Rating adjustments and unbalanced teams

If we want a perfect system, I'm not sure a purely linear calculation of rating difference is sufficient, either.  While two 1100s might be favourites against a 2000 and a 500, would the same be true of a couple 1600s against a 2500 and a 1000?  I'm not so sure; the 1000 might be able to hang on enough for the 2500's strength to tip the scales.  A 1000 doesn't have quite the same tendency to commit hari-kari that a 500 does.

People are forgetting that getting paired would be more tough when you add so many adjustments like that.

I never tried to claim my suggestion (with linear calculation of rating difference) was perfect. In my view reaching perfection isn't the goal, but rather trying to get as close as possible without adding too much complexity.

Yeah, I think your system is probably a step closer to perfection than the current straight average, which is itself a massive improvement on the previous system.  I'm not sure a perfect system is realistically achievable.

I think, though, that there's enough imprecision in an Elo system that you eventually reach the point where it's like putting caviar on a hot dog.

Do you really mean this, or the other way around? Because my blitz is 2200+ and bughouse is 1750 (and I have played quite a lot of bughouse) Did you play lots of games? 

I don't know of any 2500 in bughouse. Super GMs don't play bughouse. 

I think calculating rating on the basis of the geometrical mean of the players' ratings would be better... 

I wonder if that would be any good. Might make some good food to eat during Bughouse.

Nice idea, Martin0. I fully support such a rating system, and I agree wholeheartedly with MGleason's statement, "Yeah, I think your system is probably a step closer to perfection than the current straight average, which is itself a massive improvement on the previous system."

@Matir: I understand that. Bughouse ratings are a it deflated as it stands. However, some players might actually be better at bughouse than orthodox chess. There are some people who are extremely good at crazyhouse but only "decent" at orthodox chess.

And there are 2500s (or could be). If strong bughouse players from FICS played here, they'd reach 2500 easily. Super GMs aren't the ones you should fear...bughouse is a completely different game from orthodox chess.

@cwfrank: "I wonder if that would be any good. Might make some good food to eat during Bughouse."  LOL. You should try that and get back to us .

I put caviar on my shopping list the other day... for this exact purpose.

The only thing I'm having a hard time figuring out is how much time and money to spend on a project like this... 

... do I roll my own hot dogs out of some caviar mixture? (Yes, I know how to make sausage from scratch.) And, for that matter, since it's completely experimental, I'll probably have to use crap (fake caviar, not the real stuff, cheap crap). I even sent a question to a food scientist (albeit, confectionary), about whether or not there might be a fish membrane similar to pig gut for bundling a "specialty" item like this (you know, like making kosher or halal foods, do everything right and above board). Well, crap, now that I think about it, I know a Sushi Chef... she might know or have better ideas than the confectionary person, too.

... or, do I go with the original idea and just put caviar on top like a chili-dog?

Absurd, I know, but I will try and get back to you with a report since you asked for it.

Documentation... 

I agree with most of Croskie's analysis, except that all of B, C, D and E exclude the "A" factor (fix all the bugs... where one feature is introduced, there's likely another "A" bug that someone will complain about).

And, second... Croskie missed the F, G and H files (despite such high ratings, you can't neglect your F, G and H files). Like other missing bugs to complain about, this (agreeable) list (litany) of complaints requires attention to some basics.

Hahaha Frank you're a riot. Where's the report ?

Report: I am probably going to have to drive up north about 40 miles (resort town) to get some good caviar. Couldn't find anything I was happy with nearby. It's pretty much whenever I feel like making that kind of drive; and then when I feel like taking the time to do some experimental cooking if it's not the simple hot dog + condiment. (Although, I'm starting to think that a basic polish sausage, or maybe andouille might be better than a standard hot dog.)

See, now, this is what happens when you neglect the F, G and H files when programming... bughouse is broken today, and it's an "A" level issue. Whatever the reason for whatever change (B, C, D or E) ... "A" always comes into play, especially when you neglect F, G and H.

The programmers are missing their caviar hot dogs.

I'll get right on it. Caviar makes a wonderful condiment to add to food fights. If I throw some caviar at the people who caused the problem, do you think it might distract them, or help them work faster (as if rolling back bad code isn't simple enough).

I don't know; maybe do a test somewhere that doesn't risk negatively affecting chess.com?  Maybe go after Microsoft developers?

LOL.

What kind of chess player can afford caviar anyway.

About lopsided matchups (2000+500 vs 1100+1100, etc.):

Sometimes in such matchups, the 500 player is being told almost every single move by their strong partner. So their move quality won't be too bad, but they'll be a bit slow and the strong player's move quality will dip a bit due to focusing more on the second board than usual. Nonetheless, historically, such partnerships have tended to be successful from a ratings point of view.

Refinements to the rating system are fun to speculate about, but we can also collect a lot of data from actual chess.com games (and from the long history of publicly available FICS games, if desired) and see which conditions make upsets more or less frequent than the rating system predicts and consider adjusting the system from there.

For now, it's great that this change has been made - it's a huge improvement over the original system, so thanks!