Elo Glicko Glicko-2
Doesn’t really matter tbh.
neither there tens of better. Elo is obviously slower as was made at the computation had to such that it can be carried out on paper with table of expected values
Well glicko 2 seems to be most advanced and realistic .Elo I would say is outdated and glicko 1 that chess.con uses somewhere in the middle but you'll never get the dinosaurs in FIDE to change to glicko 2 so we all stay stuck in the middle ages cause of dinosaurs and the old school boys if there was any vision from the chess world and bringing everyone together then all would use glicko 2 then we all could get along and avoid all the questions about comparing ratings xxx
You never said if it was to rate engines or people. @Srivaths2012
I like this rating systems.
www.sp-cc.de/eas-ratinglist.htm
Imagine downloading a tournament and using the tool to discover your own EAS rating.
www.sp-cc.de/files/statistics_eas_ratinglist.txt
EAS-Score of engines compared to humans. :))
I like this rating systems.
www.sp-cc.de/eas-ratinglist.htm
Imagine downloading a tournament and using the tool to discover your own EAS rating.
www.sp-cc.de/files/statistics_eas_ratinglist.txt
EAS-Score of engines compared to humans. :))
Glicko-2 is the best, but maybe we should switch back to Elo so I wouldn't have to yell at everyone when they wrongly say "Elo!"
I wouldn't say there is best one. I'm fine with all of them.
@SimonBirch said in #4:
> you'll never get the dinosaurs in FIDE to change to glicko 2
The "dinosaur" aspect may be part of the reason but IMHO the main problem is the attitude of general public. Even if the FIDE implementation of ELO feels like patchwork held together with a set of ad hoc tweaks (some of which have rather political reasoning and make the algorithm actually work worse - e.g. the "400 point rule") and arbitrary constants, for most people it's still easier to understand than Glicko-2. They get some idea how much the gain/lose depending on the rating difference and it's not the rocket science to get that once their K factor is reduced to 20 (10), this delta is one half of what it used to be.
Glicko-2, on the other hand, operates with more complicated formulas and the updates depend on (partially hidden) rating deviation. For most people, this is something mysterious and conspicuous. Just go through this forum, how often people ask how is it possible that they lost more points of rating than someone else with approximately the same difference. (Funny enough, I don't remember anyone reporting a problem after they get more or lose less. :-) )
I did some research and went through few of the documents and reports by people responsible for the technical background and they actually admit that this is exactly the main problem they have with Glicko-2: that it operates with the "hidden" parameter of rating deviation (or volatility) which makes it harder to accept by public. Which is quite sad if you think about it as this is exactly the feature that makes Glicko-2 work better than (FIDE) ELO even with all its hacks and tweaks.
> you'll never get the dinosaurs in FIDE to change to glicko 2
The "dinosaur" aspect may be part of the reason but IMHO the main problem is the attitude of general public. Even if the FIDE implementation of ELO feels like patchwork held together with a set of ad hoc tweaks (some of which have rather political reasoning and make the algorithm actually work worse - e.g. the "400 point rule") and arbitrary constants, for most people it's still easier to understand than Glicko-2. They get some idea how much the gain/lose depending on the rating difference and it's not the rocket science to get that once their K factor is reduced to 20 (10), this delta is one half of what it used to be.
Glicko-2, on the other hand, operates with more complicated formulas and the updates depend on (partially hidden) rating deviation. For most people, this is something mysterious and conspicuous. Just go through this forum, how often people ask how is it possible that they lost more points of rating than someone else with approximately the same difference. (Funny enough, I don't remember anyone reporting a problem after they get more or lose less. :-) )
I did some research and went through few of the documents and reports by people responsible for the technical background and they actually admit that this is exactly the main problem they have with Glicko-2: that it operates with the "hidden" parameter of rating deviation (or volatility) which makes it harder to accept by public. Which is quite sad if you think about it as this is exactly the feature that makes Glicko-2 work better than (FIDE) ELO even with all its hacks and tweaks.
@AsDaGo said in #6:
> Glicko-2 is the best, but maybe we should switch back to Elo so I wouldn't have to yell at everyone when they wrongly say "Elo!"
Rating result is still Elo, Glicko-2 just like Elo's original algorithm have same base model they just estimate parameters differently. End result will be the with sufficiently big amount of games. At the end of the day they all have 200pts difference mean that stronger player over lots of games will get 3/4 of the points.
> Glicko-2 is the best, but maybe we should switch back to Elo so I wouldn't have to yell at everyone when they wrongly say "Elo!"
Rating result is still Elo, Glicko-2 just like Elo's original algorithm have same base model they just estimate parameters differently. End result will be the with sufficiently big amount of games. At the end of the day they all have 200pts difference mean that stronger player over lots of games will get 3/4 of the points.
Fide otb...you can have elo rating lichess 2000 but still noob at real otb tournament
This topic has been archived and can no longer be replied to.