webdiplomacy.net

Hey, y’all. Do reliablity ratings rebound over time/subsequent games? Thanks.

As the missed turn goes past the month threshold it'll contribute negatively less, and then once it's older than a year it'll disappear

Ah, so missed turns mete out uniform hits against reliabilty, regardless of number of turns/time banked reliably? That is to say, a player who misses one turn out of thousands played is penalized the same as a player who misses one turn outnof dozens?

If you check your profile page it'll break it down for you. There's a small component that's proportional to the number of turns you played, but for most people the main part will be a fixed penalty for each turn missed.

I think that's a fair way to do it, because the other players in the game with you don't really care how many turns you didn't miss; if you play in more games it's still up to you to be reliable and not take on more games than you can reliably play in

I'm always open to other points of view though, but I think the RR system is pretty good

Thanks for the explanation. Your call, boss. Seems like tough break to play reliably for years then get saddled with a scarlet “U” for missing a couple of turns, but I guess that’s life in the NRL. Thanks for keeping this whole thing afloat for as long as you have...

kestasjk wrote: ↑

Wed Apr 05, 2023 11:00 am

I think that's a fair way to do it, because the other players in the game with you don't really care how many turns you didn't miss

Yes, they do. Or at least I do.

I would much rather play in a game with someone who has missed 20 turns out of 1000, rather than someone who has missed 10 turns out of 50. Even if the former misses a turn and the latter doesn't, I would still rather play with the former: which of those two is more likely to be reliable on future turns? (Isn't measuring that the entire point of reliability rating?)

That's true.. I guess the counter-argument is would you rather play someone who missed 10 turns out of 100 or 10 turns out of 50 (over the same time period); for me I'd consider both just as unreliable, the person playing 100 games shouldn't be considered twice as reliable because they're playing in more games

But I've got no dog in this fight; if most people prefer a different approach that's fine

I would still prefer the 10 in 100, because they're more likely to be reliable in my game. Whether it's in the same time period as the 10 in 50 or not, missing 10% of turns is still better than missing 20% of turns.

I think the likelihood of missing turns in my game is the desirable measure.

You can analyze the existing data to assess how the number of active games affects reliability, but it will also vary by player. Thinking of that makes the simplicity of just reporting the % of missed turns look like a virtue.