Uberdude, I think what shapenaji is saying, is that in order for a 4d to go on a 20 game win streak they have to be playing like a 5d, and if they go on a 20 game lose streak, then they are playing like a 3d. I personally agree. your rank should reflect how you are playing, not the accumulation of your games over the past months. If you think about it, it doesn't really make sense that how you were playing 45 days ago is half as important to your rank as how you are playing today. in order to provide accurate handicaps, ranks need to reflect the current playing strength of the player, not a strength that they can maintain over a long period of time.

Twenty game winning streaks are quite a thing. But how does the rating system handle the guy who wins twenty then loses ten?

_________________
Occupy Babel!

he is promoted and then demoted.

According to which standard? The accelerated one or the normal one?

_________________
Occupy Babel!

So I do see a pattern in advancement, most folks who are advancing tend to hop forward. They'll be at 16k then suddenly jump to 12k, then 6k, then 1d, then 3d

It's true that I jumped through those hoops rather quickly, but I see the same pattern in others, they just sometimes spend more time at certain barriers.

My hypothesis has always been that advancement comes with a paradigm shift in how you see the game, and when someone finds that, they suddenly jump.

As you say, the KGS system doesn't always deal with rapid advancement well. But the way I see it, if a person really is at a barrier, a system with less inertia will effect them very little. They might have a good day and go up a rank, or a bad day and go down one, but they'll hover around the same spot.

The risk, in this case, is that those players might end up playing 1 stone off their "proper" handicap. I don't actually see this as a problem. Folks will fight harder to win when there's more of a risk of falling down, and more of a gain in winning.

Also, being one stone off the proper handicap is unlikely to be a blowout, at any rank.

This would make ranks considerably more fluid, along the lines of what they are at clubs. Imo a person's strength variance is usually on the order of a stone, I don't see a huge problem with their actual rank variance representing that.

_________________
Tactics yes, Tact no...

See that's the thing, I don't differentiate between a 4d on a good winning streak and a 5d, or a 4d on a bad losing streak and a 3d.

As far as how KGS works. My understanding is that in the system the games are treated equally except for the older-game-weight. Hence, just by virtue of playing more games in the past than now, the past will have a far-reaching impact. Even though your results presently may be much better.

That's why I think it makes more sense to compute an estimate of a players strength on a shorter time period. Ignore the number of games played in that time period, and just take the time period as a single datapoint. If they played more games 45 days ago, then the sigma of that rating will be lower, but it will not treat it as a large number of separate games weighted half as much

You say that based on our previous information, these streaks are not significant.

If a person is ready to advance to the next rank, the previous information would be completely inconsistent. Averaging would just be silly, even a weighted average.

Hence, I think the best system is the one that looks for inconsistent behavior and then adapts quickly.

KGS is a good rating system for slow, steady, rank improvement. But I don't think I know any people at my rank who got here that way.

_________________
Tactics yes, Tact no...

If you play games at a (roughly) constant rate, it will take approximately the same amount of time to level up regardless of what that rate is. The amount of time it will take to level up will depend on how much of a deviation from expected you are performing at. I don't have the exact link off hand, but there is a sensei's page that has the trend for a player who was instantly x stones stronger ( I think it was two or three, but I could be wrong) playing at a constant rate on KGS. Where people get caught up is that they play 100 games/wk and it takes them 100 games to promote, then the see their account they play 10 games/wk on promote after only 10 games....and yet in both cases the actual time it takes is the same...about a week. You can skew it a bit if you don't play at all then suddenly play a lot of games, or if you play a lot of games, then slow down.

The other issue that arises from inconsistent game rates (not relevant to your post, but relevant to the discussion in the thread) is survivor bias. When someone starts a new account with very few games on it, the influence of an individual game is quite large. What happens much more often than people like to talk about is that this person starts 3 new accounts. After 3-5 games on each, one does exceptionally well and ends up two stones higher, one ends up close to their "stuck" rank, and one does poorly ending up a stone or two lower. Guess which account they talk about? Guess which account never gets played on again, and is never brought up in a conversation? Even if you do ask them about it it's often written off (I was tired/I was drunk/that was a sandbagger/that's my experimental account/that's my account only for blitz/pick your own reason for losing (this isn't to say that people don't make accounts for those reasons....merely that many "new" accounts get written off as such when they don't meet the person's expectations of their rank)).

People do not suddenly jump from 6k to 1d. That's a huge advancement. Most people will suddenly jump from 6k to 5k. Similarly, most will jump from a weak 1d to a strong 1d, to a weak 2d, etc. The KGS rank system will handle these jumps quite well.

And Mef's description is correct - if you play more often, then it will indeed take more games to move up or down, but as long as you play at a fairly consistent number of games per week, it will take the same amount of time to move up.

That kind of single-rank progression is likely to be observed on KGS, but I would argue that that's because it's not actually possible to observe the multiple-rank jump on KGS.

6k to 1d is a big advancement (6k to 3k is probably more likely), and happens all the time. People will go through a period of playing more games in person, more reviews, more tsumego. That won't be observed by the KGS system



This is why I suggested the data-point change, multiple games in short time period are all measurements of the same hidden variable, which I called "performance rating"

A lot of games in a short amount of time should ONLY have an impact insofar as it reduces sampling error.

_________________
Tactics yes, Tact no...

There are several of your posts I thought about making this a response to, I wasn't sure which fit best but it pertains to this idea --

I think there's partly a philosophical difference between the systems you are describing and the system KGS implements. The ranking system you describe, along with the ranking system for GMs, and go's own professional ranking system are meant to be backwards looking evaluations of past performance. For players at that level it makes sense...we get to the stage where we know they are all strong, we want to see who did better and we want to find out who was a winner in a competition. For an amateur tournament, or really any situation where your primary concern is evaluating past performance, this can be useful -- and it also has the benefit of all relevant data being in the past. A go server rating is not only about evaluating past performance however, it is also attempting to predict future performance.

You gave one example of a 4d going on a losing streak (where they should be called a 3d) and going on a winning streak (where they should be called a 5d) - but the problem is identifying and usefully handicapping all the games in between. At game 8 of the 10 game winning streak, the server has to weigh which is more likely - this streak will continue, whether there will be a regression back to average 4d, or something in between (strong 4d with a 60% winning % against average 4d). Further -- once either of these 10 game streaks is over, how useful is it in predicting future performance? Can we reasonably expect this player to continue at 5d or 3d? Looking at a player's history can be very helpful in predicting this.

I think what a lot of players on KGS are looking for is what you describe (an evaluation of how they have already played), but what KGS aims for is the latter (an estimate of how they will play, so the games can be handicapped properly).