weighted scores

Code junkies hangout here

Moderators: ChrisThornett, LXF moderators

weighted scores

Postby Spangwiches » Wed Oct 26, 2011 7:48 am

edit: This thread is in programming, just as one would expect.

I'm making a kinda endless flag web quiz/game. Nowt serious/clever, just for the fun of it. Here:

http://uoou.info/flags/

(probably won't yet work well on older browsers)

I originally had the 'score' as the %age of right answers but then people who only played a couple of times, with all correct answers, stayed at the top of the table.

So obviously I needed some sort of weighting. I want to reward people for a combination of playing lots and being right. I currently have:

rv + cm / v + m

r = player's %age right answers
v = player's number of plays
m = global average number of plays per player
c = global average %age right answers

(c is everyone's %age score added together then divided by the number of players NOT percentage of total right answers / total answers)

Ok so this is all fine. Works pretty well for the top and middle of the table:

http://uoou.info/flags/stats.php

But not at the bottom. The bottom player, with 63% right answers and lots of plays should ideally be somewhere mid-table I think. Certainly not below someone with 40% right and quite few plays.

I'm not great at maths. I kinda see what it's doing - punishing for high plays while being below the average %age. Can anyone with more of a maths-head than me suggest a way to rectify that sort of thing without messing too much with the top of the table?

Thanks!
Last edited by Spangwiches on Wed Oct 26, 2011 8:41 am, edited 1 time in total.
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am

Postby nelz » Wed Oct 26, 2011 8:37 am

Moved to Programming.
"Insanity: doing the same thing over and over again and expecting different results." (Albert Einstein)
User avatar
nelz
Site admin
 
Posts: 8504
Joined: Mon Apr 04, 2005 11:52 am
Location: Warrington, UK

Postby Spangwiches » Wed Oct 26, 2011 8:41 am

Thank you!
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am

Postby johnhudson » Wed Oct 26, 2011 9:06 am

What about rv/cm?

Putting r over c rewards those who get higher scores; putting v over m rewards those who play more often?
johnhudson
LXF regular
 
Posts: 876
Joined: Wed Aug 03, 2005 1:37 pm

Postby Spangwiches » Wed Oct 26, 2011 9:22 am

Just tried that and it's too closely tied to 'right answers', in fact it's exactly the same order and the resultant score is ~half 'right answers'.
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am

Postby Spangwiches » Wed Oct 26, 2011 9:25 am

... that sounds wrong. Maybe I did it wrong. I shall retry.
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am

Postby Spangwiches » Wed Oct 26, 2011 9:36 am

Yeah, just verified that. Not entirely sure why it behaves that way but it essentially gives a proportion of the right answers figure (not saying it actually does - it just tracks it perfectly).

edit: Actually, it makes perfect sense. %right x total answers is of course going to relate to right answers. Hmm.

Another edit:

just tried:

%player_right * player_right_answers / %global_right * global_total_answers

Now it overly rewards number of plays. Far too far in the other direction. Maths is hard.
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am

Postby kvonh » Fri Nov 11, 2011 9:40 am

Perhaps you could post it as a question in your quiz? :D
User avatar
kvonh
 
Posts: 19
Joined: Tue Apr 12, 2005 9:46 pm

Postby Spangwiches » Fri Nov 11, 2011 9:31 pm

Ahaha :D
User avatar
Spangwiches
 
Posts: 81
Joined: Thu Sep 22, 2011 2:50 am


Return to Programming

Who is online

Users browsing this forum: No registered users and 0 guests