Time-dependent increase in phi, rating uncertainy with inactivity

[gamesguru]

See this page for a graph, but afaik it's common for rating uncertainties to decrease with frequent play, and increase with long periods of inactivity. Is there any way to accomplish this? What sort of parameter tuning would be required?

http://thetacklezone.net/glicko-new-naf-tournament-rankings-explained/

[danuker]

Sorry for the late reply ;)

(just kidding. I'm not apologizing. This didn't help you, but it might help someone in the future.)

This happens when you replace the players' ratings with their after-new-games ratings.

Note in the unit test for Glickman's example that plugging in the games to r1 returns the new rating rated.

r1 gets initialized with the default phi of 350, but all it takes is those three games to get it to 151.15... in rated, which means we got more confident about the player's mu (expected score) and will adjust it less with the next game.

Edit

I later realized that this was not at all what you were asking. In the paper, after step 8:

Note that if a player does not compete during the rating period, then only Step 6 applies. In
this case, the player’s rating and volatility parameters remain the same, but the RD increases
according to ...formula...

I guessed that to simulate a time step, you'd update the rating without any games played. And it turns out it works:

rating_later = env.rate(rating, [])