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

Hi all

I quite enjoyed streaming the Weekly blitz for a bit last night.
lichess.org/tournament/eV03teGr

However, it does concern me how the pairings work, when it ends up that the winner of the tournament - who happens to be an IM on his profile, has an average opposition of 1976 and I get an average opposition of 2149.
Winners virtual tournament:

Performance 2448
Games played 23
Win rate 96%
Berserk rate 35%
Average opponent 1976

My tournament:

Performance 2174
Games played 30
Win rate 50%
Berserk rate 93%
Average opponent 2149

I don't mean to complain too much, but I see this fairly frequently. Is it not possible to try and make the tournament a bit "fairer" in terms of average opposition faced. It just seems wrong to me that the winner had such an easy pairing all the way through basically.

Not only this, I left the tournament with about an hour to go or something. I had played 30 games. I was beserking most of the games. So I play 30 games - 7 more than the winner who didn't beserk at all, but played throughout. So I ended up playing in any case 7 more games than the winner - but I find this aspect also strange that my average opposition would be that high given the larger game sample. I would intuitively think that for such a tournament, the more games one plays, the lower the average opposition in general.

Does anyone else find this kind of scenario a little unfair - and maybe the "average opposition" should be more evenly spread out a bit? Maybe the leaders of a tournament should be waiting a little more time after finishing a game to get tougher pairings if needed?

Feedback welcome from all

Two other recent posts mentioned that the two highest scoring players never played each other, and this was against the known pairing algorithm.

The highest scoring player should be paired with the next highest scoring player which is available. If this player happens to be a low rated player, this of course would lower the average rating field.

I find this system to be fair. I don't want to be matched by rating but by the performance of each player. If I didn't agree with this concept, I would never play in a swiss tournament.

Causing anyone to wait any extra time would be unfair more than your purpose to "even" the rating field.

I have had to wait between pairings in every tournament on lichess, although there were eligible players in the lobby. This is more of a solvable and relevant issue.

BTW, how can the performance rating be more than 400 points higher than the average opposition rating? (This should only happen when there is a perfect score.)

@Kingscrusher-YouTube I'm a fan of yours great vids.

And about the Arena. I think the problem is that you start playing from the begining.
The pairing system is okay because you get paired against playeres with the same tournament/arena score. So at the begining only strong players start scoring and you get paired against them, but if you late join you get a weaker oposition because strong players are already in top positions so you won't get matched against them.
Also the actual scenario is that only a few high rated players play all the tournament. They start and then they quit. If you start from the begining you get matched against them because you score at the same pace that them. Then they quit and the players remaining aren't that high rated, if a strong player late joins he will be picking points against all those remaining lower rated players. So if you late join say 30 minutes after the begining you'll get a much easier road to top 10.

#3 - Good point. And I appreciate the fun factor here.

I suppose this does make it more fun for late joiners, to see if they can catch up the leaders, and I have late joined some tournaments in the past and experienced the "fun ride" as well.

Perhaps a simple thing might be on the tournament winners section of the page, just to add some statistics tabs like:

"Top 3 vs 2000 average opposition"
"Top 3 vs 2100 average opposition"
etc for how many rating intervals exist

To celebrate the people that actually did play the tournament from the start and didn't mind playing strong opponents all the way through.

This might be an interesting thing to experimentally add and see.

If you compare my first 12 pairings to the tournament winner there is a strong contrast:

My first 12 pairings starting with a 2406 in Round 1:

12 zayobayo 2178 1
11 Fightingphil13 2077 1
10 xabichess 2144 1
9 Alexey_Golovchenko 2086 0
8 Ahmdalysary 1885 1
7 johan1 1826 1
6 IM Klimenko_Sergey_IM 2349 0
5 Hangemhigh 2251 0
4 velryba 2188 ½
3 sakri_syam_01 1716 1
2 atomarviento 1820 1
1 IcelordBg 2406 0

Tournament winner's first 12 pairings starting with an 1108 in Round 1:

12 CM Kingscrusher-YouTube 2209 1
11 StepanOsinovsky 2154 1
10 FM Irizanin 2282 1
9 yeumate 1796 1
8 fatemeyghhhh 1538 1
7 CM Kingscrusher-YouTube 2212 1
6 jack8bc34 1927 1
5 lyahov_gena 1933 1
4 LeoMura 1715 1
3 borjchetrang 1781 1
2 Novae3 1477 1
1 Jestaedt 1108 1

Yes the top 3 against a [rating bracket/highest rating bracket] should be nice. But there should be at least five games played to get there. Otherwise if someone accidentally beats IcelordBg in his first game and then quit he'll be apearing like the top performer in the Arena page.

#5 Yes - a minimum "sample size" as you suggest would be an important element on the draft proposal modification. Thanks for that point.

If the leader of a tournament would be waiting a little more time for the pairings, than other players, then it would put the leader of a tournament into disadvantage.

#7 Yes that seems anti-fun in retrospect - i prefer now the idea of just some more stats in the tournament prize winners page.

See #4

Thanks Lance for feedback. Cheers, K

I'm absolutely noob to tournament pairing systems, so take all what I have to say as brain storming by an outsider.

However, IF it is true, that tournament pairing works based on absolute performance, THEN it is also true that joining late gives an unfair advantage, since the absolute perforamce of the late player is necessarily lower ( due to less games played ).

The solution to this could be a performance window, which means that only last N games are taken into account for pairing, with N being relatively low number ( may be on the order of 10 ). In this way a late joiner catches up quickly in performance and cannot enjoy a long lasting pairing advantage.

Thanks Tryfon for your interesting posts and ideas.

Your idea about more stats in the tournament prize winners page makes a great sense, indeed.

Best wishes to your streams and very educational chess videos!
Lance