Tournament pairing algorithm - Is more average opposition balancing a possible improvement ?

@jonesmh #20

Indeed what's more if we question ourselves about what caused the rating being skewed we will find out that late joining leads to a weaker oposition. You even can experiment it by yourself by latejoining an Arena. Rather than an accident it's a constant. And a very nice one if you ever want to get the marathon badgie.

@kenzaburo
There's no doubt that, in the short-term, this tactic does indeed give you a pool of lower ratings, however when the scores are the same, they should be getting paired from the same pool, which should prevent the average from deviating as this example has shown.

BTW, I have known about the use of starting late to get, at first, easier opponents (not necessarily lower rated), and, morally, I can't join a tournament that has already accumulated more than two or three rounds.

@jonesmh

And here comes an intrincated idea. If the duration of the arena is long enough you could probably choose the right timing to get to the first page facing the statistically easiest oposition. In part due to the fact that strong players will remain on first pages or quit the tournament while all the lower rated players will be the main pool.

Also in this case we are talking about roughly 40 minutes which seem enough to me.

@kenzaburo
I've already conceded that this scenario is possible, and that it should self-correct itself in the long-term. However, I have also qualified that I was demonstrating that in this particular instance that the difference in average ratings should not have occured. (Although I also stated that a more precise and involved investigation would be necessary to correctly determine if the rating discrepancy was actually due to the lower rated players having a good tournament or some bug in the pairing code.

I don't know why you insist on disregarding the event which KC has referenced and try to use an unlikely probability, which is before the point of interest. As this is off-topic, and you aren't interested in the logical discussion, I'm not going to respond to your overweening rant.

The main difference in our arguments is that you are starting at the beginning of the players' tournament saying that it's not difficult to get to the top page. I'm stating that with the players starting from the same time with the same score, that is when they played against each other, that their opponents should have been selected from the same pool, and the winner, by having obtained a higher score when he defeated KC, should have been paired from a more difficult--although not higher rated--pool. The lower average rating doesn't mean that the winner player easier opponents, just that these lower rated players had a better score at that point in the tournament. Once again, starting when the scores are close at the same time in the tournament.

@jonesmh

And I am saying that in this particular case the rating difference is caused by late joining. There is a big difference in the time spent playing (40+ minutes) and almost 30% more games. It's not an unlikely probability since this have ocurred since I started playing here and I've played more than 10k games on lichess. There is no bug in the pairing code it's just the way it works. And the whole point of late joining is getting an easier pairing.

What I see in your argument is that the point of "equilibrium" is achieved by the late joiner when he get's to the first page, he get's paired against KC but he got an easier paring to get there, he gets paired against him a couple of times. And since all the first page is literally busy playing their "hard" games he keeps getting paired against lower rated players. It's not that he get's the strong low rated performers that are on the first page (they didn't get to the first page so he gets them from the second, third, fourth and so on). I think that the timing he did choose (willingly or not) is so good that he managed to have all his high rated oponents involved in a game when he was getting paired.