@bruceelio said in #20:
> While Glicko does a good job of getting one's rating closer to their true performance Glicko has too much potential for overshoot in the process. And this can be strategically mined (abused) to attempt to hit various rating targets required for titles (2200, 2300, etc.). There is also potential for it to be abused for rating spots, historical records, etc. At least much more so than the existing Elo system.
You mean by deliberately not playing, getting a high rating deviation, resulting in more points gained by winning games to try to reach a target?
The simple solution is to only count rating targets for norms/records/rating spots by ensuring that when the person reaches the target, that their rating deviation must be below a certain amount. This would mean that their rating does not have high uncertainty when reaching the target. Or by having the person demonstrate continued performance above the target with a low rating deviation.
Their ratings will self-correct anyway. Plus people doing this will be obvious due to suddenly getting a high rating deviation from not playing. And it requires being lucky if they are truly not above a certain rating. And their rating will self correct. The measures above to ensure a certain rating deviation takes care of this concern.
@RuyLopez1000 said in #21:
> > While Glicko does a good job of getting one's rating closer to their true performance Glicko has too much potential for overshoot in the process. And this can be strategically mined (abused) to attempt to hit various rating targets required for titles (2200, 2300, etc.). There is also potential for it to be abused for rating spots, historical records, etc. At least much more so than the existing Elo system.
>
> You mean by deliberately not playing, getting a high rating deviation, resulting in more points gained by winning games to try to reach a target?
>
> The simple solution is to only count rating targets for norms/records/rating spots by ensuring that when the person reaches the target, that their rating deviation must be below a certain amount. This would mean that their rating does not have high uncertainty when reaching the target. Or by having the person demonstrate continued performance above the target with a low rating deviation.
>
> Their ratings will self-correct anyway. Plus people doing this will be obvious due to suddenly getting a high rating deviation from not playing. And it requires being lucky if they are truly not above a certain rating. And their rating will self correct. The measures above to ensure a certain rating deviation takes care of this concern.
Ok, that is the easy bit. Now show your work. We can start with fact that 2200, 2300, 2400 and 2500 are going to shift. Tell me what the new numbers will be. And the number of games played per month, per quarter and per year in order for a player to maintain the desired deviation to meet the title rating requirements. One could assume that at the higher titles will require a smaller deviation and hence the requirement of more games played on a more regular basis. Then graph the data showing how it compares to the existing Elo system over a varitey of number of games played and typical variance in playing strength. Now throw in the element of farming rating, buying titles, etc. that currently plague the existing system and let me know if the Glicko measures offer an improvement or make it worse.
Now, personally, I prefer Glicko based systems. Glicko provides much better accuracy of current peformance for a player but this comes at the expense of precision. And it would certainly be much more viable for juniors and players under 2000. But, once you start getting up into the ranges of titles there is likely some benefit to weighing precision as more important than accuracy. And Elo does this. The real problem is Glicko moves too fast and Elo moves to slow. Both are probably solved by playing 80-100 games per year. But, this isn't practical for everyone. Now, one could still easily hit desired ratings for a title with infrequent games with Elo but with Glicko this would likely not be possible, at least not without allowing for too much variance. So, there will be no slowly grinding one's way to a CM title.
Ha ha!
They say the whole point is that juniors are growing up too quickly and the system did not take this into account, and other highly intelligent arguments are given. In reality, everything is much more stupid, you have to think of an initial rating of 1000, taking into account the old system, when 1st rank is 1900, 2nd rank is 1700, 3rd rank is 1500. then what is 1000? oligophrenic rating???. And in general, in FIDE, on average, less prepared chess players receive ratings than, for example, on Lichess where they play everything, but even on Lichess the initial rating is 1500 and on Lichess the ratings are just balanced.
There was a complete mess going on with the FIDE ratings at the lower rankings, and FIDE noticed this only when this mess reached the grandmasters.
And even the fact that the initial rating was made at 1400 is not enough, and in my opinion, deflation continues.
@bruceelio said in 20:
> While Glicko does a good job of getting one's rating closer to their true performance Glicko has too much potential for overshoot in the process. And this can be strategically mined (abused) to attempt to hit various rating targets required for titles (2200, 2300, etc.). There is also potential for it to be abused for rating spots, historical records, etc. At least much more so than the existing Elo system.
@bruceelio said in 22:
>Ok, that is the easy bit. Now show your work.
@RuyLopez1000 said in #24:
> @bruceelio said in 20:
>
> > While Glicko does a good job of getting one's rating closer to their true performance Glicko has too much potential for overshoot in the process. And this can be strategically mined (abused) to attempt to hit various rating targets required for titles (2200, 2300, etc.). There is also potential for it to be abused for rating spots, historical records, etc. At least much more so than the existing Elo system.
>
> @bruceelio said in 22:
>
> >Ok, that is the easy bit. Now show your work.
Sure, once you answer my questions above I'll have the information I need.
The Elo system was built for GM's and stronger players above 2200 rating, so it's natural the masses getting into OTB chess are dissatisfied with a system that wasn't built for them and is being forced to take the strain. I think the Glicko-2 ratings that Lichess uses should be used up until a certain point, and then we can switch to Elo. They are more friendly towards beginners and don't mash people of wildly different skill together because there isn't enough of a range (FIDE has a lower rating of 1400 for everyone) but get skewed towards the higher end. Glicko-1, which chess.com uses, is simply worse than Glicko-2, but still better than the FIDE system.
R=R+K(S-S) -------> R=5, K=1, S=2 --------> 5=5+1(2-2) --------> 5=5+1*0 -------> 5=5+0 -----> 5=5