6.5 komi also in Japan
Joan Pons i Semelis
http://www.nihonkiin.or.jp/topics2002/brandnew-e.htm
komi will be upgraded also in Japan.
Any ideas about when western go will do it ?
(Besides ING sponsored tournaments, that have been doing it since long ago.)
Robert Jasiek
Joan Pons i Semelis wrote:
> Any ideas about when western go will do it ?
New Zealand has been using 7 for years now.
***
German championships changed their komi rule to 6 (ties in Round-Robin
or Swiss system are preferred by the majority of German championship
players) in autumn 2001 with its effect taking place since 2002. The
German tournament rules commission had suggested that and the German
federation's meeting adopted it without even saying one word about it;
that convinced is everybody that 5 komi was too low...
***
European Go Federation tournaments: I have repetitively suggested
increments in the rules commission in vain so far. Some motion about
komi will have to pass the general meeting. Let us see which European
country will be the least conservative and make an end to the most
ridiculous European Championships _in those years with functionally
different major sponsors_ when boards 1-16 use 5.5 komi and boards
17+ (hundreds of boards) use the equivalent of 7.5 komi (8 Ing komi)!
--
robert jasiek
Charles Matthews
"Robert Jasiek" wrote
<snip>
> European Go Federation tournaments: I have repetitively suggested
> increments in the rules commission in vain so far.
Unfortunate ... I think Robert means 'repeatedly' ... one does rather hope
so.
Charles
Robert Jasiek
Charles Matthews wrote:
> Unfortunate ... I think Robert means 'repeatedly' ... one does rather hope
> so.
Ah, thx, probably, from what I see in a dictionary now:)
--
robert jasiek
Bill Spight
The New Mexico Go Association adopted 6.5 komi in 1976. It also used
that komi in handicap games, given by White.
Best,
Bill
Tim Brent
IMO all it does is says White does ot have to try to win,but kust try
to stay within x points on the board.
Nor do I think it helps much for a 50/50 chance as I beleive @ iGS it
is a 52/48 split for White.
Denis Feldmann
Is it not time to ressurect the idea of bidding for komi ?
For those too young to remember (and too lazy to use Google), there was no
real objection to my last proposal (2 years ago), i.e.
1) Every player declare what is is prefered komi (at the same time as
registration in a tournament; or this could even be semi-permanently
recorded with EGF ratings), i.e. a number (integer or half integer)
2) For every even game (obviously, this doesn't blend very well with
handicap games), the real komi is the mean of the declared komi(s?)
3) The player having declared the lowest komi takes White (in case of a tie,
the colours are decided at random)
I will not give again the main arguments for this proposal (fairness, no
real possibility of psychological manipulation, openness of the procedure,
and even the possibility to determine the "right" komi (at a given moment))
But I would *really* interested to see what couter-arguments still subsist.
If none, why not give it a try?
Robert Jasiek
Denis Feldmann wrote:
> Is it not time to ressurect the idea of bidding for komi ?
It is always the right time for that! :)
> But I would *really* interested to see what couter-arguments still subsist.
Playing devil's advocate, the most famous counter-argument is given
as "A meta-game before the game is not a competition of go skill."
One can argue that only alternation would be a competition of go
skill but that positional judgement at the beginning of alternation,
i.e. just before the first move, would not be go skill.
--
robert jasiek
Jan Lucas
I think the problem is that everyone seems to think that a fair value
for komi for all players and games does exist. I think that isn't true.
If it would be possible to solve go completly, we might find that a 10.5
komi or something like that would a fair komi for perfectly played
games. But in a Reality not even pro games are perfect. A 6.5 or 7.5
komi might be fair for current pro players but I don't think it will be
fair for beginning and moderate players.
Jan
Robert Jasiek
Jan Lucas wrote:
> I think the problem is that everyone seems to think that a fair value
> for komi for all players and games does exist. I think that isn't true.
It is theoretically proven that perfect play gives a fixed result
K>=0.
What you mean is practical komi. Now what do you mean by "all games"?
Each [even] game starts from the empty board. There is no difference
from game to game.
There are different playing styles. Some favour lower, some higher
practical komi. That is a reason for bidding komi instead of fixed
komi but it is not an argument for 0 komi for all players' games.
> If it would be possible to solve go completly, we might find that a 10.5
> komi or something like that would a fair komi for perfectly played
> games. But in a Reality not even pro games are perfect.
Komi as used in games is always practical komi and never necessarily
perfect komi. So what. We talk about using practical komi in tournament
games.
> A 6.5 or 7.5
> komi might be fair for current pro players but I don't think it will be
> fair for beginning and moderate players.
I think contrarily but what do our subjective opinions mean?
If we are objective, then we can use komi with some aims like
- give every player the most direct chance to develop strategies
in the same komi environment as professionals
- give a group of (weak) players their desired average komi
- etc.
The first two might be different. The second aim is not particularly
suitable for tournaments including (also) strong players.
--
robert jasiek
Oliver Richman
black's choice to start a fight with 5 or move something like san ren sei.
You see a lot of books on the sanrensei or the chinese opening but you don't
see books specifically about playing white against those openings.
-frl
"Tim Brent" <timb...@canada.com> wrote in message
news:a6p3qu0915r4v45j4...@4ax.com...
Denis Feldmann
> Denis Feldmann wrote:
>> Is it not time to ressurect the idea of bidding for komi ?
>
> It is always the right time for that! :)
>
>> But I would *really* interested to see what couter-arguments still
>> subsist.
>
> Playing devil's advocate, the most famous counter-argument is given
> as "A meta-game before the game is not a competition of go skill."
Maybe. But look at my proposal. Point 1) eliminate this notion of meta-game
*almost* completely, as you must declare your bid for komi before entering
the tournament (or even for a fixed period of time, for all EGF
tournaments), and not necessarily knowig even the list of your potential
opponents.
> One can argue that only alternation would be a competition of go
> skill but that positional judgement at the beginning of alternation,
> i.e. just before the first move, would not be go skill.
Mmm. What kind of skill would it be, then? Chess skill?
Robert Jasiek
Oliver Richman wrote:
> You see a lot of books on the sanrensei or the chinese opening but you don't
> see books specifically about playing white against those openings.
There are if only you perceive them to be such:) IOW, I also have not
seen any explicit books on such.
--
robert jasiek
Joan Pons i Semelis
> is a 52/48 split for White.
I think that includes handicap games too, but a so called 1 stone
handicap is in fact only half stone handicap, and a 2 stone handicap
is 1.5 stone handicap, while categories are assumed 1 stone appart. In
normal handicap games white has half a stone advantage, so is normal
that white wins more games.
Douglas Ridgway
> Is it not time to ressurect the idea of bidding for komi ?
It's always a good time. The devil is in both tradition and details.
Let's talk about the details.
> 1) Every player declare what is is prefered komi [...]
There is the question of when this occurs. More gamesmanship is
available if players are allowed to redeclare for each game, e.g. a
player could diss their opponent by declaring zero, or bid extra high
if they really wanted black against some opponent. For imperfect play,
the true komi does in fact depend on the strategies of the players.
The tournament organizer might want input too: demanding particular
granularities and parities depending on the rules in use and the
desire to avoid or encourage ties. N/2, N, N + 1/2, 2N, 2N + 1, 2N +
1/2, 2N + 3/2 all seem like plausible restrictions, although some are
more plausible than others. One advantage of having discretized komi
values in an over-the-board context is that stones could be laid out
to remind the players what the komi was in this game in particular.
> 2) For every even game [...] the real komi is the mean of the declared komis
It's not clear to me that taking the mean is the best way to combine
the bids. Any komi value strictly larger than the low bid and less
than or equal to the
high big is fair. If we restrict to the values allowed by the
tournament organizer (which BTW taking the mean does not do), there
may still be multiple
choices. What choice is best? I would argue that the best choice is
the smallest value strictly larger than the low bid. First of all,
this is just the result that you would get if you held a standard
rising-price auction, with each side hanging on until their maximum is
exceeded. Also, think about the decision process for selecting a bid.
You may know the maximum you'd be willing to pay (perhaps you even
know the true fair komi), but you'd rather pay less, if you can. If
the rules set the komi at the high bid or the mean of the bids, you
would have to temper your desire to have the first move with your
guess about the bid of the other player. It becomes quite complicated
picking an optimal bid. However, if the final value is set at the low
bid + epsilon, then bidding is simple. You just bid the absolute
maximum you would pay under any circumstances. If the other player is
an idiot and bids way too low, you receive the full benefit of their
idiocy no matter how high you bid. There's no need to spend effort
guessing how big an idiot they will be. It's optimal to bid your max,
regardless of what you believe your opponent may bid. I believe that
no other rule allows an optimal strategy selected without a model for
the opponent's strategy.
I think that this is the preferred mechanism by people who study this
stuff, i.e. economic game theorists, but I'd be happy to be corrected.
So, the obstacles?
* Tradition
* Details of the rules
* Making sure everybody (players, organizers, and associations)
understands and hopefully agrees with the rules
* Practicalities, like bidding process, keeping track of the komi
during play, how to rate flexible komi games
Douglas Ridgway
rid...@dridgway.com
Robert Jasiek
Douglas Ridgway wrote:
> keeping track of the komi during play
This is a central problem in tournament play. It requires a komi
bid form if each game is bid for separately. If the tournament
organizers collect the komi in advance, then they are responsible
for storing the bids. If the whole tournament uses some mean bid
of all players, then it is easier but we get other problems.
--
robert jasiek
Denis Feldmann
Again, I repeat my proposal (dont anybody reads it?): each player gives in
advance his prefered komi (this can bemade public, in the same lists as EGF
ratings, say, and not changed too often). Then, for each (even) game, the
real komi is the mean of the prefered komis (this can be automatically
calculated with the draw; the players don't even need to do it themselves),
and the player with the highest prefered komi takes black.
What is wrong with that?
Robert Jasiek
Denis Feldmann wrote:
> What is wrong with that?
Why would you suggest a particular, possibly strange period of
time during that every player must keep his preferred komi constant?
In practice how do you expect tournament organizers to handle your
suggested style of the players' komi values? (And how may a player
object if his komi has been manipulated wrongly by accident?)
--
robert jasiek
Denis Feldmann
> Denis Feldmann wrote:
>> What is wrong with that?
>
> Why would you suggest a particular, possibly strange period of
> time during that every player must keep his preferred komi constant?
Just to prevent possible psychological manipulations. In fact, I don't
insist on that, i believe giving one's prefered komi at registration time is
enough, *exactly* in the same way that giving one's rank when entering a
tournament (with the same side effects when some players (often japanese
ones :-) overrank themselves in order to get"more interesting" games, while
other underrank in order to win :-))
>
> In practice how do you expect tournament organizers to handle your
> suggested style of the players' komi values?
In the easiest possible way :-) I. e : after pairing is done (by one of the
standard pairing programs), the mean komi is calculated, this (and the
colour of the players) are indicated with the draw (nothing really new here)
and that's it.
(And how may a player
> object if his komi has been manipulated wrongly by accident?)
Similar to wrong handicap decision in variable handicap tournaments.
Bantari
says...
> There are different playing styles. Some favour lower, some higher
> practical komi.
And there even is a difference from game to game for same players.
I, for example, favor higher komi when I have White, while I definitely
favor lower komi when I have Black. Strange, no? Go figure.
Hehe...
--
__________________________________
- Bantari
kapr...@yahoo.com
Denis Feldmann
> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote in message
> news:<ant080$t8$1...@news-reader10.wanadoo.fr>...
>
>> Is it not time to ressurect the idea of bidding for komi ?
>
> It's always a good time. The devil is in both tradition and details.
> Let's talk about the details.
>
>> 1) Every player declare what is is prefered komi [...]
>
> There is the question of when this occurs. More gamesmanship is
> available if players are allowed to redeclare for each game,
This is usually the main objection: it is (almost universally) not wanted to
add a new skill (not really go-related). This is the reason i advocate no
redeclaration.
The rest of your text essetially analyze real bidding , which I disapprove.
See my proposal again
(1) open declaration of preferred komi before you know who your opponents
will be, *and* not knowing their declarations, or 2) public
declarations,made in something like the EGF ranking lists, and valid for all
the tournaments in a given period of time)
e.g. a
> player could diss their opponent by declaring zero, or bid extra high
> if they really wanted black against some opponent. For imperfect play,
> the true komi does in fact depend on the strategies of the players.
Yes, see above.
> The tournament organizer might want input too: demanding particular
> granularities and parities depending on the rules in use and the
> desire to avoid or encourage ties. N/2, N, N + 1/2, 2N, 2N + 1, 2N +
> 1/2, 2N + 3/2 all seem like plausible restrictions, although some are
> more plausible than others.
It is always possible to forbid ties. Not very satisfactory, though. I would
say that a tie is the "normal" result of the perfect game played with
perfect komi
One advantage of having discretized komi
> values in an over-the-board context is that stones could be laid out
> to remind the players what the komi was in this game in particular.
Doesn't seem very important to me. People strong enough to adapt their yose
(or even their middle-game strategy) to the exact count are surely also
strong enough to remember the komi...
>
>> 2) For every even game [...] the real komi is the mean of the
>> declared komis
>
> It's not clear to me that taking the mean is the best way to combine
> the bids.
Strange. I thought that the argument was compelling. If A thinks that the
real komi is 8, B thinks it is 4, so A plays Black with 6 ,then A pays 2
points less that what he would have accepted, and B winds 2 point more that
what he would have accepted. What couldd be fairer?
Any komi value strictly larger than the low bid and less
> than or equal to the
> high big is fair.
I dont see why. It only garanties that both players are happy, not that they
are "equally" happy. Of course, in economic transaction (see below) one
should use the utility function, but (especially for small numbers), surely
two more points to A are of same value than two more points to B, don't
they?
If we restrict to the values allowed by the
> tournament organizer (which BTW taking the mean does not do), there
> may still be multiple
> choices. What choice is best? I would argue that the best choice is
> the smallest value strictly larger than the low bid. First of all,
> this is just the result that you would get if you held a standard
> rising-price auction, with each side hanging on until their maximum is
> exceeded. Also, think about the decision process for selecting a bid.
> You may know the maximum you'd be willing to pay (perhaps you even
> know the true fair komi), but you'd rather pay less, if you can. If
> the rules set the komi at the high bid or the mean of the bids, you
> would have to temper your desire to have the first move
What you hint at is that even with the perfect komi (for you) , you would
still prefer to play Black. Seems a contradiction in terms to me.
with your
> guess about the bid of the other player.
This is why only an open procedure (like having the preferred komi of each
player made public before the tournament)
raise no objections.
It becomes quite complicated
> picking an optimal bid. However, if the final value is set at the low
> bid + epsilon, then bidding is simple. You just bid the absolute
> maximum you would pay under any circumstances. If the other player is
> an idiot and bids way too low, you receive the full benefit of their
> idiocy no matter how high you bid. There's no need to spend effort
> guessing how big an idiot they will be. It's optimal to bid your max,
> regardless of what you believe your opponent may bid. I believe that
> no other rule allows an optimal strategy selected without a model for
> the opponent's strategy.
>
> I think that this is the preferred mechanism by people who study this
> stuff, i.e. economic game theorists, but I'd be happy to be corrected.
I am not sure; I thought the utility function was involved. But anyway, it
is a moot point.
Bill Spight
Douglas:
> > It's not clear to me that taking the mean is the best way to combine
> > the bids.
>
Denis:
> Strange. I thought that the argument was compelling. If A thinks that the
> real komi is 8, B thinks it is 4, so A plays Black with 6 ,then A pays 2
> points less that what he would have accepted, and B winds 2 point more that
> what he would have accepted. What couldd be fairer?
>
It would also be fair, I think, to regard the bids as bids in an auction
for playing Black. The player pays so much komi for that privilege.
However, taking the mean has an advantage over that. For instance,
suppose that everybody but one player bids 5.5, and that player bids 7.
Suppose, also, that she is right about the size of komi. If she gives 7
points komi, her games will be fair, but all the other games will be
fair, as well. Not fair to White, but fair to the players, who will have
a 50-50 chance of playing Black. Taking the mean rewards her for getting
the komi right. For then she would give komi of 6.25 (or equivalently,
6.5), and would have a slight advantage.
Rewarding players for setting komi correctly will lead them, over time,
to do just that.
Best,
Bill
Douglas Ridgway
> Douglas Ridgway wrote:
> > "Denis Feldmann" <denis.f...@wanadoo.fr> wrote in message
> > news:<ant080$t8$1...@news-reader10.wanadoo.fr>...
> >> 1) Every player declare what is is prefered komi [...]
> >
> > There is the question of when this occurs. More gamesmanship is
> > available if players are allowed to redeclare for each game,
>
> This is usually the main objection: it is (almost universally) not wanted to
> add a new skill (not really go-related). This is the reason i advocate no
> redeclaration.
A new skill is required, regardless of when the bidding occurs. If you
want to have komi bidding, komi bidding skill is involved, no matter
what the rules about bidding are. It's unavoidable.
> The rest of your text essetially analyze real bidding , which I disapprove.
You have to have an auction, whether the players bid once per game,
per tournament, per year, or per life. My text analyzes auctions, and
applies in any case.
> One advantage of having discretized komi
> > values in an over-the-board context is that stones could be laid out
> > to remind the players what the komi was in this game in particular.
>
> Doesn't seem very important to me. People strong enough to adapt their yose
> (or even their middle-game strategy) to the exact count are surely also
> strong enough to remember the komi...
True, most of the time. But even with the current system, where komi
is set by the organizer in a very regular way and is fixed to one of
perhaps two or three values, there have been instances in professional
play of players becoming confused about the komi during a game.
Variable komi will make this problem worse, so we'd better start
thinking now about how to deal with it.
> > It's not clear to me that taking the mean is the best way to combine
> > the bids.
>
>
> Strange. I thought that the argument was compelling. If A thinks that the
> real komi is 8, B thinks it is 4, so A plays Black with 6 ,then A pays 2
> points less that what he would have accepted, and B winds 2 point more that
> what he would have accepted. [...]
You seem to assume that each player will automatically bid their
estimate of the true fair komi. It's not obvious to me that this is
optimal play in the auction game.
> Any komi value strictly larger than the low bid and less
> > than or equal to the
> > high big is fair.
>
> I dont see why. It only garanties that both players are happy, not that they
> are "equally" happy.
By "fair", I mean that both players are playing under conditions that
they have agreed to. If player A bids 8, he or she is agreeing to play
black at komi 8.0
or less, and white at komi 8.0 or more. Any komi between the high and
low bid meets this definition of fair. If you imagine sealed bids,
then neither player would even know what the other player had bid, and
couldn't find any cause to be dissatisfied after the fact, either. I
don't have a hierarchy of fairness; a procedure is either fair (both
players agreed to play under the conditions) or it is unfair.
Could you tell me what "equally happy" means, and why it should be a
reason for adopting a particular auction rule?
I argue that if we must have an auction game, it should be as simple
as possible to analyze and play optimally, so that the role of bidding
skill is minimized.
> If we restrict to the values allowed by the
> > tournament organizer (which BTW taking the mean does not do), there
> > may still be multiple
> > choices. What choice is best? I would argue that the best choice is
> > the smallest value strictly larger than the low bid. First of all,
> > this is just the result that you would get if you held a standard
> > rising-price auction, with each side hanging on until their maximum is
> > exceeded. Also, think about the decision process for selecting a bid.
> > You may know the maximum you'd be willing to pay (perhaps you even
> > know the true fair komi), but you'd rather pay less, if you can. If
> > the rules set the komi at the high bid or the mean of the bids, you
> > would have to temper your desire to have the first move
>
> What you hint at is that even with the perfect komi (for you) , you would
> still prefer to play Black. Seems a contradiction in terms to me.
No; what I mean is that if I'm playing black, I'd rather pay the
minimal necessary komi for the right to play black. Any procedure
other than low bid+epsilon penalizes the auction winner for revealing
his or her true
preferences, and creates an incentive to deceive with a bid lower than
your true secret estimate if you think you can get away with it.
For example, using your numbers above, player A got penalized by 2
points for overestimating B's bid. If A had some idea that B might bid
that low, A could have bid 6 or 5, still won the auction, and paid a
lower komi. Under the mean-bid auction rule, this would have been
better play than bidding your true estimate of the fair komi. Under
the lowest-bid auction rule, there's no incentive to underbid [1]. My
claim is that this reduces the bidding skill required.
> with your
> > guess about the bid of the other player.
>
> This is why only an open procedure (like having the preferred komi of each
> player made public before the tournament)
> raise no objections.
How does having an open procedure simplify the process of selecting an
optimal bid?
In any case, perhaps we could agree on the following points?
* Komi bidding introduces an auction game into setting the
conditions of play
* Auction games involve skill, and optimal play is not always
obvious
* The details of the rules of the auction can drastically affect the
outcome
* Reasonable, knowledgeable people can draw intuitive conclusions
about the best rules for an auction which are widely divergent.
Generating consensus may take awhile.
Douglas Ridgway
rid...@dridgway.com
[1] However, there may be an incentive to overbid. This could be more
complicated that I thought. Perhaps I should do some reading on
auction theory.
Nikolaus Hansen
I like your idea pretty much, but...
>> Playing devil's advocate, the most famous counter-argument is given
>> as "A meta-game before the game is not a competition of go skill."
>
>
>Maybe. But look at my proposal. Point 1) eliminate this notion of meta-game
>*almost* completely, as you must declare your bid for komi before entering
>the tournament (or even for a fixed period of time, for all EGF
>tournaments), and not necessarily knowig even the list of your potential
>opponents.
...to me there still seems to be a *significant* amount of a meta-game
(see below).
>
>> One can argue that only alternation would be a competition of go
>> skill but that positional judgement at the beginning of alternation,
>> i.e. just before the first move, would not be go skill.
>
>Mmm. What kind of skill would it be, then? Chess skill?
>
No. Simply the skill to estimate the "true" komi right. If you do not,
and your opponent did, you obviously have a disadvantage which is not
related to your go skills.
You could solve the problem if only one player is afraid to have this
disadvantage. They could play the komi of the second player and with
randomly assigned colours. But if both players decide to be non komi
experts...
--
Niko
Denis Feldmann
> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote in message
> news:<ao39bd$kje$1...@wanadoo.fr>...
>> Douglas Ridgway wrote:
>>> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote in message
>>> news:<ant080$t8$1...@news-reader10.wanadoo.fr>...
>
>>>> 1) Every player declare what is is prefered komi [...]
>>>
>>> There is the question of when this occurs. More gamesmanship is
>>> available if players are allowed to redeclare for each game,
>>
>> This is usually the main objection: it is (almost universally) not
>> wanted to add a new skill (not really go-related). This is the
>> reason i advocate no redeclaration.
>
> A new skill is required, regardless of when the bidding occurs.
The term "bidding" is inappropriate, maybe.
If you
> want to have komi bidding, komi bidding skill is involved, no matter
> what the rules about bidding are. It's unavoidable.
But anyway, the key word in my previous sentence was "not really go
related". Skill in estimating the value of the komi is go-related. Skill in
outbidding one's opponent is not.
>
>> The rest of your text essetially analyze real bidding , which I
>> disapprove.
>
> You have to have an auction, whether the players bid once per game,
> per tournament, per year, or per life. My text analyzes auctions, and
> applies in any case.
Wrong. This is not a competition, but a declaration of estimation. Remember
there is no loser in my proposal, meaning this is not a zero-sum game
anyway.
>
>> One advantage of having discretized komi
>>> values in an over-the-board context is that stones could be laid out
>>> to remind the players what the komi was in this game in particular.
>>
>> Doesn't seem very important to me. People strong enough to adapt
>> their yose (or even their middle-game strategy) to the exact count
>> are surely also strong enough to remember the komi...
>
> True, most of the time. But even with the current system, where komi
> is set by the organizer in a very regular way and is fixed to one of
> perhaps two or three values, there have been instances in professional
> play of players becoming confused about the komi during a game.
If only this was the only thing a go player might get confused about :-)
> Variable komi will make this problem worse, so we'd better start
> thinking now about how to deal with it.
>
Really, you are not serious. At worst, it will make possible a new kind of
blunder. So what? here are important recorded cases of professional players
not remembering whose trun it is to play. Should we add special markers on
the side of the boards?
>>> It's not clear to me that taking the mean is the best way to combine
>>> the bids.
>>
>>
>> Strange. I thought that the argument was compelling. If A thinks
>> that the real komi is 8, B thinks it is 4, so A plays Black with 6
>> ,then A pays 2 points less that what he would have accepted, and B
>> winds 2 point more that what he would have accepted. [...]
>
> You seem to assume that each player will automatically bid their
> estimate of the true fair komi. It's not obvious to me that this is
> optimal play in the auction game.
So what is your optimal play ?(remember you have to do it not knowing your
opponent)
>
>> Any komi value strictly larger than the low bid and less
>>> than or equal to the
>>> high big is fair.
>>
>> I dont see why. It only garanties that both players are happy, not
>> that they are "equally" happy.
>
> By "fair", I mean that both players are playing under conditions that
> they have agreed to. If player A bids 8, he or she is agreeing to play
> black at komi 8.0
> or less, and white at komi 8.0 or more. Any komi between the high and
> low bid meets this definition of fair. If you imagine sealed bids,
> then neither player would even know what the other player had bid, and
> couldn't find any cause to be dissatisfied after the fact, either. I
> don't have a hierarchy of fairness; a procedure is either fair (both
> players agreed to play under the conditions) or it is unfair.
Have you ever heard of "leonine contracts"? Most people have often to agree
to do things they would rather not....
>
> Could you tell me what "equally happy" means, and why it should be a
> reason for adopting a particular auction rule?
>
> I argue that if we must have an auction game, it should be as simple
> as possible to analyze and play optimally, so that the role of bidding
> skill is minimized.
>
Again, there is *no* auction process in my proposal.
No. No more than giving one rank (which allows one to get a nice handicap,
or weaker opponents), as soon as this is an open procedure (ranks are
public).
> * Auction games involve skill, and optimal play is not always
> obvious
>
Sure
> * The details of the rules of the auction can drastically affect the
> outcome
>
Yes
> * Reasonable, knowledgeable people can draw intuitive conclusions
> about the best rules for an auction which are widely divergent.
I thought game theory would take care of that *for zero sum games*
Nikolaus Hansen
wrote:
>
>I think the problem is that everyone seems to think that a fair value
>for komi for all players and games does exist. I think that isn't true.
>If it would be possible to solve go completly, we might find that a 10.5
>komi or something like that would a fair komi for perfectly played
>games. But in a Reality not even pro games are perfect. A 6.5 or 7.5
>komi might be fair for current pro players but I don't think it will be
>fair for beginning and moderate players.
>
That means we need a formula that depends on the strengths of the
players resulting in values between e.g. 4 and 8?
Nikolaus Hansen
wrote:
>
>You seem to assume that each player will automatically bid their
>estimate of the true fair komi. It's not obvious to me that this is
>optimal play in the auction game.
>
It is quite obviously the optimal strategy here, if you assume that
your opponent is at least as smart as you:-)
In particular bidding the true fair komi is the best strategy, because
it is the only way to avoid
a) a possible disadvantage, if there is any uncertainty about your
opponents estimate, and
b) an assured disadvantage, if your oppenent knows the true fair komi
and chooses the optimal strategy.
Denis Feldmann
> Denis,
>
> I like your idea pretty much, but...
>
>>> Playing devil's advocate, the most famous counter-argument is given
>>> as "A meta-game before the game is not a competition of go skill."
>>
>>
>> Maybe. But look at my proposal. Point 1) eliminate this notion of
>> meta-game *almost* completely, as you must declare your bid for
>> komi before entering the tournament (or even for a fixed period of
>> time, for all EGF tournaments), and not necessarily knowig even the
>> list of your potential opponents.
>
> ...to me there still seems to be a *significant* amount of a meta-game
> (see below).
>
>>
>>> One can argue that only alternation would be a competition of go
>>> skill but that positional judgement at the beginning of alternation,
>>> i.e. just before the first move, would not be go skill.
>>
>> Mmm. What kind of skill would it be, then? Chess skill?
>>
>
> No. Simply the skill to estimate the "true" komi right. If you do not,
> and your opponent did, you obviously have a disadvantage which is not
> related to your go skills.
This is a go skill. Any discussion about that seems pointless. The exact
same problem lies in estimating the score during the game, and adapting your
srtrategies acurately. In fact, the strongest players are usually the ones
having thev best positional judgment
>
> You could solve the problem if only one player is afraid to have this
> disadvantage. They could play the komi of the second player and with
> randomly assigned colours. But if both players decide to be non komi
> experts...
I stil dont believe it. Ok, I make an estimation of komi at 9. It means I
believe my estimation is correct. I am then not afraid it could be wrong ...
If I am afraid, it means I don't know the value. But then , it doesn't
matter, does it? I can take any random komi, and still be afraid . How much
are you prepared to pay to have black (or, symetrically, are you prepared to
receive to have white?) Are you really saying you have no idea? This is a
little bit similar to the utility function in economics: people are usually
able to give preferences between two possibilities...
Simon Goss
>> Strange. I thought that the argument was compelling. If A thinks that the
>> real komi is 8, B thinks it is 4, so A plays Black with 6 ,then A pays 2
>> points less that what he would have accepted, and B winds 2 point more that
>> what he would have accepted. What couldd be fairer?
Bill:
>It would also be fair, I think, to regard the bids as bids in an auction
>for playing Black. The player pays so much komi for that privilege.
In an auction, after B bids 4, A's next bid is more likely 4.5 than
either 6 or 8, isn't it?
--
Simon
Robert Jasiek
Simon Goss wrote:
> In an auction, after B bids 4, A's next bid is more likely 4.5 than
> either 6 or 8, isn't it?
It depends on the bidding rules. If they allowed big jumps, then
I might make my maximal bid immediately to ensure getting black.
--
robert jasiek
Douglas Ridgway
> On 10 Oct 2002 09:59:07 -0700, rid...@dridgway.com (Douglas Ridgway)
> wrote:
> >
> >You seem to assume that each player will automatically bid their
> >estimate of the true fair komi. It's not obvious to me that this is
> >optimal play in the auction game.
> >
>
> It is quite obviously the optimal strategy here, if you assume that
> your opponent is at least as smart as you:-)
Right: it's optimal against a perfect opponent, but can be improved on
against imperfect opponents. You do better by bidding somewhere
between the true komi and your opponent's likely bid.
An appropriate analogy would be to a corner ripoff invasion which
would not live against perfect opposition. Trying such things is not
part of optimal play against a hypothetical perfect opponent, but it
is part of strong play against real ones.
However, most people would agree that ripoff invading is a go skill.
While estimating the fair komi also seems like a go skill, the kind of
tactical bidding mentioned above doesn't seem like go skill. It seems
more like poker, or bridge, or business negotiation.
Douglas Ridgway
rid...@dridgway.com
Planar
>It's not clear to me that taking the mean is the best way to combine
>the bids. Any komi value strictly larger than the low bid and less
>than or equal to the
>high big is fair. If we restrict to the values allowed by the
>tournament organizer (which BTW taking the mean does not do), there
>may still be multiple
>choices. What choice is best? I would argue that the best choice is
>the smallest value strictly larger than the low bid.
It seems to me that you are missing something: the situation is
symmetric. What is true of black with lower values of komi is also
true of white with higher values of komi. But your solution does not
respect that symmetry.
> First of all,
>this is just the result that you would get if you held a standard
>rising-price auction, with each side hanging on until their maximum is
>exceeded.
Only if you consider the rising price to be increasing komi, and the
object of the auction the right to be black. You can do the same
reasoning with "rising price" = "decreasing komi" and you are
auctioning the right to be white.
>If the other player is
>an idiot and bids way too low, you receive the full benefit of their
>idiocy no matter how high you bid.
On the other hand, if the other player is an idiot and bids way too
high, you receive almost no benefit from their idiocy. That doesn't
seem right.
Taking the average of the bids is a symmetric solution to a symmetric
problem. Remember the proverb about symmetric situations...
--
Planar
Nikolaus Hansen
wrote:
>Nikolaus Hansen <nikolau...@ingene.de> wrote in message news:<rdjbqus9pafl7olfq...@4ax.com>...
>> On 10 Oct 2002 09:59:07 -0700, rid...@dridgway.com (Douglas Ridgway)
>> wrote:
>> >
>> >You seem to assume that each player will automatically bid their
>> >estimate of the true fair komi. It's not obvious to me that this is
>> >optimal play in the auction game.
>> >
>>
>> It is quite obviously the optimal strategy here, if you assume that
>> your opponent is at least as smart as you:-)
>
>Right: it's optimal against a perfect opponent, but can be improved on
>against imperfect opponents.
For improvement it is *not* sufficient to assume imperfectness of your
opponent. You need to assume that your opponent is even more imperfect
than you which is quite a different assumption.
>
>An appropriate analogy would be to a corner ripoff invasion which
>would not live against perfect opposition. Trying such things is not
>part of optimal play against a hypothetical perfect opponent, but it
>is part of strong play against real ones.
Good example, wrong implication. In an even game with an even position
it is not strong play to invade if you would know (and must therefor
assume your opponent knows as well) how to disprove the invasion. The
main reason to do such invasion is not that you opponent is imperfect.
The main reason is that you are behind and trade "goodness" for
"outcome variance".
>
>However, most people would agree that ripoff invading is a go skill.
OK, it is a go skill in particalur because you need almost the
identical skill for any invasion.
>While estimating the fair komi also seems like a go skill,
No, if "go skill" means "skill of playing go" (there might be a high
correlation between the skill of playing go and the skill of
estimating the fair komi, but who knows?). Yes, if "go skill" means
"skill of playing go and estimating fair komi". You can decide which
definition you prefer.
>the kind of
>tactical bidding mentioned above doesn't seem like go skill. It seems
>more like poker, or bridge, or business negotiation.
ACK.
Nikolaus Hansen
<denis.f...@wanadoo.fr> wrote:
>
>This is a go skill. Any discussion about that seems pointless.
>
Well OK... than the open question is wether we want *this particular*
go skill to determine the outcome of one or another go game (more
often).
>The exact
>same problem lies in estimating the score during the game, and adapting your
>srtrategies acurately.
If you are sure about that... I have some doubts about the word
"exact".
>In fact, the strongest players are usually the ones
>having thev best positional judgment
>
Sounds reasonable.
>>
>> You could solve the problem if only one player is afraid to have this
>> disadvantage. They could play the komi of the second player and with
>> randomly assigned colours. But if both players decide to be non komi
>> experts...
>
>I stil dont believe it. Ok, I make an estimation of komi at 9. It means I
>believe my estimation is correct. I am then not afraid it could be wrong ...
Doing the best estimation and believing that this estimation is
correct is not the same. Actually believing it is correct with
probability one is foolish. Therefore I actually *should* be afraid of
beeing wrong. The question could only be wether the probability is
small enough not to worry.
>If I am afraid, it means I don't know the value.
Nobody *knows* the value. Otherwise we should use it and the whole
discussion would be needless.
>But then , it doesn't
>matter, does it? I can take any random komi, and still be afraid .
Of course it does matter, because I could loose the game due to the
wrong estimate. Taking a random komi of course makes it worse.
>How much
>are you prepared to pay to have black (or, symetrically, are you prepared to
>receive to have white?) Are you really saying you have no idea?
Well, then lets do it the other way around. What is your idea, e.g.
for a 10 kyu, 5 kyu, 1 dan and 5 dan amateur player, respectively? (I
am truly curious about the numbers). Would e.g. any 9 dan pro agree
with you?
I have the idea strong players gave me. The idea seems to vary, say,
between 4 and 8 (or 9?) probably depending on the strength. I do not
see why I myself (far away from any professional level) should be
smarter on that problem than all these people out there. My own
empirical observations on this subject are neither well enough
documented nor seem they IMHO to be sufficient to make any serious
estimate.
In spite of all this above where I somewhat played the devils
advocate: I could live very well with the procedure you suggested!
--
Niko
Denis Feldmann
No, you still don't understand. Lets say i believe the komi is 50. Ok, I am
wrong. Even I can see that i am wrong *for what should be the real komi* But
this is not the point. The point is I am happy to play black and pay only
40, since I believe 50 should still be a fair price. And of course, i am
never happy to play white with the ridiculous low komi they give in
tournaments nowadays, and with my system, I never will have too. Now, are
you seriously saying that you are afraid to play with me at whatever your
favorite komi is, knowing that you will have to play white, and get about 27
points komi ??
>
>> If I am afraid, it means I don't know the value.
>
> Nobody *knows* the value. Otherwise we should use it and the whole
> discussion would be needless.
>
>> But then , it doesn't
>> matter, does it? I can take any random komi, and still be afraid .
>
> Of course it does matter, because I could loose the game due to the
> wrong estimate.
I cannot see how. Read the following reasoning above again.
>Taking a random komi of course makes it worse.
>
>> How much
>> are you prepared to pay to have black (or, symetrically, are you
>> prepared to receive to have white?) Are you really saying you have
>> no idea?
>
> Well, then lets do it the other way around. What is your idea, e.g.
> for a 10 kyu, 5 kyu, 1 dan and 5 dan amateur player, respectively? (I
> am truly curious about the numbers). Would e.g. any 9 dan pro agree
> with you?
>
I don't know,and I don't care. I dont think it is so much a matter of
strength. My favorite komi has been 7,5 for the last 25 years, from 1 kyu to
4 dan. Ok, maybe I am just stubborn.
> I have the idea strong players gave me. The idea seems to vary, say,
> between 4 and 8 (or 9?) probably depending on the strength. I do not
> see why I myself (far away from any professional level) should be
> smarter on that problem than all these people out there.
Sure. But those are two different problem. What is the right komi is
obviously too hard for human being. What is your most comfortable komi
(eventually depending of circumstances, your opponent, etc.) is clearly
something you are in the best position to evaluate yourself.
Nikolaus Hansen
<denis.f...@wanadoo.fr> wrote:
>> Doing the best estimation and believing that this estimation is
>> correct is not the same. Actually believing it is correct with
>> probability one is foolish. Therefore I actually *should* be afraid of
>> beeing wrong. The question could only be wether the probability is
>> small enough not to worry.
>
>No, you still don't understand.
ACK
>Lets say i believe the komi is 50. Ok, I am
>wrong. Even I can see that i am wrong *for what should be the real komi* But
>this is not the point. The point is I am happy to play black and pay only
>40, since I believe 50 should still be a fair price. And of course, i am
>never happy to play white with the ridiculous low komi they give in
>tournaments nowadays, and with my system, I never will have too. Now, are
>you seriously saying that you are afraid to play with me at whatever your
>favorite komi is, knowing that you will have to play white, and get about 27
>points komi ??
>
No, I did not say that. But there are other players with other
favorite komis.
>>
>> Well, then lets do it the other way around. What is your idea, e.g.
>> for a 10 kyu, 5 kyu, 1 dan and 5 dan amateur player, respectively? (I
>> am truly curious about the numbers). Would e.g. any 9 dan pro agree
>> with you?
>>
>
>I don't know,and I don't care. I dont think it is so much a matter of
>strength. My favorite komi has been 7,5 for the last 25 years, from 1 kyu to
>4 dan. Ok, maybe I am just stubborn.
>
You are right, I don't understand. I don't understand your contentness
about a *certain value* for the komi whatsoever the value might be
(see below).
>> I have the idea strong players gave me. The idea seems to vary, say,
>> between 4 and 8 (or 9?) probably depending on the strength. I do not
>> see why I myself (far away from any professional level) should be
>> smarter on that problem than all these people out there.
>
>Sure. But those are two different problem. What is the right komi is
>obviously too hard for human being. What is your most comfortable komi
>(eventually depending of circumstances, your opponent, etc.) is clearly
>something you are in the best position to evaluate yourself.
No again, because my favorite komi does not have a specific value. My
favorite komi is the komi that results in equal chances to win the
game for both players if both players (the opponent an me) are equally
strong. I want a fair game (after the komi is fixed). The more I am
convinenced to have this komi, the more comfortable I am. (And I need
strong evidence to be convinced.)
This seems very different from your approach what your favorite komi
seems to be. And therefore different methods might be sensible to meet
the different demands.