Venting

I think every time I play bill at Go I lose at least half a stone. Had to say it. It's nerve wracking.

I hope you're not accusing him of playing slow.He's probably the quickest player on LG.
Just consider the amount of games he has to work through each day, currently over 800!
A simple calculation learns that if he takes 10 seconds for each game each day, he spends more than 2 hours on playing.
For every 5 second more you can add 1 hour of playing time. Especially in the game of Go the next move is not always obvious.
I think we owe him respect and patience dealing with this formidable task.

I think he often plays the games out to the point far beyond where they are lost. Especially in Go it takes a lot of time.

I also don't know a lot of players in twixt who never resigns. He stops trying in the end though, and just clicks outside the board.

.. but appearently he is good at reversi. I think he even won the championship.

He's fast (I've seen faster players in StreetSoccer, though) but a game of go can last for a huge number of moves, and when almost everything by your opponent is dead it gets pretty pointless to keep killing things.

To say that Bill has made any significant progress in Havannah would be stretching it, but he always replies promptly and doesn't seem discouraged by the results so far. I have no problems with that.

But Havannah is a relatively fast game. In 19x19 go I'm in moves ~190, I'll still have to move around 60 moves (more or less a short havannah game) to maybe end the game... And most of the remaining moves are killing things that are already dead. Removing stones that are put in atari. Etc etc. Well, enough venting already.

One piece of advice I've heard for playing against an opponent who's prolonging a Go game is not to capture (unless really necessary). It only opens up territory for more pointless moves by the opponent.

@ Ruben

Yes, I agree that a game that has sudden death as an inherent feature, maybe less prone to annoyance about prolongued albeit pointless play.

I think it is rather rude for someone to continue making moves after it becomes quite obvious that a game is lost. I only know bill from hex and hex19, where fortunately lopsided games are over in relatively few moves. Playing someone like this in 19x19 go must be a nightmare. Perhaps if many people used the "Message:" box when making moves to remind bill that it is polite to resign once the position is clearly hopeless, it would have a positive effect.

I agree with MarleysGhost.

There are only few players that are pushing the game to the last move. Replying with useless moves is a useful strategy, because every "normal" player is contributing to making the workload for those "inconvenient" players higher. This strategy should pose negative feedback on them, since the majority of their games are going to be clearly lost games. Maybe this state of dissatisfaction can motivate them to abandon some of their lost games and sign up for new games.

Hm, I read Marleys post once more, and now it seems that we are saying the opposite. Well, whatever... Make up your own mind. :-D

Perhaps the site could show the current score for both sides. If it says you're losing by a score of 150 to 30, you may be more inclined to resign.

In the eidogo player link there used to be a score estimate, but it doesn't seem to work anymore. Which assumptions can a computer make for 'current score'? technically after blacks first move it is something like 360 - 6.5. But I wouldn't resign as white...

Scoring a go game is relatively non-trivial in terms of computation, if I was the site admin I would not add it to keep the bills lower (no pun intended here!)

Maybe I can keep removing groups from the board to test at what point a resignation is possible, so I can counter it later (i.e. definite killing leads to resign? definite not killing leads to resign?)

What pun? Anyway, I saw that the EidoGo guy took out the scorer because it took too much CPU time. But I practice a lot against igowin, and I've never seen it make a scoring mistake. In your example of a single black stone, igowin could tell that neither side has territory, so it would report a score of 0 to 0.

eidogo runs on the user's computer (as a Java applet, IIRC). The problem is not server CPU cycles, but the boost a player could get from comparing estimated scores for several possible moves.

I've advocated automated scoring of Go games after two consecutive passes, as a preferred alternative to the current scoring method, but
1. There probably is no algorithm that can score using Japanese rules if the players pass early enough.
2. There may be some hope if the site switched to Chinese rules, but I don't know for sure that the Chinese rules can be implemented as an unambiguous algorithm, either.

But EidoGo says "Sorry, Score Estimate is temporarily unavailable. It was using excessive server resources. If you have a server with lots of spare CPU cycles, let me know."

AFAIK, the Chinese rules can be implemented perfectly. (The Computer Go Server uses them for this reason.) You can use a flood-fill algorithm to color each of the empty points, and this tells you which points belong to which player. On an n-point board it will run in O(n) time. You may need the player to capture dead stones manually, but at least this does not hurt your score under Chinese rules.

I still prefer a separate scoring phase after two passes, since sometimes a player will miss an opportunity for a little extra territory.

Oh. I stand corrected. I had no idea it communicated with its server.

"You may need the player to capture dead stones manually.."
Hmmm. During discussions of Go games that had a sequence of consecutive Count Scores (between a player who knows what he's doing and a player who doesn't), I have inquired of those who know Go algorithms better than I do whether it's possible to have totally automated scoring.

"..separate scoring phase after two passes.."
Meaning the player should be able to make a move after seeing the output of the automatic scorer, which would be run only after two consecutive passes?

It probably can't be fully automated. There is an algorithm ("Benson's algorithm") that can tell when a group is unconditionally alive. Maybe that can be used to determine live groups, and then stones inside the live groups that are small enough are marked dead. What answers did you get when you asked others?

It looks as if EidoGo is written in JavaScript, actually. I'm guessing the scorer runs in C (or something with comparable speed) on the server, and the result are fed back to the browser. This is like what the double-dummy solver on BridgeBase does. They probably want to avoid doing so much computation in JS.