It is currently Thu Oct 31, 2024 4:28 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 45 posts ]  Go to page Previous  1, 2, 3
Author Message
Offline
 Post subject: Re: Round Robin: SODOS or Direct comparison?
Post #41 Posted: Mon Jul 05, 2010 8:12 am 
Judan

Posts: 6230
Liked others: 0
Was liked: 793
topazg wrote:
What matters is anything the tournament rules state that matter. If that includes a tiebreaker, then wins is not all that matters.


This applies at moment after such rules have been set. During moments before, there is rather the question first whether and possibly why ties shall be brokwn.


This post by RobertJasiek was liked by: Harleqin
Top
 Profile  
 
Offline
 Post subject: Re: Round Robin: SODOS or Direct comparison?
Post #42 Posted: Mon Jul 05, 2010 9:19 am 
Tengen
User avatar

Posts: 4511
Location: Chatteris, UK
Liked others: 1589
Was liked: 656
Rank: Nebulous
GD Posts: 918
KGS: topazg
RobertJasiek wrote:
This applies at moment after such rules have been set. During moments before, there is rather the question first whether and possibly why ties shall be broken.


I agree, but is there honestly an argument for why the reasoning for one is more appropriate than the reasoning for another? The validity of the advantages against the disadvantages of each are in the eye of the beholder only - fairness with regards to awarding the most appropriate result is entirely subjective, dependent on what people feel individually is "fair".

Top
 Profile  
 
Offline
 Post subject: Re: Round Robin: SODOS or Direct comparison?
Post #43 Posted: Mon Jul 05, 2010 1:07 pm 
Lives in gote

Posts: 309
Liked others: 3
Was liked: 41
Rank: 5 dan
Let's assume, as done in many rating systems:
rA is a scalar measuring player A's strength
rB is a scalar measuring player B's strength

p(A beats B) = 1/(1 + exp(rA - rB))

NA number of wins of Player A in round robin
NB number of wins of Player B in round robin

I believe that the following conjecture holds:

In a round robin if NA = NB, then rA = rB.

I have not constructed a proof, but any maximumlikelihood computation on Round robin results I have done has supported it.


This post by Matti was liked by: Harleqin
Top
 Profile  
 
Offline
 Post subject: Re: Round Robin: SODOS or Direct comparison?
Post #44 Posted: Mon Jul 05, 2010 6:32 pm 
Lives in gote

Posts: 402
Liked others: 29
Was liked: 180
GD Posts: 1072
Matti wrote:
I believe that the following conjecture holds:

In a round robin if NA = NB, then rA = rB.

I have not constructed a proof, but any maximumlikelihood computation on Round robin results I have done has supported it.


If it is true, it depends on the functional form of the probability function. The AGA system uses a cumulative normal model and the assertion doesn't hold in that case.

Top
 Profile  
 
Offline
 Post subject: Re: Round Robin: SODOS or Direct comparison?
Post #45 Posted: Sun Jul 11, 2010 1:59 pm 
Dies with sente

Posts: 82
Liked others: 4
Was liked: 28
Rank: 2 kyu
GD Posts: 109
Universal go server handle: EricBackus
willemien wrote:
Code:
    a  b  c  d  e  f  | sc | sodos   
A:  =  W  L  W  W  W  |  4
B:  L  =  W  W  L  W  |  3     6
C:  W  L  =  W  L  W  |  3     7
D:  L  L  L  =  W  W  |  2     3
E:  L  W  W  L  =  L  |  2     6
F:  L  L  L  L  W  =  |  1



Who thinks that E should be 4th
and who thinks D should be 4th?

(and why)


I think D should be 4th. When DC is possible, I think it should be used before SODOS, at least for the round-robin type tournaments we are discussing here.

When I think about SODOS, it seems to me that any argument for it can be reversed into an equally valid argument for SOLOS. So the choice between SODOS and SOLOS seems arbitrary. (For example, the SODOS rewards E for winning against stronger opponents. But why not instead penalize E for losing against weaker opponents?) By comparison, the argument for DC can't be reversed into an equally valid argument for the other player. SODOS can still be useful, and I think it can be used when DC is not possible, when some method of tie breaking must be found. But in the case where DC is possible, I think it makes a better tie breaker.

I think this all becomes a little less clear when dealing with a typical McMahan tournament, because of the issue of one player possibly facing easier opponents overall than another player. But even in this case, any argument for SODOS can be reversed in to an argument for SOLOS, so even in this case I think DC is a better tie breaker.

In addition, I suspect that there may be complications in applying DC when there are cycles (A beats B, B beats C, C beats A) where each player in the cycle has the same number of wins. If we can't use DC due to the cycle, and then we use some other tie breaker to eliminate one player, do we then go back and use DC to decide between the remaining two?
--
Eric

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 45 posts ]  Go to page Previous  1, 2, 3

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group