Vortex Wars

-

Sorry to disappoint you but VW is using Glicko2 which was based on Elo rating system and is considered improvement because it takes more factions into consideration. Due to Elos "easy implementation" (it was calculated with pen and paper for chees games!) it has some nasty flaws which would make it even worse then our current rating system (fe. easier to cheat). Since we're using computers to do the math, we can use complex systems but... they need to be sable and implement perfectly. My guess is there is some trivial (yet, hard to find) error in Lopdos Glicko2 implementation (dangerous int->float cast? rounding problem? db storage "corrupts" values? dunno) - not system fault.

i agree with joriom on this.

I went through my implementation at least 3 times in detail, I even rewrote it from scratch once, but the bug is still there. I tried to google it a bit and I found few people complaining about similar problems so perhaps it's not as stable as people think it is. This is however unlikely and there is either a bug in the implementation or I am using it wrong (there are few specifics that might be screwing with it, but I was careful when designing them... one can never know for sure tho)

Lopdo wrote:I went through my implementation at least 3 times in detail, I even rewrote it from scratch once, but the bug is still there. I tried to google it a bit and I found few people complaining about similar problems so perhaps it's not as stable as people think it is. This is however unlikely and there is either a bug in the implementation or I am using it wrong (there are few specifics that might be screwing with it, but I was careful when designing them... one can never know for sure tho)

My reaction to lopdo posting

Lopdo wrote:I went through my implementation at least 3 times in detail, I even rewrote it from scratch once, but the bug is still there. I tried to google it a bit and I found few people complaining about similar problems so perhaps it's not as stable as people think it is. This is however unlikely and there is either a bug in the implementation or I am using it wrong (there are few specifics that might be screwing with it, but I was careful when designing them... one can never know for sure tho)

Is there any chance it could have something to do with you making it impossible to lose point when you win? Just a wild guess.

-

Ratburntro44 wrote:

Lopdo wrote:I went through my implementation at least 3 times in detail, I even rewrote it from scratch once, but the bug is still there. I tried to google it a bit and I found few people complaining about similar problems so perhaps it's not as stable as people think it is. This is however unlikely and there is either a bug in the implementation or I am using it wrong (there are few specifics that might be screwing with it, but I was careful when designing them... one can never know for sure tho)

Is there any chance it could have something to do with you making it impossible to lose point when you win? Just a wild guess.

It's possible, I really don't know. I guess I could do some experiments to determine where it goes wrong