HermanHiddema: Re: ((no subject)) (2008-02-26 01:14) [#4398]

This is true. The EGF text is quite recent, and as far as I am aware, no tournaments have as of yet used either SOS-1 or SOS-2.

81.129.25.98: Re: ((no subject)) (2008-02-26 01:16) [#4399]

IanDavis: Agree, for occaisions when tiebreakers must be used it is generally used - pandanet cup for example. Many players think splitting money on SOS is silly though.

194.111.136.197: EGF recommendations dissected (2008-02-26 08:00) [#4402]

The EGF list could be improved in many aspects. The bulleted text is copy-pasted from http://www.eurogofed.org/egf/toursysrules.htm

• Number of Board Wins: It can be applied only in a team tournament. There it is highly meaningful and should be the first tiebreaker.

Even more highly meaningful tie breakers for team tournaments exist, for example the sum of SOS and board wins, which seems to be expressly forbidden by its nonexistence on the list. But most tournaments are not team tournaments, so let's forget this one for now.

• Direct Comparison: Provided it can be applied at all, it is very meaningful because it might be interpreted as an already performed knockout playoff among the tied players. So, for the final results, generally it should be the first or even the only tiebreaker.

Go directly to statistics 101. Do not pass GO. http://xkcd.com/221/

Or with less sarcasm: if two players are evenly matched, as suggested by their MMS tie, will the result of their knock out game always be exactly the same as the result of their first game? If player A was "already knocked out" by player B, does it not follow from their identical MMS that player B must have been "just as knocked out" by somebody else? If A and B are the only 2 players tied for the first place, does this not mean that B has lost to a player weaker than A, and should therefore have no place in this knockout at all? (no, I do not think so even myself, but this just as logical as the given explanation.)

(caution: straw man argument ahead.) A case where the DC would consistently work is if you declare that it is possible to have "winning rings" so that A usually beats B who usually beats C who usually beats A again. In this case, it is indeed very reasonable to suspect that the winner of one game is indeed more likely to win the other game too, no matter what their MMS. If this is what you believe, then using a tournament system based on the assumption that one's ability can be expressed by a single number (say, McMahon or Swiss) is silly though, and your results would show little else than artifacts of the pairing. (straw man 0 - bass 1)

• SOS-2, SOS-1, SOS: They should be used only in McMahon or Swiss. They express a mixture of opponents' strength, statistical noise, and pairing luck. Their apparent numerical precision is greater than their true significance. Therefore they must be used with care. Application for the final results is doubtful while application for making pairings is reasonable. SOS-2 filters more noise than SOS-1 than SOS; more noise can be filtered in more rounds more easily.

This is a very good description of SOS & friends. The only things it does not mention are 1) that the source of noise and luck is not in the SOS methods themselves 2)the variants haven't been used in practice, and 3) that the "noise" filtered is actually meaningful data. Also, pairing luck is kind of mentioned twice, as it is already reflected in the opponents' strength bit, and the "doubtful" part could be omitted because of the earlier precautions of "do not use tie breakers, and even if you disobey that, beware the resolution." Overall, excellent text.

• Rating. It expresses the players' relative strength just before the tournament's start. Its apparent numerical precision is much greater than its true significance. For players with similar ratings it behaves like lottery - for high dan players with significantly different ratings it is reasonably meaningful. Hence it can be applied just before lottery especially for seeding a great number of high dan players.

Well, why not recommend using the age of the players as a tie breaker as well, since clearly this is a desperate measure not to be used by any self respecting tournament organiser ever.

• Previous Order: If it refers to a previous tournament, this information is the more meaningful the more recently that tournament was played. It should be used for similar purposes as rating.

If ever the simple concept of "predefined order to use in case of ties" was mangled beyond recognition, this must be it.

• Lottery: It should be used only as the last tiebreaker and if any other tiebreaker cannot make more meaningful decisions.

And so, we are completely out of tie breakers. In summary, we are stuck with:

 * 1 inapplicable one
 * 1 illogical one
 * SOS and variants
 * 1 desperate one
 * 1 unreadable one, and
 * lottery

So it is quite safe to say that SOS is "in the upper half" of tie breakers.

• Global tiebreaker: In special tournament systems, it might be more meaningful than lottery if it fits the tournament system well and is defined carefully.

It is good to see that at least the list can be changed after it is written and declared absolute. At least I cannot figure out why this should be placed after lottery otherwise.

 -Bass

This rant was powered by a powerful mixture of frustration and caffeine, and is to be taken with at least as big a shipment of salt as you would use on general recommendations of international committees.