Declared komi (today Montpellier, tomorrow the world)
Denis Feldmann
resort city at 30 kms), ans so I could introduce (for the first time ever)
my proposal of "declared komi", which I reproduce below (as it will be
explained to the players tomorrow) ; I hope to be able to report the results
Sunday evening (GMT) ...
*********************************************
The "declared komi" system
1) Every player in the (beautiful) Montpellier tournament makes known at
registration a number (integer or half-integer) called his "favorite komi"
(fk). If he forgots (or declines to do so), his fk is settled at 5.5
2) For every even game, the "real komi" (rk) is the mean of the "favorite
komis" of both opponents.
3) The player with the larger "favorite komi" begins (and plays with black);
the komi of the game (i.e. the compensation given to White) being obviously
the "real komi" (rk). In case of equality of fk, colors are drawn
(randomly?) by Gotha (the Luc Vannier's organising program).
-------------------------------
Commentaries
1) Using this method, let's say that A declared a fk of 8 and B a fk of 4.
This means they feel the game is fair with those respective komis. The
procedure gives a real komi of 6, and A begins. A should be satisfied, as
she must pay only 6 points to be Black while she thought this advantage
should cost her 8 points ; B should be equally satisfied, as he gets a
compensation of 6 points, and was prepared to receive only 4.
2) Some similar systems have been criticized for introducing a psychological
dimension (a poker game) in the bidding period: let's say A knows that C
hates playing with White, and is willing to declare 12 as fk to be sure to
get Black. A has then only to propose 11.5 to get White and a komi of 11.75
points, while she could not hope for more than 8 if C knew A's fk in
advance.This is the main reason for point 1) above : no such kind of
manipulation is really possible, as every player must declare his fk before
knowing who will be his opponents.
3) Of course, if rk is an integer, the problem of drawn games ("jigos") is
back.
But is this such a large inconvenience in practice ? Anyway, Gotha manages
it perfectly...
mullens
I looked, but could not see an announcement of this tournament, other than
its date, at http://ffg.jeudego.org/tournoi.html , on the Montpellier site
http://ffg.jeudego.org/FRANCE/CLUBS/34Mo.html or the Paris site
http://www.go-paris.org/forum/viewforum.php?f=4
When this occurs, It seems perhaps that the numbers may be limited, or that
in reality it may be a picnic by the seaside (not that this isn't a good idea).
I confess that I must have become confused, because I thought that it was to
be held at the site of the Stage of Go - but I now see that this was for
Toulouse.
I was in Sète earlier this year and it seemed a nice enough place - apart that
is from the traffic jams.
Is there somewhere else that I should be looking for tournament announcements ?
Please correct me if I am wrong, but notifications on the www seem to happen
less frequently nowadays (eg Antony).
Anyway, I hope that you have a very good tournament.
Richard
Chris Lawrence
> my proposal of "declared komi", which I reproduce below (as it will be
This system seems quite fair. Is there any inherent disadvantage or
unfairness to it? The only one which I can think of is if A wishes for
just a slightly higher komi than B but is a relatively much stronger
player than B. A then gets to play first even though B really needs to
be black or even take a handicap.
Have I missed something there? Also, is it common practice to omit komi
when a handicap is used? I have seen mixed opinions.
This komi method reminds me of the problem of how to divide a cake in
half so that both parties are satisfied there was no bias. The first
person makes the cut and the second person chooses which half he would
like.
Montpellier is indeed beautiful. I flew there a couple of years ago
when visiting some friends in Lozere. The temperature in the UK was a
mild low 20 but in Montpellier it was about 37. I had to wait four
hours for the next train to Mende because the flight was delayed and
made me miss my train by just two minutes :-) I may go again because I
would like to check out the Sophia Antipolis region with a view to
possibly moving there in the future.
--
Chris
Denis Feldmann
> On Fri, 20 Jun 2003, Denis Feldmann wrote:
>
>> my proposal of "declared komi", which I reproduce below (as it will
>> be
>
> This system seems quite fair. Is there any inherent disadvantage or
> unfairness to it? The only one which I can think of is if A wishes
> for just a slightly higher komi than B but is a relatively much
> stronger player than B. A then gets to play first even though B
> really needs to be black or even take a handicap.
No, remember this is an even games tournament, where the purpose is to
determine the strongest player. So fairness is only that every player
thinks the initial conditions are at least even, or in his favour....
>
> Have I missed something there? Also, is it common practice to omit
> komi when a handicap is used? I have seen mixed opinions.
You can affine handicap by adding komi or reverse komi (like 2 stones +2.5
pts for Black). But my system doesn't cater for such, and in fact I cannot
imagine any internal procedure of fixing handicap that both players would
think fair ...
>
> This komi method reminds me of the problem of how to divide a cake in
> half so that both parties are satisfied there was no bias. The first
> person makes the cut and the second person chooses which half he would
> like.
Sure :-) How do you think it was found? My only personal touch is to have to
make your choic of komi *prior* to knowing your opponents...
Denis Feldmann
> A figment of my imagination whom I shall call Denis Feldmann wrote:
>> 3) Of course, if rk is an integer, the problem of drawn games
>> ("jigos") is back.
>> But is this such a large inconvenience in practice ? Anyway, Gotha
>> manages it perfectly...
>
> getting any from me... I would probably choose something like 7.99 or
> thereabouts.
Some people feel that drawn games are annoying ; some pairing programs don't
manage them. My system asks you to choose n or n+1/2 (it could be a
mistake :-) ); the probability of a drawn game is low, but not zero.
>
> - Michael
Carl Johan Ragnarsson
> 1) Every player in the (beautiful) Montpellier tournament makes known at
> registration a number (integer or half-integer) called his "favorite komi"
> (fk). If he forgots (or declines to do so), his fk is settled at 5.5
Isn't it Montpellier that is beautiful, and not the tournament... I
would have the players choose either integers or half-integers only.
Probably integers. (I would choose 8 :) )
I think the actual problem is that there is some difference playing
white and black, and a good player should master both. There could
easily be some player choosing 10 as his komi, just to make sure he is
playing black in all games. Still, I really like your idea, I hope
this or similar will be used in more tournaments.
And while we are at it, why not use the pie rule to its full extent...
one of the players make a move, then the other player gets to choose
sides.
regards,
Carl Johan
Ben Finney
> The "declared komi" system
>
> 2) [real komi is mean value of players' favourite komi]
> 3) The player with the larger "favorite komi" begins (and plays with black);
This requires some extra amount of organisation before and during the
tournament.
The "bidding komi" system, on the other hand, allows the players to sort
between themselves, at the time of the game, what the komi will be:
- One player starts by bidding an amount of komi
- Players alternate gidding higer komi amounts, until one player
passes
- The player with the highest komi bid takes black; the other player
takes white and receives that komi amount.
Thus, the player who gets black pays the price (komi to the other
player) that he thinks the advantage is worth; the player who gets white
is compensated higher than her own bid, so is also satisfied.
The "bidding komi" system seems to have all the advantages, and none of
the organisational overhead, of the "declared komi" system.
--
\ Ben Finney | "Whatever you do will be insignificant, but
`\ | it is very important that you do it."
_o__) big...@zip.com.au | -- Mahatma Gandhi
Pascal Cuoq
> - One player starts by bidding an amount of komi
> - Players alternate gidding higer komi amounts, until one player
> passes
> - The player with the highest komi bid takes black; the other player
> takes white and receives that komi amount.
>
> Thus, the player who gets black pays the price (komi to the other
> player) that he thinks the advantage is worth; the player who gets white
> is compensated higher than her own bid, so is also satisfied.
This has been discussed before.
This very interesting post gives some answers:
http://groups.google.com/groups?selm=aobvts%242dh%241%40ites.inria.fr
I will try to summarize the other good points in the previous, long,
discussion on that topic:
Firstly and importantly, offering the possibility to raise your bid is
*not at all* pertinent. Who in his right mind would bet less than
what he considers fair komi in the first place, and risk ending up
with Black in a game with that komi? Each player should need to bid
only once.
Taking the average of the bids is then the only fair way to split up
the bounty when players disagree on the value of first move. There is
no reason to give all to white.
It is disadvantageous to speak first, so the bids should be
simultaneous.
There is the problem of "mind games", where you have a guess at your
opponent's bid (this is related to the previous point, of course.
It's all too easy to "guess" your opponent's bid when he was the first
to bid). You then bid a value just above or just below his expected
bid, instead of what you sincerely think is fair, in order to get
more/give less compensation than you would with your normal bet.
These don't make the system unfair (you have as much to lose as you
have to win by playing these mind games, if your opponent plays them
too). But some people rightly argue that these games have nothing to
do with go, and would rather not have each game of go begin by
playing a different game just to decide of the komi.
Denis Feldmann then suggested to solve that last problem by augmenting
the granularity and limiting the players to one bid per tournament.
That is rather smart.
Pascal
Ben Finney
> Ben Finney <bignose-h...@and-zip-does-too.com.au> writes:
>> [points in favour of bidding komi]
>
> This has been discussed before.
>
> discussion on that topic:
These are all interesting points. Thanks very much for that excellent
summary; saves us re-hashing it again.
--
\ "Just because nobody complains doesn't mean all parachutes are |
`\ perfect." -- Benny Hill |
_o__) |
http://bignose.squidly.org/ 9CFE12B0 791A4267 887F520C B7AC2E51 BD41714B
Pascal Cuoq
> Who in his right mind would bet less than
> what he considers fair komi in the first place, and risk ending up
> with Black in a game with that komi?
Argh!
I meant White in this sentence, of course.
Pascal
David J Bush
> Argh!
> I meant White in this sentence, of course.
At the risk of being pedantic, I repeat the proposed bidding description:
- One player starts by bidding an amount of komi
- Players alternate gidding higer komi amounts, until one player
passes
- The player with the highest komi bid takes black; the other player
takes white and receives that komi amount.
Here is your (corrected) response:
Firstly and importantly, offering the possibility to raise your bid is
*not at all* pertinent. Who in his right mind would bet less than
what he considers fair komi in the first place, and risk ending up
with White in a game with that komi? Each player should need to bid
only once.
I don't understand your use of "risk" here. What risk? If you bid 0.5
komi and your opponent passes you wind up with Black not White. Suppose
you believe komi should be 7.5 but your opponent believes it should be
6.5. If you bid 6.5 your opponent would pass and you get to play Black
with a lower komi than you would have if you bid 7.5.
Am I mistaken about something?
David
Pascal Cuoq
> I don't understand your use of "risk" here.
Indeed, the mistake I corrected was in fact a deep misunderstanding
on my part about the proposed system.
I see now how the progressive climbing of the bids from a ridiculously
low value allows both players to be happy about never giving too much
information at once (bidding 1 gives no information at all, then
raising the bid to 2 gives almost no more information since most
players would be expected to have a higher estimation of the komi, and
so on...).
I still think there's something wrong about it not being symmetric.
The original post said this :
> Thus, the player who gets black pays the price (komi to the other
> player) that he thinks the advantage is worth; the player who gets white
> is compensated higher than her own bid, so is also satisfied.
which seems to imply that it's advantageous to white "who is compensated
higher than her own bid", but I now think that that sentence is a little
misleading, and that it's advantageous to Black, who never had to reveal
how much he was willing to give to white. The "higher" in the original
sentence should always be one point or zero (if it is allowed to bid at
the same level reached by your opponent).
So, the problem I have with this bidding system is : why should Black be
allowed to give only 5 points of komi, when he thinks that having first move
is worth 9 but his opponent thinks it's worth 4?
I agree that neither player can really complain: White was willing to
play second for only 4 points and got 5, but if the game ends up to be
very close, White can complain about the fact that if the auction had
been one of decreasing bids for the privilege to take white, he would
have won.
Pascal
Denis Feldmann
> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote:
>
>> 1) Every player in the (beautiful) Montpellier tournament makes
>> known at registration a number (integer or half-integer) called his
>> "favorite komi" (fk). If he forgots (or declines to do so), his fk
>> is settled at 5.5
>
> Isn't it Montpellier that is beautiful, and not the tournament...
Well, actually the tournament was held in Sčte (30 kms west of Montpellier,
on the seaside)
I
> would have the players choose either integers or half-integers only.
> Probably integers. (I would choose 8 :) )
In fact, 11 players on 64 did choose something <> 5.5pts ; those 11 players
had very good results, which of course might be a coincidence ;-)
>
> I think the actual problem is that there is some difference playing
> white and black, and a good player should master both. There could
> easily be some player choosing 10 as his komi, just to make sure he is
> playing black in all games.
Yes, this is a somewhat perverse side effect. Still, it means that he is
prepared to pay 10 pts komi, this feels large. In Montpellier, people choose
between 4 and 7.5
Still, I really like your idea, I hope
> this or similar will be used in more tournaments.
>
>
> And while we are at it, why not use the pie rule to its full extent...
> one of the players make a move, then the other player gets to choose
> sides.
A completely different proposal , meaning the first player much make a
mediocre move (I presume there is no komi). Interesting, but no longer real
go, or is it?
>
>
> regards,
> Carl Johan
Robert Jasiek
Denis Feldmann wrote:
> A completely different proposal , meaning the first player much make a
> mediocre move (I presume there is no komi). Interesting, but no longer real
> go, or is it?
IMO, it is interesting to play that way.
--
robert jasiek
Ben Finney
> I still think there's something wrong about it not being symmetric.
Fine; use nigiri to decide who starts the bidding.
> [Ben Finney] said this [about bidding komi]:
>> Thus, the player who gets black pays the price (komi to the other
>> player) that he thinks the advantage is worth; the player who gets
>> white is compensated higher than her own bid, so is also satisfied.
>
> which seems to imply that it's advantageous to white "who is
> compensated higher than her own bid",
That's not my intention; Black is also getting an advantage (that of
playing Black). In an even game, we're trying to minimise that
advantage to a degree amenable to both players.
> The "higher" in the original sentence should always be one point
I agree that bids sould only increase by integer amounts.
> or zero (if it is allowed to bid at the same level reached by your
> opponent).
No, this would simply result in a bid stalemate, if both players wanted
to play Black at the same komi.
> So, the problem I have with this bidding system is : why should Black
> be allowed to give only 5 points of komi, when he thinks that having
> first move is worth 9 but his opponent thinks it's worth 4?
Both players have their own idea of how much komi is fair. If they
believe they can win as white with a low komi, they don't have
justification to complain about the size of komi when they are proven
wrong.
> I agree that neither player can really complain: White was willing to
> play second for only 4 points and got 5, but if the game ends up to be
> very close, White can complain about the fact that if the auction had
> been one of decreasing bids for the privilege to take white, he would
> have won.
Then he should have stipulated a higher komi himself. It's rather
disingenuous to state that a komi is fair for your opponent (in an even
game) but unfair if that same (or higher, in your example!) komi is
given to you.
--
\ "No wonder I'm all confused; one of my parents was a woman, the |
`\ other was a man." -- Ashleigh Brilliant |
Douglas Ridgway
> This [komi bidding] has been discussed before.
>
> This very interesting post gives some answers:
>
> http://groups.google.com/groups?selm=aobvts%242dh%241%40ites.inria.fr
>
> I will try to summarize the other good points in the previous, long,
> discussion on that topic [...]
A couple of other points:
* Bidding games are complicated, not in the rules, but in the play.
Optimal play depends not only on your estimate of the distribution of
outcomes of games
between you and the other player(s) at a particular komi, but also
your estimate of their bidding behavior.
* The bidding phase therefore requires not only go skill, but also
skills which are not obviously go-related.
It seems to me that having the tournament organizer or an expert
committee readjust the komi every few decades is simpler, and just as
accurate and fair as any bidding scheme. If the game was entirely
novel, and we knew nothing aboutwhat the first move was worth, we'd
probably have to bid for it, or do you-cut-I-choose on the opening.
But go has been studied for thousands of years, and experts have a
pretty good idea of what going first is worth.
doug.
rid...@dridgway.com
Douglas Ridgway
> > "Denis Feldmann" <denis.f...@wanadoo.fr> wrote:
> >
> >> 1) Every player in the (beautiful) Montpellier tournament makes
> >> known at registration a number (integer or half-integer) called his
> >> "favorite komi" (fk). If he forgots (or declines to do so), his fk
> >> is settled at 5.5 [...]
> In fact, 11 players on 64 did choose something <> 5.5pts ; those 11 players
> had very good results, which of course might be a coincidence ;-)
[...]
> In Montpellier, people choose
> between 4 and 7.5.
Denis,
Congratulations on your tournament, and very interesting results!
Would you be willing to share the full distribution of bids? That
would be useful both for people setting up fixed komi tournaments and
also people bidding in future similar tournaments.
For me, already it shows that optimal bidding is complicated. For
example, my guess for a true fair komi would be around 6.5 or 7.0 --
that seems to be consensus these days. That's not what my bid should
be, however. If I had bid 6.0, I would have gotten black against every
single one of the 53 players who kept the default 5.5 komi, at a cost
of only 5.5 (5.75 == 5.5), an advantage of about one full point for
the large majority of games played. I lose a fraction for the small
minority of people who outbid me (eg I get only white at 6.5, instead
of random or alternating), but it would be well worth it.
Of course, I'd still lose the tournament, because I'd be crushed on
the go board. A one point handicap is not enough to put me into the
top ranks. And bidding results in the next tournament will be
different. In any case, strong players will need to carefully consider
not only what their estimate of true komi is, but also what others are
likely to bid, in order to maximize their advantage in the bidding
phase.
doug.
rid...@dridgway.com
Nick Wedd
Ragnarsson <abst...@abstractboardgames.com> writes
>And while we are at it, why not use the pie rule to its full extent...
>one of the players make a move, then the other player gets to choose
>sides.
I believe that the pie rule, in Go, would clearly favour the second
player. (Ok, I know that it theoretically favours the second player in
any game. But in, say, hex, there are starting moves where it is not
clear whether it is right to swap.)
If the first player moves anywhere on the 1-line, the second player
should accept this, with an advantageous position. If the first player
moves anywhere on the 2-line or higher, the second player should swap,
with an advantageous position.
Nick
--
Nick Wedd ni...@maproom.co.uk
Denis Feldmann
> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote in message
> news:<bd55u0$u2o$1...@news-reader4.wanadoo.fr>...
>
>>> "Denis Feldmann" <denis.f...@wanadoo.fr> wrote:
>>>
>>>> 1) Every player in the (beautiful) Montpellier tournament makes
>>>> known at registration a number (integer or half-integer) called his
>>>> "favorite komi" (fk). If he forgots (or declines to do so), his fk
>>>> is settled at 5.5 [...]
>
>> In fact, 11 players on 64 did choose something <> 5.5pts ; those 11
>> players had very good results, which of course might be a
>> coincidence ;-) [...] In Montpellier, people choose
>> between 4 and 7.5.
>
> Denis,
> Congratulations on your tournament, and very interesting results!
> Would you be willing to share the full distribution of bids?
Sure :-) Of the 64 players, 51 (not 53, as I wrongly wrote) took the
usual 5.5, 2 took 4, 1 took 5, 3 took 6, 3 took 6.5, 1 took 7 and 3
(including myself) took 7.5. Remember this is area counting, so for all
practical purposes, 5.5=6=6.5 ;-) (another fact I didn't properly took onto
account)
That
> would be useful both for people setting up fixed komi tournaments and
> also people bidding in future similar tournaments.
>
> For me, already it shows that optimal bidding is complicated. For
> example, my guess for a true fair komi would be around 6.5 or 7.0 --
> that seems to be consensus these days.
7.5 in area counting , yes (but remember Ing times)
That's not what my bid should
> be, however. If I had bid 6.0, I would have gotten black against every
> single one of the 53 players who kept the default 5.5 komi, at a cost
> of only 5.5 (5.75 == 5.5), an advantage of about one full point for
> the large majority of games played.
Yes, I agree, but this is only because people don't agree (well, bid as if
they dont agree) with your estimate. You couldn't know what people would do
(it was done at registration, and there was no publicity of the bid already
done). Of course, in the next such tournament, they might want to
reconsider ;-)
I lose a fraction for the small
> minority of people who outbid me (eg I get only white at 6.5, instead
> of random or alternating), but it would be well worth it.
>
> Of course, I'd still lose the tournament, because I'd be crushed on
> the go board. A one point handicap is not enough to put me into the
> top ranks.
Yes, that's true. But to have black in every game, this could make quite a
difference (if only because you can more or less give some direction to the
fuseki)
And bidding results in the next tournament will be
> different. In any case, strong players will need to carefully consider
> not only what their estimate of true komi is, but also what others are
> likely to bid, in order to maximize their advantage in the bidding
> phase.
Yes, I am beginning to suspect that my proposal is not as watertight as I
thought... ;-(
>
> doug.
> rid...@dridgway.com
Nick Wedd
Ridgway <rid...@dridgway.com> writes
>Would you be willing to share the full distribution of bids? That
>would be useful both for people setting up fixed komi tournaments and
>also people bidding in future similar tournaments.
Auction komi (different from Denis's system in details, similar in
effect) was used in the 1990/91 London Open Go Tournament, both for the
main tournament, and for the "rapid" side event. The statistics of all
the final komis are given in the British Go Journal issue no. 82.
In summary:
6 was the most popular final komi for both events, at all levels of
play. (A final komi of 6 meaning that one bidder was willing to bid 6
or more, the other was willing to bid 5.5 but not to bid 6).
Stronger players tended to bid more than weaker players.
Black won most of the games played with up to 7 komi in the main
event; and with up to 5 in the "rapid" event.
Bernd Gramlich
> The "bidding komi" system, on the other hand, allows the players to
> sort between themselves, at the time of the game, what the komi will
> be:
>
> - One player starts by bidding an amount of komi
> - Players alternate gidding higer komi amounts, until one player
> passes
> - The player with the highest komi bid takes black; the other
> player takes white and receives that komi amount.
This seems slightly unfair to the player with the lower bid.
Therefore, I suggest a more symmetric bidding system:
- Nigiri assigns roles to both players: B is bidding for black, and
W is bidding for white.
- The first bid is simultaneous (with hands closed just like in
nigiri): B offers b stones komi to W while W demands w stones
komi from B.
- Players take turns in changing their bids. Upon his turn, B may
increase b by one or pass, while W may decrease w by one or pass.
- If both players pass in succession or if w<b, bidding ends and
komi is (b+w)/2. Colours are determined as follows: If w<b, B
takes black, otherwise B takes white.
--
Bernd Gramlich [bE6nd "gRamlIC]
Bill Taylor
> I If the first player
> moves anywhere on the 2-line or higher, the second player should swap,
> with an advantageous position.
I'm not at all convinced about this!
If my identical twin started at 2-2 or 10-2 I would feel quite content
taking white against him.
What do others think?
------------------------------------------------------------------------------
Bill Taylor W.Ta...@math.canterbury.ac.nz
------------------------------------------------------------------------------
Some moves are merely KYUte, but others are truly DANgerous
------------------------------------------------------------------------------
ro...@telus.net
Taylor) wrote:
>Nick Wedd <ni...@maproom.co.uk> wrote
>
>> I If the first player
>> moves anywhere on the 2-line or higher, the second player should swap,
>> with an advantageous position.
>
>I'm not at all convinced about this!
>
>If my identical twin started at 2-2 or 10-2 I would feel quite content
>taking white against him.
2-2 is worth more than you think: IIRC Wimmer used to hustle Japanese
6-dans using it. IMO 10-2 is closest to a half-value move, but I
would still take Black. Moves at 10-2 are often valuable quite early
in the game if there is a stone at 10-4.
-- Roy L
Bill Taylor
> 2-2 is worth more than you think: IIRC Wimmer used to hustle Japanese
> 6-dans using it. IMO 10-2 is closest to a half-value move, but I
> would still take Black. Moves at 10-2 are often valuable quite early
> in the game if there is a stone at 10-4.
Sounds reasonable.
So let me ask you Roy, and anyone else who wants to add in,
can you think of an opening THREE (half-)moves,
that IYHO gives as near to equality as possible.
i.e. a baddish move for black, a good move for white, then another baddish
move for black, that gives equality (with white to move next OC); so that
you would be equally happy to play on with black or white.
Three-move equalization, in fact.
------------------------------------------------------------------------------
Bill Taylor W.Ta...@math.canterbury.ac.nz
------------------------------------------------------------------------------
Semi-eyes win semeais
------------------------------------------------------------------------------