lichess.org
Donate

I suggest adjustments to the rating change in berserk games

Hello.

Berserk is a nice feature, but I don't think it should affect players' ratings the way it does. Players' ratings represent each one's strenght. When one berserks he has half the time and can't be as strong. As a result, people who berserk often will have A MUCH lower rating compared to those of the same strenght that never berserk. This alters ratings' reliability. In Rapid I've a rating higher of 97% of players, but since a lot play berserk there are probably a lot of players rated 2100 or even 2000 that actually are stronger than me (I'm around 2250 usually, actually I'm slightly below 2200). I'd like to have a more reliable rating.
Would it be an idea to give the berserking player a rating reduction of e.g. 100 points before used in the glicko equations and then apply the obtained rating change to the real rating? In this way a berserking player's rating would stay roughly the same even if his performance level is 100 points below his real rating during a series of berserking games. Also the berserking player's opponents would receive a little less points for draws/wins against him. Maybe this would be better for the rating system and to some degree make berserking accounts unnecessary.
Is the rating system is only designed to help lichess pair people in games?

And if you beserk, and your rating becomes lower than your actual strength does that matter? OK so Lichess might pair you against lower rated players because they match your lower(but not accurate) rating, but so what if you're just going to beserk against them too? And, while I don't beserk at all, I thought it was a tactic reserved for when you are facing a FAR lower rated opponent who you expect to crush easily and therefore don't need all your time and want double points?
Prof. Glickman surely had never imagined berserking when he devised Glicko rating system. Berserked games should be unrated (unless both sides berserked).
The problem is that some players might try using the button as a form of rating insurance.
The whole point of berserking is trying to win with less time, and therefore less chances. Also, lower rated players would berserk against higher players they don't have a chance against.
> "but so what if you're just going to beserk against them too? "
That case is not a problem, it is basically a berserking account then.

> "Prof. Glickman surely had never imagined berserking when he devised Glicko rating system."
True, maybe we need Glicko-3, a generalized handicap version :)

> "It's your own choice how much you berserk"
Well, yes of course it is...

> "The problem is that some players might try using the button as a form of rating insurance."
They would still have to perform well to maintain their rating. We are not talking about a huge rating compensation here and the difference would be hard to notice for each game.

> "Also, lower rated players would berserk against higher players they don't have a chance against."
Why is this a problem? It would just spice up the game. If they win they deserve the rating points, right?

Btw, why is mutual berserking allowed? There is no risk involved when both players berserk so why is there a reward? The reward comes both in terms of a possible extra point and being able to play more games. If both players want to berserk maybe the higher rated player could be the one to get the berserking rights?
Add this feature in the prioritization:
Sort the players by wins with berserk(B) and then wins without berserk.
Example 4 Winners two berserked in round one.
Round 2 Sorted
2200 (B)
2200
2150 (B)
2150

Round 2 Pairing
2200 (B) vs 2150 (B)
2200 vs 2150

Take the same approach to the players that drew and the players that lost.

The idea is assuming the Berserker has an under inflated rating.
With two equally rated players, the odds are equal only when the time is equal.

With only one Berserking it should be using the Blitz rating vs the Bullet rating numbers.
Pairings should try to match Blitz with Blitz and use the blitz rating even if its a rapid tournament.
I like the simple and easy to implement solution of #2.

Not to change the rating per se, but only to use a slightly lower rating during the calculation of the increase/decrease K.

Example, if a 2100 player (A) berserk against a 2099 player (B), then consider the rating as (A)=2000 and (B)=2099.

That means, without bersek, both would probably change very little, such as +2 for win/loss and probably +0 for draw.

But with berserk,

* if player (A) wins, lets say, it wins more points, say +10, ending up with 2110
* if player (A) loses, it loses less points, say -5, ending up with 2095
* Same for player (B), wining, it is his obligation, not much to gain. Say goes +5 to 2104.
* Player (B) losing, loses -10, goes to 2089

And in case of draw, it is not +0 anymore, but something like (A) keeps +1 = 2101, (B) loses -1 = 2098.

Of course, those figures are just to exemplify. The correct glicko formula is to be used as is, only considering a slightly less rating for the berserked player.

I think mathematically it makes a lot of sense to penalize, bonify such cases.

This topic has been archived and can no longer be replied to.