In message <[email protected]>, Don
Dailey <[email protected]> writes
In swiss tournaments or knockout tournaments it's important to rank the
players to the best of your ability before doing the pairings. This
does indeed increase the chances of ending up with the best player at
the end and the effect is not insignificant.
It's very easy to see this with knockout tournaments. Imagine a
tennis tournament where the top 4 seeds played each other in the first
and second round so that only 1 remains. For everyone else, the
pairings would be much easier and every players chance of surviving
till the final round has gone up dramatically. You would see players
you never heard of in the semi-finals, etc.
The same basic principles apply to swiss tournaments. You have much
fairer tournament if you know in advance how strong each player is and
can pair accordingly. This is not just to increase excitement, it's
mathematically sound.
I understand this if there's a rule that two opponents don't play each
other twice. But the KGS tournaments aren't like that. See, for
instance, the cross-table at
http://www.weddslist.com/kgs/past/69/index.html . You will see that the
cells near the diagonal are generally full, some even having two games
in them, while those furthest from the diagonal are empty: Zen and pachi
never played SimpleBot and WeakBot50k. The KGS tournament scheduler has
deliberately arranged it that way, because a game between Zen and pachi
gives more information about which is the strongest player than does a
game between Zen and WeakBot50k.
You are proposing that the tournament should start by pairing strong
players with weak players, and claiming that this is more likely to
result in the strongest player winning the tournament. I don't see it.
Nick
In swiss pairings, you don't just select the players initially by
strength, but when you make score groups you also rank the players by
strength. If you have 10 players with 3 points, you don't play the
best 2 players, you play the top 5 against the bottom 5 in the proper
order. Otherwise, you will likely have the ridiculous situation where
the best players is less likely to win this round that the weakest.
Imagine these ratings for the 3 point group:
2500
2490
2300
2200
2100
2000
1900
1800
1700
1200
If you pair the 1700 vs the 1200 and the 2500 against the 2490, you
have ridiculous situation where the the 1700 player is far more likely
to advance than the 2500 player! Since the problem is recursive,
you will now have a 1700 player in a group in which he probably does
not belong and the 2500 player will be playing way down in the next
round. It's not just a matter of fairness to these players, but to
the ones who have to play them in later rounds.
Some rulebooks that describe swiss tournaments will have some clause in
them that tells you to estimate the rating of the players if you don't
have an actual rating and that it should be the best estimate you can
muster. So even if you are wrong, you should at least attempt to
estimate the ratings of the players for pairing purposes.
Don
On Sun, May 8, 2011 at 2:22 PM, Nick Wedd <[email protected]> wrote:
In message <[email protected]>, Ingo Althöfer <
[email protected]> writes
Hi Nick,
thanks for the information.
>KGS seems to use a sub-optimal pairing program: ...
>There are 12 participants.
>How can you pit, in a CH-system, four of the five
>strongest against each other in the very first round?
It does not use any knowledge of the players' strengths.
And if it did, it would _prefer_ to pair the stronger players
against
each other, so as to obtain more information.
That might be good for sparring tournaments.
However, in "title tournaments" a normal pairing
scheme should be used, to get highest chances
that the best participant wins.
A normal scheme for such a goal would
be to pair the upper half of the participants
against the lower half in the first round.
There are reasons for doing a pairing such as you describe. But I
don't believe that "highest chances that the best participant wins"
is one of them. Indeed, if that is your objective, such a pairing
is pessimal.
Nick
--
Nick Wedd [email protected]
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
--
Nick Wedd [email protected]
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
