ccomeaux / boardgamegeek4android

BoardGameGeek application for Android (unofficial)
GNU General Public License v3.0
228 stars 47 forks source link

Suggestion for improvement in play stats #77

Open quasipedia opened 7 years ago

quasipedia commented 7 years ago
  1. The "skill" field should have an information pop-up as "utilization" has, explaining how it is computed.
  2. Introduce a mechanism to eliminate outliers in the skill computation.

More on suggestion 2: the skill computation seems to take into account the skill of the adversaries (good), but it lacks a mechanism to cut off outliers (bad). For example, in my stats for "Roll for the Galaxy" I have a skill score of 278 with 19/23 wins (82%). However, it just so happen that I lost one of my early games to a friend I normally never play with and who now - with a 1/1 (100%) wins - has a skill score of 600. Essentially the natural statistical noise is confused with a proper signal, and a stroke of luck with skill. Moreover, unless a re-match happens, this friend will now have his leading position uncontested forever.

Ideas on how to address this:

  1. Cut off the top and worse X% performances from the computation
  2. Provide a weight bonus on skill linked to the number of plays
  3. Require a minimum number of matches for a player's skill to be computed
  4. Read https://en.wikipedia.org/wiki/Elo_rating_system anc come up with something genius
  5. ...?