#1: Kadou (Kadou) on Nov 18, 2010
Read some of the rating war thread, and got tired of it.#2: Teresa K (fasstar) on Nov 18, 2010
So what measures did Jan take in the end?
Did he change the way ratings are averaged?
I personally would require a unique email address per member. I vote per member. Suspicious voting could be looked after using the IP address.
Allow voting only after a puzzle is completed. People who print the puzzles (which only may be a few) who want to vote (which could only some of the few) can recopy their solution online to get to the completed screen where everyone votes.
Please let me now.
Kadou, if you read the rest of that thread, most of your questions would probably be answered, as Jan has explained what he did, how his program works, and why he does it the way he does. We have to remember that he does this for free for us and it takes a great deal of his time to straighten out some of these problems. He is pretty wise, as well as generous, and his rationale usually makes total sense.#3: Kadou (Kadou) on Nov 18, 2010
I'm surprised Jan made changes on the effects and not the cause of the problem, which was a lot more work.#4: Joe (infrapinklizzard) on Nov 18, 2010
The effects were obviously the falsified ratings.
The cause being MULTIPLE accounts rating puzzles NEVER actually solved.
A question remains: Are higher and lower ratings removed from the average calculation ? And if yes, how many?
Since the range of ratings is so limited (1-5), I don't think trimming the top and bottom values would help much after the first few voters. There just isn't room for egregious outliers.#5: Kadou (Kadou) on Nov 18, 2010
So the values can fluctuate wildly to begin with, but they settle down to a pretty constant number after about 15-20 voters. Thus puzzles won't show up on the "Best Puzzles" page until they have a minimum number of raters. (15 I think?)
Joe, I'm asking about trimming because this is the solution applied by Jan - not my idea.#6: Jan Wolter (jan) on May 6, 2011
Since you mention it, is the Best Puzzles really necessary? Anyone can use the Find puzzles and set it to find the best ratings AND sort them as wanted. Best Puzzles is the only tab from the Home page where results are by chronlogical (small # 1st) order and cannot be sorted otherwise. The NEW, SAVED, MY initially are sorted in reversed chronological order (big # 1st), but can be sorted by clicking any of the headers.
The Find Puzzles button on the left side menu is a repeat of the Find puzzles tab on the Home page.
A light touch of color in the background would make the site less paper-like and easier on the eyes. The puzzle would stand out more.
Jan made an extrordinary software, I only want to see it look better.
I voice my opinion, you are free to disagree.
It didn't really end. There are still people occasionally trying to tilt the ratings. The trick of eliminating some high and lows only helps a little. Mainly I built some new tools so that (1) I can tell the system not to count a user's ratings toward the score, and (2) I have various ways of scanning statistically for people who exist solely to warp the ratings. Between the two, I can eliminate most of these bad ratings in a few minutes, which, given that it presumably takes hours to create them, isn't too bad a deal. But it does need to be done manually, so it doesn't happen when I've been neglecting webpbn too long.
Goto next topic
You must register and log in to be able to participate in this discussion.