Life In 19x19

On virtue of the IGS/Tygem style system is that it is easier to figure out how to throw in little hacks for those cases. Winning 19 of 20 games could put you in a special accelerated ranking change category that lets you rank up quickly.

In contrast for a KGS system, you can just change things arbitrarily, or you might corrupt the system's rating integrity. Perhaps there are ways to fix the potential for stuck ranks, but you can't just guess and hope they'll work.

hyperpape wrote:In contrast for a KGS system, you can just change things arbitrarily, or you might corrupt the system's rating integrity. Perhaps there are ways to fix the potential for stuck ranks, but you can't just guess and hope they'll work.

On KGS if you truly feel stuck, the easiest thing is to play games handicapped how you think they should be at your correct rating. Assuming you are correct, your rating should stabilize at that level fairly quickly. Unlike the other systems listed it does this with no penalty your opponents who helped you fix your rank.

oren wrote:On the opposite end, I just played on Pandanet IGS yesterday for a change of pace. I have a really old account and decided to check how many games it would take to 1d. My account is at 8k+ right now. Even if I win every game in a row, it would still take me 100 games to get the account to 1d.

One advantage of the IGS rating system is simplicity, as evidenced by your ability to make that calculation. Can you come up with a similar calculation for a KGS rating?

mitsun wrote:One advantage of the IGS rating system is simplicity, as evidenced by your ability to make that calculation. Can you come up with a similar calculation for a KGS rating?

Nope, I'm not defending the KGS system either. I think Pandanet may be the heaviest to move though. Tygem or wbaduk, 100 straight wins gets you to the top pretty fast. Also 100 straight wins on KGS on an account you went back to after leaving it stale for a bit would move even quicker.

oren wrote:

mitsun wrote:One advantage of the IGS rating system is simplicity, as evidenced by your ability to make that calculation. Can you come up with a similar calculation for a KGS rating?

Nope, I'm not defending the KGS system either. I think Pandanet may be the heaviest to move though. Tygem or wbaduk, 100 straight wins gets you to the top pretty fast. Also 100 straight wins on KGS on an account you went back to after leaving it stale for a bit would move even quicker.

No need to leave it stale, especially not if we're discussing a 100 game winning streak. There will be an initial lag with kgs because a player near promotion is expected to have a high winning %, but after that first promotion the confidence in your rating will drop, and you will rapidly continue promoting.

Mef wrote:

hyperpape wrote:In contrast for a KGS system, you can just change things arbitrarily, or you might corrupt the system's rating integrity. Perhaps there are ways to fix the potential for stuck ranks, but you can't just guess and hope they'll work.

On KGS if you truly feel stuck, the easiest thing is to play games handicapped how you think they should be at your correct rating. Assuming you are correct, your rating should stabilize at that level fairly quickly. Unlike the other systems listed it does this with no penalty your opponents who helped you fix your rank.

Is this necessarily true for small differences (you're one stone below your true rank)? I've never seen someone actually do the math on playing at handicaps other than those indicated by the ratings system.

In any case, my point was about the system itself, not user attempts to maximize their ratings change.

hyperpape wrote:In any case, my point was about the system itself, not user attempts to maximize their ratings change.

My apologies, I think I misunderstood the context your post was meant to be in...If you mean to say that with altered parameters (game half life, rating distribution width, etc) you may easily mess up the system's calibration, that is true.

Mef wrote:On KGS if you truly feel stuck, the easiest thing is to play games handicapped how you think they should be at your correct rating. Assuming you are correct, your rating should stabilize at that level fairly quickly. Unlike the other systems listed it does this with no penalty your opponents who helped you fix your rank.

Please tell me more about this. I always assumed that in a game with an incorrect handicap, the higher rated player would suffer a more severe rating loss than if he lost at the correct handi.

Is this documented somewhere?

Thanks.

wineandgolover wrote:I always assumed that in a game with an incorrect handicap, the higher rated player would suffer a more severe rating loss than if he lost at the correct handi.

I think that's what most people assume (including me till this very thread). Which can make it hard to get even rated games with higher rated players, because they may "fear" losing more.

It certainly sounds logical that if a solid 1d loses against a solid 13k, then it will have a greater negative impact on the 1d's rating than losing against another solid 1d would.

But I have no idea how the KGS rating math works.

I think what mef means is that if the 13k plays a bunch of games as a 1 dan, and wins a lot of them, the system will retroactively treat those games as a 1 dan beating another 1 dan.

That is, the effect of a game on your current rank is based on the system's current estimate of what the strength of the opponent was (at the time of the game). The effect on your rank is not based on the fact that when the game happened, it had previously estimated that player's strength as a 13k.

That's why the system is prone to ratings drift: if you beat someone and the system changes what it thought their strength was, you get a retroactive strength change.