Why break ties
On different pages on tie breakers there is a lot of discussion as to whether ties should be broken at all. This page attempts to centralize the discussion.
Some players dislike ties, and prefer a more fine-grained ranking. Others prefer to be tied if the primary scoring result is identical.
- A single winner may be more conducive for publicity.
- Somebody with a very unfair draw may deserve to emerge as winner.
- Sometimes it's necessary to break ties, for example to select an entrant to the next stage of a championship
- Players that dislike ties will be happier
- Players can be separated despite there being no real difference in their performance
- Prize money, or prizes, can be unfairly allocated
- Players that like ties will not be happy
tapir: I guess this will end as a discussion page anyway :) In my opinion, breaking ties is unnecessary. Make komi 6 to allow for more diverse results (with jigo), and share places if necessary. If there are prizes not possible to share, make a rematch or nigiri to decide the winner. Nobody can complain about unfair tiebreakers anymore if it is skill (rematch) or fate/luck (random).
Bass: I do not believe your last statement about fate/luck. Say we are in some running competition, and we both get "10 min 23 sec" as our result. Also suppose you were some half a second (a couple of meters) ahead of me when we finished. Would you truly not mind at all if the competition organizer looked at the results and decided to toss a coin to determine which one of us would get the gold medal? I'd say that "fate/luck" is a good description of the random tie breaker, especially with the initial letters swapped.
tapir: You obviously have strong opinions about random tie breaking. Your comparison with running competitions still doesn't fit - most go tournament systems have win / lose only as a result - there is no equivalent to "10min 23sec" in go.
isd: Are you, tapir, not confusing winning a game with total scores in a tournament? There is an idea that a tournament performance rating of 2500 should win over tournament performance rating of 2000 where total score is level. It is not essential to break a tie. Yet if it can be shown that one player has performed better than another (run faster) then a tiebreaker might be used. This is the idea that Bass is suggesting. If you want to say that a tiebreaker can never give any indication of performance strength then that is fine, it is your opinion.
tapir: I love it that you only use such a broken analogy - are you not confusing a score with a time measurement? Time is continuous and can be measured always more fine-grained in case of tie. The score isn't, it is a multiple of 1 or 0.5 respectively. If you want more fine grained scores, why not use the Hahn tournament system? The SOS is an independent score (= sum of opponent's scores) showing something different (=the overall performance of your opponents and not your own) and it does not resemble a more fine-grained time measurement at all. 60 points difference in SOS do not make 1 point difference MMS as in 60sec = 1min. In my very humble opinion, it is more appropriate to compare opponent scores with the difficulty of the course in other sports (that may change during an outdoor tournaments due to wind or rain or even starting sequence) - this surely affects the result but since it is 1) another scale not comparable to the first one and 2) not in the hands of the athletes to change it, it is hardly used to decide winners. (If you need an example of a second scale used for tie breaking, take weight lifting. It has two independent scales. The weight lifted and the weight of the athlete and uses the second one for tie breaking.)
isd: I misread the starting line of this hare, but no matter. :) I don't think you need to get quite so hung up on the technical details of the analogy - it is after all just an analogy. A MMS is an indication of how a player performed during the tournament, it is also formed from the number of wins and losses they gained in that tournament. SOS is a sum of MMS. The idea for using SOS is that we can look at the level of opposition a player faced. (To quote from swissperfect.com "The idea is that the same score is more valuable if achieved against players with better performances in a given tournament.")SOS is of course distinct from, but subtly related to the MMS. If (and i'm not sure if) you believe that nothing can be derived from any tiebreaker such as SOS, SODOS, DC etc that is fine. Others will disagree :). I think the majority usage of SOS in Chess and Go indicates that most feel it is not quite equivalent to pure luck. This is not to say that I suggest SOS is perfect, swiss or holds a divine right to declare victory. For SOS surely contains some elements of both luck and skill - just as certainly as MMS does. Anyway, we will never get rid of the suggestion that a tiebreaker is unfair, this is as certain as the possibility that all tournaments will not double in round number or that all tournaments will not find time for a rematch to occur. For most tournaments a single winner is not needed at all. I believe we should not rush to use tiebreakers or rematches. When we do want to use one, then we should consider carefully what we feel we want to reward and how we can identify it.
Kirby: I think we've covered these points already in previous discussion. A tournament win means only what it's defined as. A coin flip brings greater accuracy to the result of a tournament where the winner is defined as "the person with greatest MMS, and in the event of a tie, break it by a coin flip". Arguing about what brings the "greatest accuracy" seems to be a moot point, because it can only be defined by what you are measuring (unlike a running competition where what's being measured is almost always defined in the same way). Vague concepts like "best performance" and "best skill" are not equivalent to game results. So we pick a system that we're going to use, we look at how the winner is determined, and choose a strategy to optimize our chances of winning. What system is the "best" system to use? Leave that up to the sponsors.
Bass: I find this discussion interesting, but why not let's leave "use this other tournament system X" and "leave it to the sponsors" out of it: They are the trivial cases, and I think everyone participating in this discussion is able to take those cases for granted, and focus on the remaining bit, which also happens to be the most usual occurrence: tournaments that use a tournament system just because it is customary, good and recommended.
Kirby: You should leave it to the sponsors because it's silly to think that you can truly capture "best performance" or "best skill" in a tournament. Claims that system X is definitely better than system Y are just opinions based on what the individual believes to be a good model of reality. It's impossible to do this definitively, and I find any system to be fun and acceptable. If a tournament gives money to the guy who lost the most games, that's cool, too. It's fine to call him the winner, too, if that's how the tournament rules are set. I still claim: let the sponsor chose a system that he likes.
Bass: Yes, let's let a sponsor choose any system he likes. Now, as I said earlier, why not let's drop that subject and focus on the case where there is no omnipotent being to conveniently impart absolute truths about which tie breaker is best, shan't we?
willemien Rematch (play-off) with nigiri is a form of tie breaking. And even with rematch there can be some problems (mostly because there is not time enough for a game under the same timelimits as in the tournament) see http://main.uschess.org/content/view/8462/456 for a chess - example. I don't know the timelimits for the normal games but the deciding game was a 5 minute per player one...
I agree with what you say about whole number komi. (see description under komi -> the correct komi)
isd: I think this has been said before on other pages. Usually there is no real need to break a tie. Most prizes and titles can be shared. If you need a unique winner there are a series of tiebreakers you can choose from. The European Championship used to use a rematch as a tiebreaker, it took place long after the european go congress.
Willemien This page was an idea to movea all those discussions to one place.
tapir: I like this: "The tie-break order should be Nigiri, SODOS, SOS. TMark" (Yes, Nigiri without rematch. Where possible I would not mind a rematch, but often schedules don't allow for it.)
willemien I don't understand, what do you exactly mean by nigiri? I understood it to mean a game where the colours are decided by nigiri am i wrong here? Tmark is not a tiebreaker (although i guess he sometimes likes to break them. (I spoke with him yesterday (30/12/09) , I should have asked...) SODOS is generally thought of as no good, i guess that has to do with complications because mostly the rankings are negative (almost only people above the bar have a positive score, all others have a negative score. having less defeated opponents therefore gives a higher SODOS score.
Here, it refers breaking ties randomly and arbitrary by Nigiri (and omitting the game thereafter). Tapir