Hi, this is fantastic news!
(Though now I will need to go through my history and figure out which ones should really be marked as unknown.)
Top tip: It seems that things you’ve marked as watched on an unknown date will still appear on your History page, but in the position where the Unix epoch i.e. 1st January 1970 would be.
Initially, I expected films marked with an unknown date to appear at the end of the list, but because I used to mark films as watched on their release date if I didn’t know when I had watched them, I had some films marked as watched before 1970, which meant films marked as watched on an Unknown Date would appear above them:
This is truly a great feature, but the episodes I marked as “unknown” are definitely messing up my all-time stats. It seems they all get marked with a date in January between 1:00 and 1:59 on a Thursday. At least I’m sure that I didn’t watch that many episodes in Juanary or between 1:00 and 1:59.
This is great! Thank you so much for implementing this. In fact, this feature being available now made me a paid subscriber.
I have one question though: When I look at my watch history CSV export, I can see that “unknown date” gets exported as 1970-01-01T00:00:00Z, i.e. a Unix timestamp of 0 seconds. I don’t mind that, but I’m curious whether internally you’re indeed simply storing a zero timestamp, or null (which then becomes a zero timestamp on export and display and stuff). In other words, do you have any way to differentiate between an “unknown date” watch and a 1970-01-01T00:00:00Z watch?
Because it doesn’t seem like you have. If I check in at “another date” and enter 1970-01-01T00:00:00Z, after saving, it says “unknown date”. I don’t see a problem if that’s how you’ve implemented it, I’m just curious. Are unknown date watches stored as null or as 1970-01-01T00:00:00Z?
To me the actual date I watched something isn’t that important if I can’t be 100% sure when I did, But I’ll be honest here, it makes absolutely zero sense to me to have it all default to 1970-01-01. In numerous cases that would precede the date the show / movie became available and thus really skew statistics for those who want that.
I think it’s fine if you do what the Trakt website seems to do and interpret 1970-01-01T00:00:00Z as a special value, and exclude it from certain statistics. It will take some time for the Trakt website, app, etc. to be updated with the new logic, but I think the stats should look fine when that’s done.
Arguably, the statistics were already skewed by people recording unknown plays in all manner of ways (including myself who would use the release date). So you’d probably find a surprisingly high number of plays on the release date.
Whereas now, at least there’s a consistent way to record this, so unknown plays can be excluded from statistics calculations, which should improve their accuracy.
It depends what stats these unknown dates are being included and excluded from.
I think a better explanation of how this is going to effect stats and everything else.
Up to now I have all my “unknown” watches marked as 1 January 1950 00:00, at least then I know anything on that date is an item I know I’ve watched previously but don’t know when.
Like most features, everyone uses the system differently. The logical solution is to have an option to exclude unknown from the stats and let the user decide which is more useful to them.
Trakt has various charts showing your plays broken down in different ways, such as by month, day of week or hour:
Since you have recorded all your unknown watches as 00:00 on Sunday 1st January 1950, that means that these charts will show a disproportionate amount of plays for Januarys, Sundays and 00:00.
That’s not to say you’ve done anything wrong, but now that Trakt has the “unknown date” feature, this opens up the opportunity to exclude “unknown date” plays from these specific charts, which will make them more accurate. Of course, your total number of plays should continue to include those plays.
As I know I have added “unknown" at that date I’m aware of the “odd" results for those times.
The thing is how does the “unknown" date affect those stats or is it that all unknown will not be included in any stats.
I wonder because my watched stats went haywire the other day and have gone from Items and Plays being the same to me having lost 25,000 odd items (tagged Justin on the thread to ask what has happened, I’m not the only one).
So it makes me wonder what this “unknown" date affects.
If my stats return to normal and we are clear of the “unknown” consequences I can/will start a change of my 1950 date to unknown.
Let’s see if Jason can throw any light on my problem.
Personally I have just always either guessed roughly when I watched the movie or series. Marking something as watched in 1970 when it released in 2010 for example would seem utterly silly to me, but to each their own.
I realize this was something so many asked for but I truly feel like there will be no solution to how the “unknown” factor gets handled as everybody has their own method of handling it.
Unless of course as surmised by @scy above they are actually marked as Null and not the time / date we think they get marked as.
The troubling thing to me is that this seems to have caused “plays” to be broken.
I really would wish that when changes like this are forthcoming that there would be at least some kind of communication in the future with a poll or similar, asking users how they would like it to be handled. Give us all options rather than make the changes and then have users discover the change and feature isn’t exactly as they envisioned.
It seems there are some bugs at present, so I wouldn’t take the current functionality to be the intended behaviour.
If 1970-01-01T00:00:00Z is always treated as a special case, then it effectively doesn’t really matter how it is actually recorded in the database. It’s just an implementation detail.
Justin is aware of the bugs, and says he’ll fix them. Once they’re fixed, I really don’t see there being an issue. The total number of plays should include all your plays (including those on “unknown date”), and the chart for the number of plays on each day of the week should correctly exclude those with an unknown date (because if you don’t know what date you watched something, it makes no sense to include them on such a chart).
Hopefully @justin can get this sorted soon and post a report on exactly how it works and its effects.
For me it’s a good idea once it can be sorted out to work correctly as it saves me having to manually input a date and time for old shows/movies I come across I remember watching at some point.
It’s just slightly annoying that a new feature doesn’t work how it should and has the possibility to cock up other things on it’s implementation.
Having said that I have great respect for the guys and gals that can do this sort of thing as it is way beyond my capability.
This is something we were thinking about. Thank you for your feedback!
I think we should group them better than this but also, yeah, a lot of the old “mark at release date” should in fact be moved to “unknown date”. Which is also something we are thinking about.
We are storing the date as 1970-01-01T00:00:00Z, not null. We aimed to make it backward compatible with a lot of other things like export/import, public API,…
This is an implementation detail, the goal is to make it transparent for users. So yeah, as a user who doesn’t care about the watch date, you shouldn’t even see that it’s stored as a specific date in the database (unless you look for it).
The goal is also to not mess with the date-based stats. Justin is still working on polishing this. Your counter-based stats shouldn’t be affected either.
Date-based stats/graphs will not take unknown dates into account to avoid saying things like “you watch more things on January at 00:00”.
Counter-based stats will be no different than before. If you watch a 100 movies at an unknown date, it will show that you watched a 100 movies.
I’ll ask Justin to have a look at this, this seems like a bug that may be unrelated
It is currently not possible but something we are also looking into. Thanks for your feedback!!
Date-based stats/graphs will not take unknown dates into account to avoid saying things like “you watch more things on January at 00:00”.
Counter-based stats will be no different than before. If you watch a 100 movies at an unknown date, it will show that you watched a 100 movies.
I’ve just marked a few movies as unknown date and it’s bumped up the number of plays in 1970
Date-based stats/graphs will not take unknown dates into account to avoid saying things like “you watch more things on January at 00:00”.
Counter-based stats will be no different than before. If you watch a 100 movies at an unknown date, it will show that you watched a 100 movies.
I’ve just marked a few movies as unknown date and it’s bumped up the number of plays in 1970
I’d like to suggest an expansion of this feature, though: allowing multiple unknown-date watches.
Want to +1 this heavily, the only thing that’s keeping this (great) feature from being perfect for me. I have shows I’ve seen multiple times on unknown dates, and it’s crucial to me that I be allowed to track them as such. Hoping this gets added soon!
