search

lanesawyer / gshe

The Ginormous Stackrank of Human Experience
MIT License
3 stars 0 forks source link

Create algorithm that quickly determines the general area in which a new experience will ultimately reside #7

Open lanesawyer opened 9 years ago

lanesawyer commented 9 years ago

Generally, with a brand new experience we want to compare it to an established experience that resides in the middle of the list. If the new one wins, compare it with the middle of the top 50%. If it loses, the middle of the bottom 50%. Continue for a few iterations until it lands where it belongs.

Stackranker commented 9 years ago

That's a great description. Eventually we'll probably be able to establish some "anchor"- experiences that are universal (e.g. taking a nap or being yelled at) that very reliably rank at about .5, .25, .75, etc. in the stackrank. Since the anchors are universal, we can match new experiences against them to quickly place the new experience roughly where it will end up after hundreds of rankings. I bet this problem is already solved somewhere- not sure where we'd snag the code from, though.

Stackranker commented 9 years ago

I wonder if there's a developer or some code from the online gaming world that's solved this problem. They have a similar challenge- ranking relative skill among thousands of players from all over the globe. Either in tournaments or ongoing online battles, if we found a 1 vs. 1 game online (such as online chess) that already figured out how pair closely-matched opponents- we could probably copy their approach or code.

I note that Free Internet Chess Server might share that code/approach (http://www.wikiwand.com/en/Free_Internet_Chess_Server). Their contact info is here (http://www.ficsgames.org/about.html). Some of their interfaces are compared here (http://www.freechess.org/interfaces.html), and related contact info is here (http://www.freechess.org/contact.html).

The Glicko article also says: "Both Glicko and Glicko-2 rating systems are under public domain and found implemented on game servers online (like Free Internet Chess Server, Chess.com, Chesscademy.com, ChessTempo.com, PokemonShowdown, Lichess, Nodewar, SchemingMind, and Guild Wars 2[1]). The formulas used for the systems can be found on the Glicko website." (http://www.wikiwand.com/en/Glicko_rating_system)

Do you think this approach is productive? If so, how do we execute it?

Stackranker commented 9 years ago

If we don't end up adopting or adapting an existing solution, we can just execute the bifurcation exercise for, say, the first 10 times the new experience appears in an A v. B rating. Then, we can treat it just like every other experience (pair it according to our normal pairing approach #18).