I'm ignoring the scoring of go software.

[Alberich]

There is simply no reason why I should even bother with the scoring systems in any of the software titles I've seen, whether for the iPad or pc. It's my opinion they're simply wildly inaccurate and incapable of giving reasonable estimates. In my opinion both KGS and Tygem are laughably cartoonish in their estimates while IGS when using Panda iGo does relatively well if you don't forget to manually assess the position using the edit function after the game is over...and I'm referring to the pc version. The only other software that comes reasonably close in good estimates in my opinion is Compogo for Windows which doesn't seem to be supported anymore and that's a real pity.

Given this state of affairs I don't think its possible to get people interested in a game that has no clear scoring mechanism in place to allow reasonable estimates of who really prevails in this activity regardless of ones proficiency in the game. Chess doesn't have this problem because the goal is much clearer...check mating the king and this can be calculated with precision using chess engines. If I have go software that assesses a game where one side wins with 3.5 while another says the other side wins with 285 points then this remains in my opinion a sore spot for this sport and that's a shame because this really is better than chess in my opinion.

[illluck]

...no clear scoring mechanism in place to allow reasonable estimates of who really prevails in this activity regardless of ones proficiency in the game.

Completely wrong. IMO you can't be called a go player if you can't determine what the result of any typical game is (allowances obviously made for more exotic situations).

It's good that you have realized software scoring is generally not completely reliable, but you should really go find someone who can explain scoring to you in real-time.

[Alberich]

...no clear scoring mechanism in place to allow reasonable estimates of who really prevails in this activity regardless of ones proficiency in the game.

Completely wrong. IMO you can't be called a go player if you can't determine what the result of any typical game is (allowances obviously made for more exotic situations).

It's good that you have realized software scoring is generally not completely reliable, but you should really go find someone who can explain scoring to you in real-time.

I saw a YouTube video on this. In the clip the announcer claims the easiest way to count the game score is to take all your stone colors and group them by tens and then simply add the prisoners to the game score to find the winner.

[oren]

I saw a YouTube video on this. In the clip the announcer claims the easiest way to count the game score is to take all your stone colors and group them by tens and then simply add the prisoners to the game score to find the winner.

It would probably help to have someone explain it to you locally. It's not difficult to do, but the first time it can seem confusing.

[shapenaji]

I actually agree with OP, I think a big part of the difficulty in the game's reach is the difficulty with ending the game.

It can certainly be taught, but when you compare it to chess (where people can/do accidentally checkmate) it's a barrier.

It's also a negotiating step, and that always makes things hard. Heck, I look at kyu games and see a lot of boards where I'm like "That board is not done, now get into that moyo and exploit some aji"

[palapiku]

I don't think its possible to get people interested in a game that has no clear scoring mechanism in place to allow reasonable estimates of who really prevails in this activity regardless of ones proficiency in the game. Chess doesn't have this problem because the goal is much clearer...check mating the king and this can be calculated with precision using chess engines.

How do you think people played games before computers? I remember chess was pretty popular even before there were strong engines.

And what games do you want score estimation for? If it's your own games, surely estimating the score should be your own responsibility. If it's high dan or pro games, which is what most people tend to watch when they're not playing, then one of the reasons they're difficult to score is because they tend to be really close the whole time. You're not likely to see anything interesting from an accurate score estimator in the middle of a high-dan game.

[Mike Novack]

Alberich -- what are you saying?

a) That even the stronger progams might give a slightly different projection than the eventual score?

b) That the stronger programs give an incorrect final score.


I have observed "a". But if you claim "b" I would like you to show us some examples. Please limit that to the programs capable of playing at a couple dan on a reasonably powerful PC. In other words, don't tell us that something running on a phone can't get it right.

[hyperpape]

I actually agree with OP, I think a big part of the difficulty in the game's reach is the difficulty with ending the game.

It can certainly be taught, but when you compare it to chess (where people can/do accidentally checkmate) it's a barrier.

It's also a negotiating step, and that always makes things hard. Heck, I look at kyu games and see a lot of boards where I'm like "That board is not done, now get into that moyo and exploit some aji"

Teach stone scoring, and this is a lot easier. With two absolute beginners, you might see them play far longer than they need to, which is bad, but probably easier than the alternative. With a teacher, the teacher's pass can be the sign to end the game.

[Phelan]

While I understand your complaint about score estimation, the problem is that accurate score estimation is a skill as any other in go, and one that is particularly hard for computers. What computer programs do for estimation is usually to play out the game for both sides until the end with a Go AI playing program.
Since several of them use different AI programs, the estimates will vary wildly unless the game is close to an end.

[Alberich]

While I understand your complaint about score estimation, the problem is that accurate score estimation is a skill as any other in go, and one that is particularly hard for computers. What computer programs do for estimation is usually to play out the game for both sides until the end with a Go AI playing program.
Since several of them use different AI programs, the estimates will vary wildly unless the game is close to an end.

But as we see in actual iGo games - not all of them are played out to the bitter end to prove one side is truly the winner. But even when games are stopped like maybe around move 100 there are some strange readings by the Go software. I've seen games resigned by either side when the assessment by the software says exactly the opposite!

I can say with assurance there were some games I've played where the result was recorded completely differently by the scoring mechanism used by the programs I used. One said I won by 3.5 points while the other said I lost by 295 points. Unfortunately I don't have a record of the game score as I erased some of the results. But the next time I run this experiment I will post the game score and verify the software programs I used to show what I've been seeing.

[Boidhre]

While I understand your complaint about score estimation, the problem is that accurate score estimation is a skill as any other in go, and one that is particularly hard for computers. What computer programs do for estimation is usually to play out the game for both sides until the end with a Go AI playing program.
Since several of them use different AI programs, the estimates will vary wildly unless the game is close to an end.

But as we see in actual iGo games - not all of them are played out to the bitter end to prove one side is truly the winner. But even when games are stopped like maybe around move 100 there are some strange readings by the Go software. I've seen games resigned by either side when the assessment by the software says exactly the opposite!

I can say with assurance there were some games I've played where the result was recorded completely differently by the scoring mechanism used by the programs I used. One said I won by 3.5 points while the other said I lost by 295 points. Unfortunately I don't have a record of the game score as I erased some of the results. But the next time I run this experiment I will post the game score and verify the software programs I used to show what I've been seeing.

Which program are you using for scoring? Also, early game scoring by, for example, Smart Go is waaay off because of how complicated the question is.

[Bantari]

There is simply no reason why I should even bother with the scoring systems in any of the software titles I've seen, whether for the iPad or pc. It's my opinion they're simply wildly inaccurate and incapable of giving reasonable estimates.

I think you are absolutely right in the above! You should not bother with scoring systems of any off-the-shelf go software you can get. They are all wildly inaccurate and pretty much useless. The trick is to learn to count (or estimate) the score yourself - and then you won't need the scoring systems. Try it - it will help you improve your game as well.

On the other hand - relying on automated scoring systems just ruins your game.

[jts]

Alberich, the first time you made a thread like this several people asked you to post an example to that we could help you. Please do so. I promise you it will help.

At the end of the game, you should be able to calculate the score yourself, just by counting. Start at one, go from there.

SE on kgs is a mess of a program - wms didn't design it and he freely admits he has no idea how it works. It only works on Japanese scoring, and even then only if it gets l&d correct.

[Alberich]

I'm getting better at the game. I've figured out how to do better against the engines and now my losing scores are going down. Before my results were horrible but now I've figured out the trick on making the engines sweat I'm getting better moyo and aji. Life and death is still a problem though.

[topazg]

FWIW, if you're a Windows user, the best estimator I've ever found is Dariush: http://ricoh51.free.fr/go/dariusheng.htm

I find that's surprisingly accurate except for not being able to assess groups occasionally in particularly complicated group fights and liberty races - even then it's fairly good.