![[Welcome to Sensei's Library!]](images/stone-hello.png "Sensei's Library")
SOS is a direct result of who a player is against. Losing a game pairs you lower in the field than winning a game. Having a pairing lower in the field means that your SOS will be lower. It appears that losing the first game of a tournament has a much greater effect on SOS than losing your last game. Losing your first game results in a significant chance that all subsequent opponents will be weaker. By contrast losing your last game does not change the field of opponents you will face at all.
Is this correct?
Yes, that is correct. The Swiss/McMahon tournament system is engineered to work so that the data collected on the earlier rounds is taken into account when finding the more important pairings for the later rounds. This "intelligent pairing" trickery enables Swiss to produce results in fewer rounds than the round-robin system, and it also has the effect that losing early will give you easier opponents. (Losing the first 2 games is sometimes jokingly called "stepping into the McMahon elevator" because it will give you the easiest opponents if you wanted to score 3 wins out of 5 only.)
I recently played a tournament where I lost the first two rounds (against a 1d and a 4d player), but still scored 3 out of 5. After these two losses, I got only easy opponents (1k, 1d, 1k). This is wildly different from tournaments where you win the first rounds. In the European go congress this summer in Villach, I won the first two rounds (against two 3d players), then also ended up on 3 out of 5 (losing to two 5d players in rounds 3 and 5, beating another 3d in round 4).
For me, it is obvious that I performed much better in the second example than in the first, even though both are still basically 3/5 results. In this example, SOS will reflect this quite clearly
But not because of the quality of SOS. Average rank of your opponents would have produced your observation without fail. With SOS you only have a likelihood that also it might produce that observation.
So you're basically saying that Sum of Opponent Ranks is better than Sum of Opponent (McMahon) Scores. If you feel that initial McMahon Score (which is based on rank) is better than final McMahon Score (based on rank and performance), why are we having a tournament at all?
RobertJasiek: No, I am not saying that. I have meant it for the specific purpose of collecting information about the opponents' ranks. - Your concluding doubt is justified, of course.