Ratings Implementation - Githubissues

chaitu236 / TakServer

Server to handle online TAK games

https://www.playtak.com

GNU General Public License v2.0

35 stars 8 forks source link

Ratings Implementation #26

Open Abyssal-Tak opened 7 years ago

Abyssal-Tak commented 7 years ago

With this code, ratings should be operational, though there will be no way for any users to see their ratings (that is yet to be implemented). Currently, ratings are being recalculated from the "beginning of time" (April 23rd), but it's only a slight time waste at the moment, and you can probably fix that quite easily (I didn't fix it because I don't know how you want the data stored). Also, sorry for the ugly pull request, but I'm not a git expert yet, and it's not letting me pull your master branch so easily.

chaitu236 commented 7 years ago

Sorry for the delayed response - I was on travel. I remember we had a chat about this but its been a while and I forgot some of it, sorry if I ask the same questions again.

Why aren't we considering botgames for rating? - even if we should discard botgames, hardcoding botnames in server is a bad idea. We should either do that in the properties.xml file or have all names with Bot in them be classified as bots or not discard botgames at all.
Do we have live ratings? - as in are the ratings updated for each game?
I haven't looked in too much detail yet, but you're using only the ELO ratings right?

You can rename the database columns from r4..r8 to anything you want.

Don't worry about the aesthetics of the pull request. I can resolve the conflicts and merge. I'll keep going over the code whenever I find time. And thanks a lot for working on this!

Abyssal-Tak commented 7 years ago

1) Bot games are less reflective of actual human skill than human vs. human games are, and in my opinion we now have enough data that we can afford to not count them. As a compromise with Nohat, I do count bot games for bots, but not for humans.... that is to say, bot ratings are only comparable to other bots, and not to humans. I do agree that hard coding botnames into the server is bad, and we should move to a .xml or similar file in the future. 2) Yes, the Elo ratings are calculated live, though they are not dumped to database live (yet). 3) Yes, the Glicko are not operational at the moment, and I've decided to postpone using them, though I've still included them in the source code in case we ever want them in the future.

Also I probably should have refactored the names r4 through r8, but I'm sort of already accustomed to them so.... shrugs.