Doubling the Canon 2021 Update - Results!

[Angel]

Welcome to our annual update.

What is this?
Begun in 2007, Doubling the Canon is a yearly project which aims to expand the Film Canon as crafted by the They Shoot Pictures, Don't They? website.

Who can participate?
Anyone can join. Requirement for newbies: Please note that you will be requested to identify yourself, just to make sure that the poll is answered (only once) by cinephiles of good will.

Where?
In this thread.
In the SCFZ thread: https://scfzforum.org/phpBB3/viewtopic.php?f=5&t=681
Via e-mail: doublingthecanon@hotmail.com

What is new in this update?
New Canon: TSPDT's 2,000 Greatest Films + 21st Century’s 1,000 Most Acclaimed Films.
New formula to calculate our top rated films: A Bayesian estimate instead of the former 'Adjustment for Popularity'.
Any film that is newly admitted to the Doubled Canon must be subjected to a revote after 5 years so we're going to revote the 2016 entries.

2020 update:
https://forum.icmforum.com/viewtopic.php?f=6&t=4790

Phase I: Nominations for titles not on either the TSPDT lists or the Doubling list.
Deadline: March 20.
Each participant can nominate a maximum of 12 films. Shorts, TV-movies, and TV mini-series are eligible. Every nominated movie goes onto the roster as a nominee. Films released in 2020-21 aren't allowed this year for our exercise.
Along with a list of the titles, please include the corresponding list of bare IMDb urls for the same films.
Please, don't edit your post! To change your ballot (undesirable) or add new titles you have to write a new reply.

Phase II: Ratings.
You're invited to rate the films you have seen (whether you have participated in the phase I or not). The nominees with the highest ratings make it into our list.
You can cast your votes anytime between now and Monday, April 26. Only 'write-in' updates will be accepted after the deadline.
Please, sign your ballot. Example: This is Angel Glez.

Rating system: 0-6 (3.90 was the cutoff last year)

Procedure 1: Just download the ballot (if you prefer I can send you the file) and email it back to me (doublingthecanon@hotmail.com)
DDL: https://we.tl/t-4UpUvGPf5h
Procedure 2: Online (PM me for details)

Don't remove anything, don't reorder. Untouched ballot with ratings, please.

After submitting your ballot, post a reply here to make sure I received the file. You can also add comments or statistics if you wish.

Early ballots would be appreciated. Then, if you watch new films or forgot to rate some, send me a 'write-in' ballot.

Thank you in advance for joining us in this game.

P.S.: If you know anyone adventurous enough (on or off the forum) that you think would also enjoy participating, feel free to invite them as well.


***********************************************************************************************************************


Results (ongoing)

https://1drv.ms/x/s!ArCAuhmbxCeRgQdbFut ... I?e=ndCRqh

https://forum.icmforum.com/viewtopic.ph ... 40#p708620

DtC list 2021 (ongoing)

https://1drv.ms/x/s!ArCAuhmbxCeRgQwhDIt ... Q?e=H4fahD

https://forum.icmforum.com/viewtopic.ph ... 60#p709721

Letterboxd

ICheckMovies

[Angel]

TSPDT The 2,000 Greatest Films (2021)
https://www.theyshootpictures.com/gf100 ... _table.php
https://www.icheckmovies.com/lists/they ... yquestion/
https://www.icheckmovies.com/lists/tspd ... 1001-2000/

TSPDT The 21st Century's 1,000 Most Acclaimed Films (2021)
https://www.theyshootpictures.com/21stc ... _table.php
https://www.icheckmovies.com/lists/the+ ... med+films/

Doubling the Canon (2020)
IMDb list
iCM list
Letterboxd list

All lists in the same file (TSPDT 2021 + DtC 2020)
https://docs.google.com/spreadsheets/d/ ... sp=sharing

Search links

Search link: https://www.imdb.com/search/title/?titl ... 2019-12-31

Nominations 2021
https://www.imdb.com/list/ls089692158/
https://letterboxd.com/angelg/list/doub ... -nominees/

[greennui]

The Flying Train (1902)
Twice Lived (Max Mack, 1912)
Nus Masculins (Francois Reichenbach, 1954)
Salomé (Pierre Koralnik, 1969)
Crash! (Harley Cokeliss, 1971)
Passing Strangers (Arthur J. Bressan Jr., 1974)
Chilly Scenes of Winter (Joan Micklin Silver, 1979)
My Eye is Reflecting (Nina Hedenius, 1992)
Vassourinha: The Voice and the Void (Carlos Adriano, 1998)
Parallel I-IV (Harun Farocki, 2014)
American Reflexxx (Alli Coates, 2015)
Knife+Heart (Yann Gonzalez, 2018)

Links:

https://www.imdb.com/title/tt0390011/
https://www.imdb.com/title/tt0002592/
https://www.imdb.com/title/tt12470144/
https://www.imdb.com/title/tt0355989/
https://www.imdb.com/title/tt2420990/
https://www.imdb.com/title/tt0071975/
https://www.imdb.com/title/tt0079278/
https://www.imdb.com/title/tt13842418/
https://www.imdb.com/title/tt0334662/
Not on imdb
https://www.imdb.com/title/tt10570744/
https://www.imdb.com/title/tt6426028/

[sally]

every year i say i can't do this, then i decide to do this and then i don't. so this year i'm gonna say...the TSPDT link above links to a starting list of 20,430 films apparently. (how did they compile that? cuz there's some films there impossible to see and obvious ones missing) or there's a link to the 1000 greatest films, but where did the 2000 films come from?

am going to wait for angel's coming soon all lists in same file

[MrCarmady]

there's two TSPDT lists on ICM - one is the classic 1-1000 list, one is a 1001-2000 extension they've added recently, which means DtC has gotten that much more obscure as a lot of the 'canon but not quite mega-canon' stuff was in that 1001-2000 bit according to their methodology of ranking the 20,430 total
then there's the third TSPDT list that's specifically 21st century and overlaps with those two a bit but i'm guessing you're not that bothered about it anyway?
and then of course there's DtC itself

i submitted stuff last time and got one on the list but the rest got brutally slaughtered so gonna put a bit more thought into it this time. can't believe they did fallen angels dirty like that...

[thoxans]

since none of my noms ever make it, i just like to use this as an opportunity to downvote every other nom on the list

i find that revenge is a dish best served in an excel spreadsheet

[sally]

there ought to be some kind of formula to weight your votes with how old & cranky you are. i used to be way more openminded, but now i can generally guess if i'm going to like something or not. i don't know if that's experience or just brain calcification but i don't feel as impartial as i used to be. it doesn't seem fair really

also i know there's no data and it shouldn't matter but i'd really love to know the gender ratio of participants.

[Silga]

1. The Harder They Fall (Mark Robson, 1956) - https://www.imdb.com/title/tt0049291/
2. They Drive by Night (Raoul Walsh, 1940) - https://www.imdb.com/title/tt0033149/
3. Affliction (Paul Schrader, 1997) - https://www.imdb.com/title/tt0118564/
4. Watchmen (Zack Snyder, 2009) - https://www.imdb.com/title/tt0409459/
5. The Reflecting Skin (Philip Ridley, 1990) - https://www.imdb.com/title/tt0100469/
6. The Big Blue (Luc Besson, 1988) - https://www.imdb.com/title/tt0095250/
7. The Grey Fox (Phillip Borsos, 1982) - https://www.imdb.com/title/tt0085622/
8. Howards End (James Ivory, 1992) - https://www.imdb.com/title/tt0104454/
9. Streets of Fire (Walter Hill, 1984) - https://www.imdb.com/title/tt0088194/
10. The Heart Is a Lonely Hunter (Robert Ellis Miller, 1968) - https://www.imdb.com/title/tt0063050/
11. The Last Days of Disco (Whit Stillman, 1998) - https://www.imdb.com/title/tt0120728/
12. I Served the King of England (Jirí Menzel, 2006) - https://www.imdb.com/title/tt0284363/

[greennui]

also i know there's no data and it shouldn't matter but i'd really love to know the gender ratio of participants.

Had a look at the last years thread and I counted 5 people that I know are women out of 90 voters, would have been 6 if you had voted.

[Angel]

since none of my noms ever make it, i just like to use this as an opportunity to downvote every other nom on the list

i find that revenge is a dish best served in an excel spreadsheet

The less successful you are, the more important your contribution is.

I repeat it every year like a mantra. Otherwise I wouldn't host the show.

[Curtis, baby]

can't believe they did fallen angels dirty like that...

the wkw? i would give that maximum points

[flip]

New formula to calculate our top rated films: A Bayesian estimate instead of the former 'Adjustment for Popularity'.

curious about this, if you have a chance to clarify. what was the previous formula and what is the new one?

[flip]

and if we're 'doubling the canon', and the tspdt 'canon' is now 2000+ films, shouldn't the 'doubling' list also be 2000+ films?

[Curtis, baby]

Don't know the numbers but I believe the more votes a film recieved the higher average vote it had to receive to make the list

[flip]

Linda Linda Linda (Nobuhiro Yamashita, 2005)
Film, Film, Film (Fyodor Khitruk, 1968)
Anantaram (Adoor Gopalakrishnan, 1987)
Blue Mountains, or Unbelievable Story (Eldar Shengelaya, 1983)
Play (Gholam Reza Ramezani, 2005)
Ley Lines (Takashi Miike, 1999)
The Well (Russell Rouse/Leo Popkin, 1951)
Ladies in Retirement (Charles Vidor, 1941)
The Girl Chewing Gum (John Smith, 1976)
Josef Kilian (Pavel Juracek/Jan Schmidt, 1963)
Mandala (Im Kwon-taek, 1981)
The Toth Family (Zoltan Fabri, 1969)


imdb urls:
https://www.imdb.com/title/tt0468795/
https://www.imdb.com/title/tt0212943/
https://www.imdb.com/title/tt0092557/
https://www.imdb.com/title/tt0087349/
https://www.imdb.com/title/tt0490803/
https://www.imdb.com/title/tt0258766/
https://www.imdb.com/title/tt0044202/
https://www.imdb.com/title/tt0033802/
https://www.imdb.com/title/tt0125118/
https://www.imdb.com/title/tt0057421/
https://www.imdb.com/title/tt0082704/
https://www.imdb.com/title/tt0064498/

[Curtis, baby]

Linda Linda Linda (Nobuhiro Yamashita, 2005)
Film, Film, Film (Fyodor Khitruk, 1968)

the beginnings of an interesting "Genre"

[Curtis, baby]

Don't know the numbers but I believe the more votes a film recieved the higher average vote it had to receive to make the list

or mb it was the other way around? not sure which it was, or even which that i think makes more sense, tbh. hmmmm angel??

[rischka]

it was the other way around. a film with very high scores and a few votes can survive as a holdover. a film needs 8 votes to enter the canon

the cut off score is decided by number of voters/films submitted? I think?

[thoxans]

what was the previous formula and what is the new one?

x2 curious about this as well

[flip]

the linda linda linda / film film film duo was intentional

i know how bayesian rating estimators work, and i'd be concerned that by using one, dtc would be hurting the chances of highly-rated 10-views films, the kinds of obscure gems that i've always thought dtc did such a great job of highlighting in the past. as i understood the previous system, as long as a film reached a certain views threshold (around 7 or 8 ratings in total), then its average rating was all that went into the final ranking. when a film didn't meet the minimum view threshold, if it had an especially high average rating, it could hold over to the next year, and be subject to a new vote that might get it over the bar. so there was a cutoff, but a film with 12 views with an average rating of 4.5 would do better than a film with 36 views and an average of 4.4, as i understood things. that is very unlikely to remain true using a bayesian estimator.

when using a bayesian rating estimator, you set the value of a constant x, which you could sort of think of as a number that represents the minimum number of ratings you'd want so that you felt you had evidence about how people feel about a film. so for this exercise, i'd imagine that constant would be set to a value around 8 or 10. then you just assume that there are x people who gave every single film an exactly average rating. then you use all the genuine ratings, along with, for each film, these x brand new average ratings, and compute each film's new 'bayesian average' from there, in the normal way. so it's just a weighted average calculation, one that regresses every single film's rating towards the mean, but which has a more pronounced effect the lower the number of genuine ratings a film has. if the constant x were set to 10, for example, and if 3.0 was the average rating for a film, and you had two films:

film A: 10 ratings, average of 5.0 --> new bayesian rating is 4.0 (just a weighted average of 10 ratings of 3.0 and 10 ratings of 5.0, which is a straight average in this case)
film B: 30 ratings, average of 4.6 --> new bayesian rating is 4.2 (just a weighted average of 10 ratings of 3.0 and 30 ratings of 4.6, which produces a weighted average 'three times closer' to 4.6, since that group has three times as many ratings)

notice film B wins. philosophically i don't think i agree with using bayesian estimators in this way because i think they're based on an incorrect philosophy about bayesian priors in this context, but i'd need to think that through a bit better to either articulate why or to revise what i think.

[Angel]

Quoting lee-109, DtC's father (Classic Film Board, 2007):

Adjustment for Popularity
The Doubling exercise appears to have a built-in bias towards movies that have been seen by relatively few people -- movies that even most Classic Film buffs consider obscure. To slightly lessen this effect, there has been a small adjustment for popularity built into the exercise which increases the vote averages for movies that have been seen by many.
If a movie receives 10-19 votes, we drop the lowest vote. If a movie receives 20-29 votes, we drop the two lowest votes. If a movie receives 30-39 votes, we drop the three lowest votes. And so on...

This adjustment is capricious and rather unfair, but at the same time it's part of what gives the DtC its unique feel.

The Bayesian estimate works as flip has explained. I'm not sure this is a better adjustment method, but at least it's not capricious.
What can we expect? A different ranking, that's for sure, but the films would be mostly (I think) the same.
Coeur fidèle, with 43 votes, would be the new #1, whereas Himmel und Erde, with only 11 votes, would drop to #25, but it would still be there.

[Angel]

and if we're 'doubling the canon', and the tspdt 'canon' is now 2000+ films, shouldn't the 'doubling' list also be 2000+ films?

I invite you to try.

[sally]

i can't tell if any of these are on any lists or not already and i used the stupid titles on imdb (mostly)

The Beauties of Italy, Triptych of Picturesque Views (Piero Marelli, 1912)
Blue Blood (Nino Oxilia, 1914)
Le brasier ardent (Ivan Mozzhukhin/Mosjoukine, 1923)
Abwege (G W Pabst, 1928) (who the hell calls it the devious path?)
West of Zanzibar (Tod Browning, 1928)
Dood water (Gerard Rutten, 1934)
Mark of the Vampire (Tod Browning, 1935)
The Sacrifice (Gösta Werner, 1946)
Monelle/Les amoureux sont seuls au monde (Henri Decoin, 1948)
Banketten (Hasse Ekman, 1948)
Aila, Pohjolan tytär (Jack Witikka, 1951)
Prebroyavane na divite zaytzi (Eduard Sachariev, 1973)

https://www.imdb.com/title/tt1469332/
https://www.imdb.com/title/tt0201063/
https://www.imdb.com/title/tt0293777/
https://www.imdb.com/title/tt0018615/
https://www.imdb.com/title/tt0019563/
https://www.imdb.com/title/tt0025060/
https://www.imdb.com/title/tt0026685/
https://www.imdb.com/title/tt2270454/
https://www.imdb.com/title/tt0039150/
https://www.imdb.com/title/tt0040148/
https://www.imdb.com/title/tt0121370/
https://www.imdb.com/title/tt0177138/

[john ryan]

It Rains in My Village! (Aleksandar Petrovic, 1968) https://www.imdb.com/title/tt0124294/
High Hopes (Mike Leigh, 1988) http://www.imdb.com/title/tt0095302/
The Twentieth Century (Matthew Rankin, 2019) http://www.imdb.com/title/tt7969836/
Amateur (Hal Hartley, 1994) http://www.imdb.com/title/tt0109093/
The Connection (Shirley Clarke, 1961) http://www.imdb.com/title/tt0054763/
Goodbye CP (Kazuo Hara, 1972) http://www.imdb.com/title/tt0234669/
The Chess Players (Stayajit Ray, 1977) http://www.imdb.com/title/tt0076696/
Savage Messiah (Ken Russell, 1972) http://www.imdb.com/title/tt0069225/
Yum, Yum, Yum! A Taste of Cajun and Creole Cooking (Les Blank, Maureen Gosling, 1990) http://www.imdb.com/title/tt0100997/
Have Sword, Will Travel (Cheh Chang, 1969) http://www.imdb.com/title/tt0064067/
Appointment in Honduras (Jacques Tourneur, 1953) http://www.imdb.com/title/tt0045512/
Who's Camus Anyway? (Mitsuo Yanagimachi, 2005) http://www.imdb.com/title/tt0456873/

[flip]

Quoting lee-109, DtC's father (Classic Film Board, 2007):

Adjustment for Popularity
The Doubling exercise appears to have a built-in bias towards movies that have been seen by relatively few people -- movies that even most Classic Film buffs consider obscure. To slightly lessen this effect, there has been a small adjustment for popularity built into the exercise which increases the vote averages for movies that have been seen by many.
If a movie receives 10-19 votes, we drop the lowest vote. If a movie receives 20-29 votes, we drop the two lowest votes. If a movie receives 30-39 votes, we drop the three lowest votes. And so on...

This adjustment is capricious and rather unfair, but at the same time it's part of what gives the DtC its unique feel.

The Bayesian estimate works as flip has explained. I'm not sure this is a better adjustment method, but at least it's not capricious.
What can we expect? A different ranking, that's for sure, but the films would be mostly (I think) the same.
Coeur fidèle, with 43 votes, would be the new #1, whereas Himmel und Erde, with only 11 votes, would drop to #25, but it would still be there.

First, that old adjustment is effectively just a modest randomizer. There's no reason to think it would achieve its intended effect; If a film has received 10 votes or if it has received 30 votes, in either case the old adjustment discards the worst 10% of the film's ratings. Assuming those low ratings are zero, the two films come out the same if they were otherwise equally rated; the more-viewed film doesn't win. It's also clearly the wrong thing to do just by considering limit cases - say these were all the votes for three different films:

film A: 18 votes of 6.0
film B: 18 votes of 6.0, 1 vote of 0.0
film C: 18 votes of 6.0, 2 votes of 0.0

I think any reasonable person would agree that in the final rankings, A should finish first, B should finish second, and C should finish third. With the old popularity adjustment, they all end up tied.

The old adjustment was bizarre because it helped films with 10-11 votes, or 20-21 votes, or 30-31 votes, much more than it helped films with 18-19 votes, or 28-29 votes, or 38-39 votes. Films really shouldn't get a bonus just because you get a small remainder when you divide their number of votes by 10.

So a more progressive (and less capricious, I guess) adjustment does seem better. But the philosophical premise behind using a Bayesian adjustment is, now that I've thought about it further, completely wrong (and I read the discussion on the ICM forum about it, or at least what I found near the end of the 2020 DtC results thread, and that discussion was often very misguided, mathematically speaking). I'll post about that later.

[sally]

Coeur fidèle, with 43 votes, would be the new #1, whereas Himmel und Erde, with only 11 votes, would drop to #25, but it would still be there.

i would thoroughly approve any system that gets himmel und erde out of the final list altogether.

[greennui]

i would thoroughly approve any system that gets himmel und erde out of the final list altogether.

What's so bad about it? I haven't seen it but I've come across it a lot due to it being at the top of this list. I've never been able to decipher what the film is actually like which is why I haven't watched it. Not that keen on just diving into a 285 min film.

[sally]

oh well, you're probably best off asking someone that likes it about it? i just didn't get anything from it other than oooh technology bad, dour nostalgia for quaint countryside, and wow rural communities are people too, dude -type voyeuristic stuff, punctuated by deep & profound (sarcasm) quotes from ancient chinese philosophers etc oh god. it's many hours of my life i'm not getting back for zilch gain. people praise its 'sensitivity'.......so maybe one for the sadbois?

if you like that kind of thing, fine. but there's no way it should be number 1.

[greennui]

900 minutes of Heimat was enough heimat for a lifetime.

[RogerTheMovieManiac]

Yay, I finally managed to log in (after forgetting my password for yonks)! Hello again, SCFZ!

Thank you to flip for that teasing out and examination of the Bayesian formula principle. Maths is not really my strong-point but I think DtC might be losing part of its very unique charm by discarding the Adjustment for Popularity (which I had to ask for clarification around, when I started taking part three or four years ago...).