Suggestion to improve PSI algorithm

[Loerst]

I like the idea of PSI and find it very interesting. However I could not help myself but to notice, that it could be improved even further, mainly by taking in account some movie and user parameters, that are already included in the database.

For instance, PSI could take into account:

- release date: I noticed, that in my top 50% there're only three movies made before 1990, while they consist about half of my bottom 10%. This was not done intentionally and it took me some time myself to realize that, but very obviously does tell something about my taste. However, exactly half (10/20) of my top 20 PSI movies are old (<1990, some even <1950) movies. I know I could easily filter them out, but I do think, that they shouldn't be put so high on a list in the first place. With +180 films ranked I thought that issue will eventually clear by itself, however it doesn't.

- country and popularity: I love Turkey. I really do. It's a great country, with very friendly people, delicious food and tons of interesting things to see and I'm already looking forward to spending my next holiday there. But their movies are not something I would like so see, however they keep poping up in the first page of my PSI list. I guess the problem is, that in my top 1000 users there're always at least 3 Turks, that absolutely love some local film no one else ever heard of and gave it a 100 points, efectively catapulting it to the top of all charts. I don't know what's the criteria for a filter when deciding "popularity" of the film, but in my case more than 95% of all films are at least "well known" which I find somehow hard to believe.

- genre: everyone has his/her favorite genre and certain movie types that doesn't like so much. That could reflect in a PSI as well. Of course one could argue, that it already is presented in a PSI in a way, but I would give it some extra attention that would come from my previous picks, not opinion of my best TCI users.

- popular actors, directors: there're some people who pick a film according to their favorite actor or a director and some who don't really care and vice versa (that some actors are hated and films avoided because of that). I'm sure a simple algorithm could easily find out about them.

Suggestion:

After calculating PSI they way it is, there could be some additional points added or subtracted to the rating, like +-3 points for every of the above mentioned subjects. Of course that would mean 4 extra algorithms and consequently some extra calculations but I don't think it would be too time consuming as it could be done only when a user would change his ratings and it would be dependent only on user data, not on interaction with thousands of other users.

[somnivore]

I'm not sure that these factors are such that would truly belong in the PSI calculation. The way you have worded your suggestions, it sounds like there are certain properties of films you want to completely filter out.

You can currently create a filter for all those preferences and apply it to your PSI list (http://www.criticker.com/?fl&filter=op).

I would not want any of those factors to influence PSI as I am open to viewing any good films, regardless of genre, language, etc, and any correlation between PSI and such factors may be coincidental or indirect.

One purpose of the PSI is to measure factors that cannot be easily quantified or at least not easily queryable in a typical film database. You don't really need an aggregate score for factors that are easily quantifiable, I think.