CryptRat
Arcane
- Joined
- Sep 10, 2014
- Messages
- 3,575
That's why the bayesian average is useful here.The thing is, a game rated as a 9 by two persons isn't the same as a game rated as a 9 by 50 persons.
Last edited:
That's why the bayesian average is useful here.The thing is, a game rated as a 9 by two persons isn't the same as a game rated as a 9 by 50 persons.
True. But then you also get spammers who vote for everything - even stuff they haven't played.Theoretically, that means that someone with less game played has more weight in his top games than someone who has a longer list of games. Of course, the latter could just as well distribute points to only a few of his favorites, but nonetheless, there's intrinsic tendency for the system to regard the preferences of those with fewer entries as more intense.
Fair enough.True. But then you also get spammers who vote for everything - even stuff they haven't played.Theoretically, that means that someone with less game played has more weight in his top games than someone who has a longer list of games. Of course, the latter could just as well distribute points to only a few of his favorites, but nonetheless, there's intrinsic tendency for the system to regard the preferences of those with fewer entries as more intense.
Mind you it'd mean re-coding the voting system I did last year and that's unlikely to happen soo...
All I need at this stage is a list of games.
Of course, spreading a budget of 25 points over games already gives the ranking implicitly, so the extra field would in that case redundant. But if you'd have that likert scale instead, the extra field would yield additional information.Kev Inkline: Adding another field wouldn't be a huge issue if required. Just that re-coding it so people have a limited number of votes (EG: spreading 25 points across all games) would be a slightly more involved effort (which is to say, do-able but I don't know my time availability over the next few weeks.).
I think this would certainly be one fruitful way to carry this out. The assigned points implicitly yield the rankings, with which one can do all sort of funny things.Why not do both? Have each Codexer rate each game they've played both on a 1-5 scale, AND have them assign each game 0-5 points out of 25 total. I don't think that would be overwhelming information overload.
Sounds very good! However, I don't know if you would like to restrict to grading 1-5 to only those games you give at least one point out of 25? Well, if we don't restrict that, it's one way of testing voters' consistency .Ok, how about this: we do the voting with the 1-5 scale, one page for each year, then the last page is a text box where people sort out their 25 points.
That way we get both methods, DU doesn't have do perform great l33t hacking feats and we can play data analysis to our hearts content.
EDIT 2: Here's the WIP list of games: https://docs.google.com/spreadsheets/d/1Onb7VZP-AfPZ1v4OfcptZgc3TSZqjf9G4xhjO51vANU/edit?usp=sharing
2012:EDIT 2: Here's the WIP list of games: https://docs.google.com/spreadsheets/d/1Onb7VZP-AfPZ1v4OfcptZgc3TSZqjf9G4xhjO51vANU/edit?usp=sharing
Can't it be with 1-10 scale? Otherwise There will be too many 3s. Or if 1-5 then allow +0.5 in-between.voting with the 1-5 scale
Can't it be with 1-10 scale? Otherwise There will be too many 3s. Or if 1-5 then allow +0.5 in-between.voting with the 1-5 scale
Maybe because unless you're a masochist you play a game of at least average quality (hence 3s are most common). In the case of 1s, these are usually games you couldn't finish or didn't even bother to play (like Fallout 3 or 4). 1-10 will at least let us better differentiate between an average game and above average game. When I see an average RPG which isn't that bad but isn't that good either I would rate it 3 instead of 2 but with 10 point scale it would be 5.The trouble with a 1-10 scale is that most people underuse the bottom half
2016:EDIT 2: Here's the WIP list of games: https://docs.google.com/spreadsheets/d/1Onb7VZP-AfPZ1v4OfcptZgc3TSZqjf9G4xhjO51vANU/edit?usp=sharing
Hyperdevotion Noire
It's an ordinal scale anyways, so it really doesn't matter if he bottom half is under represented.Can't it be with 1-10 scale? Otherwise There will be too many 3s. Or if 1-5 then allow +0.5 in-between.voting with the 1-5 scale
The trouble with a 1-10 scale is that most people underuse the bottom half, so that it is essentially a 6-10 (or 7-10) scale plus 1/10. A 1-5 scale accepts this reality in a more fitting way. Also, most games probably are 3s.
I think we must not try to do too much with one system. Simply aggregating ordered preferences (w/missing data) is already a huge challenge. If you want to account for magnitude of preference ("I really, really, really loved my #1 much more than my #2") I think it's taking on too much at once.
To whom?"recommend" or "not recommend"?
To whom?"recommend" or "not recommend"?
Meh, I think a 1 to 5 scale is completely sufficient. You can give DA:I a 5 ("riding the bull never looked that good") and NEO Scavenger a 1 ("that's not why I bought my new graphics card"), and everything else fits nicely in between (PoE 3 "gorgeous backgrounds, but why can't I move my camera?"). That's basically all you need.