I dont find the recomendations/predicted ratings reliable. I have ranked more than 100 movies, but still there can be huge diffence between the predicted rating and the rating I would give the movie. Criticker predicted I would rate a movie 80, but I couldnt give it more than 30 after watching it tonight. When I look at the lists of recomendations, I sometimes wonder if it is just random selected movies - I cant tell that it should have anything to do with my taste at all (Cowboy movies...?!, a lot of movies from the 50s, action movies....which it thinks that I would rate very high) Is there anything I could do to make it more reliable? (I also use Jinni, which I usually can trust. I know that many people find Criticker more trustworthy. Maybe I just have a special taste that doesnt resemble any TCIs out there......)
111 ratings probably isn't quite enough to generate particularly reliable ratings, and the less conforming your taste is the less useful the PSIs will be. futhermore, criticker doesn't generate recommendations by the metadata of films (such as genres or actors), but only by the scores of your closest TCIs.
nonetheless if you're interested in genres, you can always go to
explore > film database > genre
then sort by # rankings, regenerate PSIs and see what happens.
Thanks! I will do as you both suggest, and this will also be good methods to find movies that I HAVE seen, so that I can quickly increase the amount of ratings (one obviously has to be a neird to use criticker) The hardest job is to find movies that one DOESNT like, its very difficult to recall those tittles, but I assume that it is important to have a lot of those too. Yes, I know it doesnt use metadata. I find it interesting to combine with jinni, high predicted ratings on both MUST be a garantied enjoyment. I often watch movies with a friend and would be great if Criticker could be used for finding movies that had high precicted ratings for both of us. If she makes a profil, I understand that I would be able to see the TCI that I have with her, but will I also be able to somehow find the movies that have high PSIs for both of us?
oplars wrote:Thanks! I will do as you both suggest, and this will also be good methods to find movies that I HAVE seen, so that I can quickly increase the amount of ratings (one obviously has to be a neird to use criticker) The hardest job is to find movies that one DOESNT like, its very difficult to recall those tittles, but I assume that it is important to have a lot of those too.
Under the Interact tab, there is a tab labeled Rank More Films. It's a useful tool for ranking a lot of films quickly when you're first starting out. Spend even an hour on that page and you're likely to get up to 500 or so movies ranked, at which point your PSIs will start to get accurate.
Also, feel free to ask for suggestions in the forums. The forums aren't super busy, but there are a number of knowledgeable movie enthusiasts who can probably make great recommendations for you while you build up your rankings on the site. Enjoy!
oplars wrote:I often watch movies with a friend and would be great if Criticker could be used for finding movies that had high precicted ratings for both of us. If she makes a profil, I understand that I would be able to see the TCI that I have with her, but will I also be able to somehow find the movies that have high PSIs for both of us?
Well if you're like me with nonconforming tastes I don't even use Criticker for the recommendations anymore. I use the site more to keep track of what films I've seen and how much I like them. I find the mini reviews to be very helpful as they get right into the thick of why someone likes or doesn't like a film. Also I'm a subscriber and I really like all of the stats Criticker keeps track of.
If you want a good recommendation engine the old Netflix engine is still available to DVD subscribers. Accurately rating movies on Netflix is the best way to get recommendations for someone like me.
They're never reliable for me either, and I've got over 6000 ratings 'we predict 85' = I give it 10. I don't really bother with the recommendation system, unless I'm curious as to what my kumpels agree on.
wau, 6000 ! You have given it a fair try then! I guess that for some it works, it probably all matters about what is the criteria for liking each of the films. Say it is because one likes movies which includes women with red hair, which a very unusual criterion, so probably its recommendations will be of no help. I had imagined it would be of more help. There once was a video shop in the city where I lived. Somebody told me that you could ask the owner for advice. He would ask you to tell him tree movies that you really like. Then he would pick one for you. I went there one day to try it out, told him tree favorite movies and he gave me "Moon", which I really, really enjoyed. This is just from tree movies. Not 6000. So I find it strange this doesnt work better (for me).
Netflix use Jinni, dont they?..... Jinni is much more precise (in my case), but still it is terribly wrong some times. It should contain some kind of filter, so that one could chose movies withOUT certain genes. It seems that it easy gets the impression that you like a certain gene if it has been in movies that you prefered for a totally other reason. There is no way to say, please no more sentimental movies, other than finding sentimental movies and give them lowest score. And they never answer any questions or comments in their forum.
I mean, there's always going to be issues with a system for recommendations that isn't intelligent. For instance I love time travel, so any time travel movie is going to get a higher rating from me than it 'deserves' but there's no way for the math behind the site to know that.
