Glicko Rating System
From The Othello Wiki Book Project
Giuseppe Menozzi is the italian creator of the Glicko rating system used by the Italian federation. He gives below some explanations about this system:
The DAN system was proposed by me and accepted in a version changed by Francesco Marconi. Originally, I didn't put any range of 5 years, but the DAN points were constantly summed up for the whole life, while Francesco limited that to those 5 years. It was, I think, around 1998, and obviously the DAN have been calculated also for earlier players.
It's true, different ages players can be compared with DAN, but that's actually not its purpose. Instead, the idea was to have a rating based on important results, and not just on singular matches or tournaments. When you deal with a rating system, you always have to take into account many possible deviations and errors, although they generally are statistically rare. For example, someone can get many points playing only in a restricted and closed group of friends, without corelation with the rest of the players. Some other player could get points in light matches, and lose in important matches for tension, others feel the importance of the event and do the exact opposite.
The DAN system does not care how's the player behaviour during the year, it only considers top tournaments results.
There is also an historical reason that made the DAN request necessary. In the same period the old rating system (Privitera) was being migrated to the Glicko system, because it was creating too much inflation: there were too many Masters. For reasons that you can understand, the old Masters have all been kept in the Glicko system, although they would not have ever become such using the new rating. The DAN system does not suffer of any inflation, and it has been used to make a distinction among strong Masters and weak Masters.
About the Glicko. As I said, Biagio Privitera developed a very interesting rating system in the 80's (I think). It has been used a lot, and the ratings were published every 3 months on "Othello News", the old FNGO's bullettin (now substituted by Nero su Bianco, a smart name given by Alberto Viviani). That system was also well integrated with Benedetto Romano's program TORNEO, that was used to make a swiss-like tournament organization. Unfortunately, in that system there were corrections that led to inflation: players could gain as much rating as they wanted, but could not lose more than a certain amount. Not only every rating system has inflation, due to new players, this system unfortunately also introduced it artificially. With Donato Barnaba we decided it was the moment to change, and proposed two systems: he proposed classic Chess Elo and I proposed the experimental Glicko. For both, we created excel files and imported a huge number of previous matches, going back to early years of FNGO. Chess Elo is an intelligent system, that is studied to manage thousands of players and tournaments: it subdivides players in many different ranks, and for this reason we thought that it was not really the case of FNGO. Glicko is described here: http://en.wikipedia.org/wiki/Glicko_rating_system . Basically, it uses the Elo statistical prediction formula, adding a standard deviation, that impacts on how much both playing users' rating oscillate. The s.d. deviation is lower when a player is very active, and is high when a player is new, passive, or often plays with oscillating players. The important idea behind the Glicko is that it's not a Sum-Zero rating: one player can lose 10 points while his opponents gains 20. This apparantly gives less control on possible inflation, while actually the rating is much more stable on a statistical point of view.
I think that FNGO has been maybe the FIRST federation ever implementing that system, I personally talked with Dr. Glickmann for some information, then developed an excel file that automatically imported many different results (including TORNEO, Swiss Perfect and papp), parametrized it all to recalculate all the history of FNGO and give a suitable ranking to start in 1998. Donato Barnaba and I worked a lot to tune the 3-4 parameters, every time recalculating the whole FNGO history with new parameters. Then, the Glicko passed to "production", in parallel with Biagio Privitera, for several months or years: I don't know when it was finally decided to make it the official Italian rating system. Currently, it does not appear to have much inflation and its behaviour is very good also against closed groups, passive returning players.
I'm rather happy of this double-system, it actually should be seen as an unique system made of two modules: DAN and Glicko.
I also developed a similar DAN system for Happy End, but I never managed to keep it updated.
