| |
Perff
Administrator
Posts: 1684 |
New rating calculation.
As I posted in another tread I have made a new 'weighted average' function that tries to find and eliminate the fake (extreme) votes.
This should result in giving the down/up voters a hard time, but I don't promise a miracle function - it's just something I figured out for myself. :)
Because of this all the ratings have been recalculated using this new function, so all the charts have changed a bit, and the ratings have in general grown.
Any reactions?
|
|
| |
Nafcom
Registered: Apr 2002
Posts: 588 |
Good idea, so thanks and let's hope it will work!
|
| |
T.M.R
Account closed
Registered: Dec 2001
Posts: 749 |
Is there any plan to stop people voting for others in the same group as them...? [evil grin =-] |
| |
Puterman
Account closed
Registered: Jan 2002
Posts: 188 |
Yes, TMR's pathetic self-downvoting has to be stopped! |
| |
Dwangi
Registered: Dec 2001
Posts: 130 |
Perff: how does it work then? does it really work well? I first tried to upvote metalvotze. metalvotze has atm 6 votes on 1 that is 1.0 and i changed my votes to 10 which would with no doubt be an extreme vote, but the averge was 2.3.
Then i tried to downvote comajob by crest+oxy and i changed my votre to 1 and got an average of 6.8 so in those cases i dont think my extreme votes were spottet or am i wrong?
And yes i changed my votes back to normal again!
|
| |
Perff
Administrator
Posts: 1684 |
In those two examples I think the problem is that there simply is to few votes total to surly spot the extreme votes.
But some finetuning on my function might help. :) |
| |
Steppe
Registered: Jan 2002
Posts: 1510 |
I give Second Reality a 10 and the average rises to 7.3. I reconsidered my vote and eventually gave it a 9. Now the average is 7.4! Funny, ey? ;-) |
| |
Perff
Administrator
Posts: 1684 |
Heh.
What happened there was that there were so many 10's already (compared to the number of other votes) that my function desided that some of the 10-votes, including yours, must be a fake vote - and removed them when calculating the rating.
When you changed it to a 9 it was no longer considered a fake vote and then the average got higher.
:)
BUT! I'm not done working on my find-fake-votes-function yet!! |
| |
Slator
Registered: Jan 2002
Posts: 277 |
It is good to work on the calculation problem, but still there should be some auto-exposing of the lowest and maybe highest votes.
|
| |
Steppe
Registered: Jan 2002
Posts: 1510 |
Perff, I already figured something like this might be the case. :-)
Maybe you should exclude the public voters from your calculation? |
| |
Perff
Administrator
Posts: 1684 |
Hi there.
I just worked a bit on this function which tries to find crap-votes. This has resultet in all the ratings in CSDb have changed slightly - hopefully to some more correct values. :)
If anyone finds some ratings which looks very odd let me know. |
