game. If I was thinking about resigning because my opponent looked to be ahead, I would double-check to make sure it was a wide enough margin to justify not continuing play. I've since stopped this practice since in no way does it help me improve, but as a beginner it was a way to feel a little more confident about resigning, which is of course good etiquette when an opponent is dozens of stones ahead. What are people's thoughts on the ethics of doing this?


Use of SGF editors and Computer Go programs during games last edited by 50.23.115.116 on January 27, 2015 - 05:29
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library