Variations in the formula for TCI calculation

Ideas to improve Criticker and new feature requests, as well as announcements about new enhancements.
Freundh
Posts: 18
0 Ratings
Your TCI: na
Joined: Sat May 01, 2021 7:05 pm

Variations in the formula for TCI calculation

Post by Freundh »

I'm no statistician, so I don't know if these are good ideas, but I had some thoughts about how to calculate TCI.. Perhaps you've already done plenty of experiments and found that the current formula works best. Please don't take this as a complaint about how the site works :)

I was thinking that intuitively, it feels like my movie taste might be more defined by the movies I love or hate, and less by the ones I don't feel so strongly about. I.e. maybe it would generate even better TCIs if my top and bottom percentiles were weighted more than the ones in the middle?

The second idea I'm even less sure about, but perhaps if you have some kind of test fixture it might be interesting to use a formula similar to RMS instead of just plain average of differences? I think the effect would be that outliers (I mean films that me and the other person have very different rating for) would be have a bigger impact on TCI, and small differences in rating would matter less.

mpowell
Posts: 3897
1 Ratings
Your TCI: na
Joined: Fri Sep 09, 2005 10:22 am

Re: Variations in the formula for TCI calculation

Post by mpowell »

Thanks for those thoughts!!

We've experimented in the past with ideas similar to this, such as more heavily-weighting titles that are at the very top or bottom of the ratings list. But in the end, the gains didn't seem to be worth the added complexity to the algorithm.

That doesn't mean those ideas are discarded by any means! We've kind of got a permanently-open action item to review and revise our algorithms... right now, though, our focus is on modernizing the site, and we don't want to do anything to make Criticker more difficult to understand.

One thing we'd love to do in the future is allow for more personalization -- like, allowing people to tweak their own TCI algorithms, for example in the ways you're asking about. It'd be a cool way to experiment with live data... and would really help us learn what works best!

Freundh
Posts: 18
0 Ratings
Your TCI: na
Joined: Sat May 01, 2021 7:05 pm

Re: Variations in the formula for TCI calculation

Post by Freundh »

No worries, it was just some random thoughts and I wanted to throw them out there in case you wanted some inspiration for things to test :)

If you're working on making the site more accessible, please make it easier to find the import page.. I still don't know where it is so I have to use google to get to it :)

tev
Posts: 14
30 Ratings
Your TCI: na
Joined: Wed Mar 15, 2023 8:56 pm

Re: Variations in the formula for TCI calculation

Post by tev »

Freundh wrote:
Sun Sep 24, 2023 12:01 pm

If you're working on making the site more accessible, please make it easier to find the import page.. I still don't know where it is so I have to use google to get to it :)

"Resources" menu - then "Feeds and More" - then in the "Import Your Scores" heading, "Go to the Import Tool" link :)

Freundh
Posts: 18
0 Ratings
Your TCI: na
Joined: Sat May 01, 2021 7:05 pm

Re: Variations in the formula for TCI calculation

Post by Freundh »

Ah, it was hiding behind the "and more" :) Thanks!

paulofilmo
Posts: 2586
5 Ratings
Your TCI: na
Joined: Mon Apr 07, 2008 3:40 pm

Re: Variations in the formula for TCI calculation

Post by paulofilmo »

I love this,, it's similar to how I use kumpels (i think).

i was thinking how a small test batch would turn out, and how could it possibly not be an improvement(?)

Post Reply