Life In 19x19

SE is known to be inaccurate with AGA rules, but what I've never understood is the specific reason it would be. Moreover, while I can understand it being a point or two off, I'd gotten the impression from kibbitz that it can be off by several points, which blows my mind.

I don't actually know what acronym you are using. I'll take a guess. SE=Score Estimation?

Because it wasn't written by WMS nor has it been updated in A LONG time, nor will it from what he's said. It only takes in account Japanese rules.

Semi-related: http://www.lifein19x19.com/forum/viewtopic.php?f=24&t=604

ZeroKun wrote:Because it wasn't written by WMS nor has it been updated in A LONG time, nor will it from what he's said. It only takes in account Japanese rules.

I think you are missing the point. AGA rules were specifically designed to produce the same score as Japanese rules. So if the SE uses Japanese rules to estimate then it should prouce the same result regardless.

DrStraw wrote:

ZeroKun wrote:Because it wasn't written by WMS nor has it been updated in A LONG time, nor will it from what he's said. It only takes in account Japanese rules.

I think you are missing the point. AGA rules were specifically designed to produce the same score as Japanese rules. So if the SE uses Japanese rules to estimate then it should prouce the same result regardless.

Actually, the AGA rules were specifically designed to produce the same score as Chinese counting.

I assume this discussion is about the score estimator, which guesses the score during an unfinished game, and not the score counter, which determines the score of a finished game. In which case, who cares if it's off a few points? It's only an estimate.

One of the reasons SE fails at Area based rule sets is prisoner count (yes, AGA rules are considered "area based"). Area based rule sets don't count prisoners, where as Territory based rules sets do.

If both players play correctly, AGA and japanese rules will be within 1 point of each other (barring unusual seki or superko). But if players play incorrectly, the gap can be much larger. For example, if two players are tied before dame, and one fills in dame while the other passes, under japanese rules the score won't change, but under AGA rules the player filling dame is gaining points. There can be similar issues with playing inside, which costs a point in japanese rules but not AGA. I'm not sure if that's what throws of the score estimator though.

xed_over wrote:One of the reasons SE fails at Area based rule sets is prisoner count (yes, AGA rules are considered "area based"). Area based rule sets don't count prisoners, where as Territory based rules sets do.

As far as I remember SE is a lot worse in games with area rules than it is with territory rules. And that is very strange.
If you take the pass count into consideration the rule differences should only amount to one point(excluding certain uncommon situations). So it's strange that the SE is so much worse for area-scoring than it is for territory scoring. If it showed its territory based estimate it would probably more accurate than its current estimate when area rules are used.

emeraldemon wrote:If both players play correctly, AGA and japanese rules will be within 1 point of each other (barring unusual seki or superko). But if players play incorrectly, the gap can be much larger. For example, if two players are tied before dame, and one fills in dame while the other passes, under japanese rules the score won't change, but under AGA rules the player filling dame is gaining points. There can be similar issues with playing inside, which costs a point in japanese rules but not AGA. I'm not sure if that's what throws of the score estimator though.

The example above is not quite incorrect, but I will respond simply to make sure that the poster understands.

AGA rules are designed for use in AGA tournaments, that is, face to face games. They are also designed to make the final score equivalent to Chinese scoring. This is the purpose of white having to play last, and having to "lose" a point if it becomes his turn and there is no place to play. This will NOT change the result vis a vis Japanese counting, absent an unusual seki.

So yes, dame do count - but in practice not to the extent emeraldemon suggests. Since 99% of AGA games are counted in the usual Japanese style - what he suggests would not happen. If you are really playing AGA rules and black plays a dame and white passes, white would have to pass a stone, give a point. If this continues, yes the dame filler is gaining points - but it would be from the pass stones, not the dame stones. Needless to say, in practice, the passing player would only make this mistake once, and would begin filling dame instead of passing stones. However, the damage could be done, between the pass stone and changing the odd/even total of dame, it could be a two pt mistake and could change the result.

Note that the default counting in AGA is Chinese - so at the end of the game, a player may insist on Chinese counting. Then, the benefit of the pass stones would disapear, replaced by the benefit of playing the dame.

I hope this helps somebody.

It matters (a bit) because you sometimes use SE to judge the score when life and death is settled, and all major endgame has been played--just dame left, or perhaps a half point ko. In these instances (I think: really this post is to try and find out) SE can still work poorly with AGA rules.

And yes, this is laziness--I should just count.

Horibe: while AGA rules are designed for face to face, they're sometimes used for matches played on KGS.

hyperpape wrote:It matters (a bit) because you sometimes use SE to judge the score when life and death is settled, and all major endgame has been played--just dame left, or perhaps a half point ko. In these instances (I think: really this post is to try and find out) SE can still work poorly with AGA rules.

And yes, this is laziness--I should just count.

Horibe: while AGA rules are designed for face to face, they're sometimes used for matches played on KGS.

view offline --> change rules to japanese --> use SE