I have a quiz online game. For each game there is a prize (number of scores). Now I have three ratings: top for the last day, top for the last week, total top (based on the sum of scores). Sum is not bad, but there is an issue: the more you play (even with bad results), the higher you are because more games leads to more points. I want to build a new rating, that takes in account not only sum of scores, but performance in games to. Just an average prize is not a good idea for me, because it means that winner of a single game will be higher, than a person who won 9 games and got not-maximum score in the last one.

Do you have any ideas/experience, what formulas could be used?



I'd probably ignore these scores and make a rating similar to the Elo rating system.

In short:

  • Have an invisible rating number for each and every player. They start at some fixed number, like 100 or 1000.
  • You then try to determine how good the player will be (either against other players he's playing with/against or simply "how many questions he'll answer right").
  • Once that's done, you let them play.
  • After the game compare the real result to the expected result calculated before.
  • The bigger the discrepancy, the bigger the change in the Rating will be. For example, you expected the player to guess 50% of the questions correct. If he did better (and wins), he's gaining points based on how good he was (only slightly better? -> less points; a lot better -> more points). In case of a lost game, you do the same - just reversed.

For a highscore or ranking, you can now compare these ratings. You could do things like "highest rating overall" or "highest gain last week". To avoid people stop playing to Keep a good position you simply lower their score over time if they don't play at all.

Just find a good name/system to actually visualize this. Just don't display raw numbers, as this might get People into number crunching, trying to manipulate it, etc.

  • \$\begingroup\$ I'd say definitely not the perfect answer, but still way better than the system the asker was using. \$\endgroup\$ – o0'. Feb 10 '13 at 10:21
  • \$\begingroup\$ Yeah, it has ist flaws, esp. given with the vague question, as we don't know how the actual quiz works. \$\endgroup\$ – Mario Feb 10 '13 at 11:06
  • \$\begingroup\$ Yeah. I felt compelled to write this comment as a compromise: I (personally) really hate ELO-like systems, but this deserved a +1 anyway. \$\endgroup\$ – o0'. Feb 10 '13 at 11:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.