QMDB Recommendations - Part 2

[Quicky]

Here you go, td888: http://thequicky.net/qmdb/td888-qmdb-re ... 080624.pdf
My recommendations perform as a 1.30 TCI for you, Criticker does slightly worse with 1.34. Oh and you love Clint Eastwood, Steven Spielberg, Harrison Ford, Robert De Niro, and Mel Gibson .

[td888]

I have to say, your system (algorithm) is working very well. I'm indeed a big fan of Gibson (as director), Tarantino, Eastwood and Clive Owen. Also I don't like Banderas (and Cruise, but I have to admit he makes good movies).

The top5 genre also fits my preference.

The one thing I don't get is Michael Cain in my botton 15. He's one of my favorite actors.... I think I gave most of his movies a pretty good rating, except for "On Deadly Ground" from Steve Seagal....

But for the rest, excellent info!

[Quicky]

Hmm... I did a quick check on Michael Caine. You've rated Caine's movies on Criticker a 70 on average. That classifies as "Not Good" in your ratings. I also don't have the movie 'Zulu' in my database yet, so this pulls the average down to 69. I've been pondering on this for the last hour however and I think the problem lies with the fact that I'm taking the average of your ratings which in this case is heavily influenced by one or two outliers. I'm thinking that maybe it's better to take medians (a pain in the ass to calculate) or to take the average of the rankings (pretty much like tiers). So I've rewritten some of the code so that now it will average the rankings for each name, and not the average rating. This should lower the effect of outliers like 'On Deadly Ground'. I'm testing the new code at the moment. I'll keep you posted .

But for the rest, excellent info!

How do you like the recommendations by the way? The extra info is all fun and interesting, but it's the recommendations that are the whole point of what I'm doing here. Do you like the new varied top recommendations? If so, why? If not, why not?

[td888]

When I was looking at all the movies of Caine, I saw that I didn't rank Zulu. So I ranked it...

Actually, the recommendations in general are not that interesting to me, because of all the old movies. As a rule of thumb I don't watch old movies (eg before 1970)... I know that there are some classics out there (they keep popping up as recommmendations from my kumpels and high in my PSI list), but I'm just being stubborn.

But the recommendations of recent movies seem very interesting. Those are movies I already decided I want to watch.

Maybe you can filter on movies newer than 1970? I'm curious what will pop up.

[Quicky]

I've added a line to my code in Excel that says: IF user.Trait = archaeophobic THEN min.Year.Visible = 1970
No seriously, I did add some code that allows me to put a minimum year of release in place for certain users. So if someone else wants this measure in place, let me know .

As for the Michael Caine issue... It turned out that when I changed my code, the performance of the whole system deteriorated significantly. So I reverted the changes except for a small detail and made a new list for you with the minimum visible year set to 1970: http://thequicky.net/qmdb/td888-qmdb-re ... 080625.pdf
(lol at your number 10 most hated keyword: "flatulence"...)

Let me know if that's better.

[td888]

Yes, these recommendations are more to my liking. Thanks!!

[paulofilmo]

MAJOR UPDATE!

I can't be on long, so I'll just say that I love these updates sooo much, and as soon as I get some films I've got waiting out the way and maybe move some rankings about, I'll be hounding you for an update.

much lovels

paul

x

[Quicky]

This is really cool, and I hope you can do one for me as well. I especially like your idea of calculating a TCI for both your recommendations and criticker's.

Here you go: http://thequicky.net/qmdb/peaceanarchy- ... 080625.pdf
For you it gets to a TCI of 1.48. Your lowest user TCI is 0.79 (is that the other person in your head or something? :p), but your second lowest user TCI is 1.65, so my recommendations do better than that at least.

Oh and these are you 5 most favourite names: Takashi Shimura, Akira Kurosawa, Billy Wilder, Toshirô Mifune, Tatsuya Nakadai. I'd say you should move to Japan .

[PeaceAnarchy]

Oh and these are you 5 most favourite names: Takashi Shimura, Akira Kurosawa, Billy Wilder, Toshirô Mifune, Tatsuya Nakadai. I'd say you should move to Japan .

Only if I can go back in time to meet Akira Kurosawa. The other three are mainly there because of his movies. The rest of the names are mostly what I expected, Rohmer, Bergman, Kubrick for directors and Stewart and Grant who can make anything enjoyable. I have no idea who Porter Hall is though. I'm surprised Nicolas Cage and Jim Carrey, Christopher Walken are in my bottom 15, since I really like them, but they do make some terrible movies sometimes.

I really like that last page. I'm surprised how evenly distributed everything seems, except for the fact that I've seen a lot more movies from the last 10 years, obviously.

[Quicky]

I'm surprised Nicolas Cage and Jim Carrey, Christopher Walken are in my bottom 15, since I really like them, but they do make some terrible movies sometimes.

Of course you have to realize that my system is trying to guess whether you'll like a movie based on the fact that a certain person was involved. If some of your favourite actors always turn up in movies that just plain suck, well... then it's probably a good guess that the next movie he or she will be in is also gonna suck. So... What I'm calculating is not who you love the most, but which people are most often involved in movies that you like, whether they made the movie good, or whether they just happen to run along.

I really like that last page. I'm surprised how evenly distributed everything seems, except for the fact that I've seen a lot more movies from the last 10 years, obviously.

I was quite surprised by that as well.