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

I think the pairing is based a fortiori on the game points you have when pressing 'ready'. you get paired with the first available matching points candidate. This might produce a 1900 vs 1250 (in case a late joining 1900), but gives the desired quick pairings. Swiss works with default settings so no pairing know how or experience is needed. Swiss pairs based on the last round standing.
In the swiss alike pairing system at lichess you risk by entering late playing some very low pl as 2000+, so this will cap your TPR as it is weighted. If you berzerk, it seems to make no effect on the TPR, which is logical.

The Keizer system combines flexibility with retrospection.
In the Keizer system I believe there is a default or setting to avoid pairing a elo diff > 360. Depending on the version, and intelligence of the programmer, an implementation of the Keizer system can have many useful options and settings.

Now for smtg boring...

If you stumble over n over again seeing the awkwardness and rigidness of the swiss pairings, notably visible in the last half, take a look into the Keizer (= 'Ceasar') system .

Keizer pairing system is flexible: it is made for handling players entering after the 1 pairing, or adding extra rounds, even allowing pairing 2 times the same players like in a round robin. Another option: rating a chosen bye (in case you indicated on beforehand not to play the next round knowing u wont be there and there are a uneven amount of pl) better than a given bye.

Maybe that's an idea: a new setting or toggle button to indicate that you want to wait longer for a more suitable match. be it standing wise, TPR, or rating .

#4 I suggest there's another problem: strong players don't mind tough opposition the entire event, however that doesn't provide latecomers chances to play against the champion.
github.com/ornicar/lila/blob/227cd3303e1307afa35c7f4d942b41a78070d167/modules/tournament/src/main/arena/PairingSystem.scala

Tournaments are paired using the en.wikipedia.org/wiki/Blossom_algorithm , not necessarily based upon "Join" time.

Here's an idea: let's not pair anyone with anyone. This prevents the less able people from the pain of loss, the more able people from the pain of an upset, and prevents the whining of @Kingscrusher-YouTube about inane situations. No one has the possibility of getting their feelings hurt.
Although suggesting helpful criticism is viewed as good, it just provides a platform for people to complain about their petty problems. Many use suggestions have been provided, but they all contain a situation where someone could logically complain. Until a unbiased solution is presented just cope with the current feasible format.

Arguably KC's feedback (embracing the current mess and making it more enjoyable) is more constructive than mine.

They joined the tournament late, and therefore had to climb their way up to first page, by beating lesser ranked players. That's why the average opponent rating is lower.
Once they reached first page, they had the same tough opponents as other first page players, and the winner did the best job at beating them, therefore reaching first place.

Why I'm not getting paired in a tourney, where there has not been a game played for half an hour, a player comes in, whom I never played, we sit there for more than a minute. No pairing. Player pauses. Then an other players comes, whom I played in the first game, six games since then took place with different opponents. Again we sit there for a minute. No pairing. Player pauses.

This just kills off variant tourneys. I simply don't understand this.

lichess.org/tournament/WmbDpXkj

All_Hail_The_King and Navarra2, both decent players, I could have played a 3+2 game against them. No way.

To clarify the "late start" theory:
KingCrusher and the winner played each other three times, so you can say that they were in the tournament at the same time with similar scores.
After their first meeting, the winner played 16 more games with the average rating of 2082. KC played 11 more games with the average rating of 2222.
The first opponent for the winner was 1538 having a good tournament. His second opponent was 1796.
The first opponent for KC was 2002, and the second was 1910.

As i'm not interest enough to investigate further, those more concerned can do the harder math. You can use game times to see when the next pairings took place and check the math to find the scores of each opponent when they were paired.

I'm going to just assume that this situation is a rare event and disregard it.

@jonesmh #17

The winner had 23 games.
KC had 30 games.
Say an average 30 moves game with 50% time consumption per player. The winner berserks 36% of the time and KC berserks 16%. Those 7 games can easily translate into 5 minutes per game plus the 2.5 minutes per 3 which is 36% of 23 minus 16% of 30.

We get that KC played for 42 and half minutes longer. So the "late start" theory apparently works and isn't a rare event.

#pairmetoo

@kenzaburo We know when in the tournament the two played. By disregarding the previous games, we can use that point to as a "starting" point as both players should have a similar score. By using only data after this pairing, we can determine that the average ratings were skewed.
To expand, I, without further evidence, assume that the rating average difference is a rare event--not the theory as I infer from your statement.
BTW, the number of games, rate of going berserk, nor the length of playing time has any effect on the oppositions' average rating. The average is still the average.