| I'm actually not quite sure about this suggestion, but I'll throw it out there. Last time I brought up the idea of changing around the Anime Compatibility index, it involved switching from comparing anime watched to just comparing scores. The system I think can still be improved a bit further, but I think it may be easier than last time. The biggest weakness of the current algorithm seems to be differences in the way people score. Basically, if one person tends to score higher on average and another tends to score lower on average, chances are not likely that those two would get a good compatibility rating. For example, if we had the following situation: You: Death Note - 7, Tokikake - 8, Fullmetal Alchemist - 9, Haruhi Suzumiya - 10 Person B: Death Note - 10, Tokikake - 9, Fullmetal Alchemist - 8, Haruhi Suzumiya - 6 Person C: Death Note - 4, Tokikake - 5, Fullmetal Alchemist - 6, Haruhi Suzumiya - 8 Compatibility with B: 9/4 = 2.25 Compatibility with C: 11/4 = 2.75 (I just took the top 4 anime on MAL guys, I'm not trying to start a flame war here =P) You obviously like Haruhi a lot better than Death Note, but Person B likes it the other way around. On the other hand, Person C likes Haruhi better than Death Note as well, and also rates the four anime in pretty much the same order you did, whereas Person B rated them in reverse order. Problem is, C rates in general a lot lower than most people, thus B gets a much higher compatibility rating. I've been wondering why I've had such low compatibilities with everyone who I consider have similar tastes, and the answer is because I score so low compared to most people. We already have an important piece of information we can use to incorporate into the algorithm, and that's the value at the bottom of every list: "kei-clone's scores average about 1.25 points lower than the average scores posted by other members." I'm hoping now that some mathematicians better than I here can figure out a way to do this (just like last time we changed the anime compatibility index). Now hopefully everyone understands what i'm trying to say first...>_> |
| I think see what you are saying. So basically we need to incorporate everyones standard on how they score and "[user] score y pts +/- than average" will help us figure out how to incorporate. Side note: "you" is exactly how i score those series. |
 |
| I'd personally rather work on recommendations...or about 10,000 other things first. |
| Well aside from the fact that xinil already stated that he isnt intersted in doing such a thing at the molment, another thing you would want to throw into the equation is that some people tend to have a few extream scores one way or another on just a few anime that goes agenst how most people score. Take how I scored Gantz for instance, I scored it a 2 for both seasons of it while most people tend to score it rather high and that tends to mess my scores up with most people. I think that if you were to drop the 2 highest divations it would also help get more accurate results. I also agree with weighting the scores based off of a users "Score Deviation" found in there anime list, but I think that droping the top 2 diviations would also help a great deal in the equation |
 "WoW is like a kiddie pool, it looks fun on the package, but is ultimately too shallow to enjoy." |
| what would the chances of having this implemented be if the full equation was figured out here (rather than having xinil figure it out)? Would it be a simple find/replace in the code? would that allow it to be easily put in place? |
| Don't know. I'd have to see what kind of equation you guys came up with. |
More topics from this board
» Interest Stack Suggestionquartz_goblin - Dec 7, 2024 |
1 |
by kur0ba
»»
1 hour ago |
|
» Export list to CSVlavratti - Dec 5, 2016 |
22 |
by deg
»»
11 hours ago |
|
» Martial PeakMauo7pK - Jul 17, 2021 |
22 |
by Darkonius
»»
Yesterday, 11:31 AM |
|
» Checkbox to mark that you watched the anime dubbed.EstrangeSonej - Yesterday |
0 |
by EstrangeSonej
»»
Yesterday, 9:47 AM |
|
» Hyperventilation (manga)Elisa1974 - Feb 2, 2024 |
4 |
by SewerBrat
»»
Dec 31, 2024 7:23 PM |