Home - Hal and Dee at the Movies Mail Hal C F Astell - Site Map

Change in the IMDb Top 250

Quick Jump

If you're used to this data and just want to look at the current year's grab, here's 2018.


When I moved to the US in 2004, I wasn't allowed to work until the government gave me permission. That took six months to grant and I took advantage of the free time to delve into classic film courtesy of the newfound wonder (to me) of cable television and Turner Classic Movies. While I watched as much as I could generally, I also tried to have some focus to ensure I was finding an appropriate grounding.

There are many lists of 'the greatest films of all time'. I maintain archived copies of a bunch of Top 100 Lists here at Dawtrina.com, for example, and there are plenty of others out there to play with. These are generally static lists, created by a single person or a focused group of people, and that's fine. However, there's another list that's been around for a long while that is constantly updated and it's voted for by the largest audience of film fans there is: people who frequent the Internet Movie Database.

The IMDb Top 250 is a fascinating, albeit flawed, creature and I grabbed a static copy sometime in mid-2004 to work through. I've kept that up over the years, though I've never managed to watch everything on the list. Over the last few years, I've hovered between the 200 and 210 mark, though I'm a little lower now.

IMDb do attempt to ensure a strong list by applying rules and algorithms to get weighted ratings. They don't disclose all the details of how they do this, but some of them are available here. Put simply, they filter down to theatrical features that have received a certain number of votes (I can't find the current threshold but it used to be 25,000), then reject what appears to be bad data. They do a pretty good job.

Some flaws are still obvious, of course. This is based on popular voting, so it's open to the tyranny of the majority. It's not too surprising to find a strong bias towards recent pictures, especially big Hollywood blockbusters which leap into the list on release and then slowly (or quickly) drop back out again.

What I found over time, though, is that it holds up pretty well. As I write, for instance, I'm keeping track of my ratings in 38 different official lists. My rating is currently highest for the American Film Institute's 100 Years... 100 Movies list, where I've rated 79 of those 100 films for an average, in my system, of 6.67 out of a maximum of 7. However, I've rated 207 of the films in the IMDb Top 250 for an average of 6.68, so, from an entirely personal and an 80%-ish complete standpoint, the IMDb list is 'better' than the AFI's (and, to a varying degree of completion, the other 37 lists I'm following). That still seems odd to me, but data doesn't lie. [Edit: maybe it does, as I'm finding that the AVERAGE function in LibreOffice isn't behaving properly.]

So, in order to keep an eye on this data, I started grabbing a fresh copy of the IMDb Top 250 every New Years Day, starting in 2013. That allows me to see how that data changes annually. I'm sharing that data on pages here for wider reference:

2018, 2017, 2016, 2015, 2014, 2013 and some time in mid-2004.

Here's a summary table:

The Data

Element 2018 2017 2016 2015 2014 2013 2004
Mean 1985 1984 1983 1983 1982 1980 1973
Median 1993 1993 1988 1988 1988 1986 1976
Mode 1995 1995 1995 1995 1995 2003 2003
Hal Rated 176 176 207 188 200 200 210
Hal Average 6.59 6.55 6.57 6.59 6.57 6.58 6.67
Dee Rated 172 171 203 184 197 199 210
Dee Average 6.40 6.37 6.40 6.41 6.44 6.51 6.61
Same 47 30 28 20 17
Up 84 37 82 73 69
Down 105 158 124 140 136
New 14 25 16 17 28
Variation 6.56 10.17 5.95 9.16 10.83


Here's what these data elements mean.


The mean, median and mode are ways of calculating averages.

The mean is what most people would call the average. It's calculated by adding up all 250 values and dividing by 250.

When all 250 values are sorted in order, the median is the value in the middle. In other words, there as many films in the list newer than 1993 as there are older than 1993.

The mode is the most frequently represented value. In other words, according to this list, 1995 is currently cinema's golden year.

Our Ratings

The Hal and Dee numbers represent how many of the 250 films my better half and I have rated (which means we've seen them since 2004) and the mean of our ratings. My rating system ranges from 1 (lowest) to 7 (highest).


The rest of the elements reflect change since the previous year:

Same is the number of films which stayed in the same spot as the previous year. Up is the number that moved up. Down is the number that moved down. New are the number of films in this year's list that weren't in the previous year's. However, some of them may have been in the list prior to that.

Variation marks how much the list has changed overall over the previous year. It counts how many places in the list each film moved and calculates the mean of that.

Basic Analysis

The change in mean tells us is that the films represented in the IMDb Top 250 get newer each year. That's not surprising as new movies are released all the time. The median tells us that after three years of being stuck at 1988, the distribution of them shifted notably newer over the last couple of years.

Our ratings suggest that my wife and I both prefer the oldest list that I grabbed in 2004 and it's become a little less valuable to us since then. However, it's also reflected our views a little more over the last year.

The up and down numbers suggest that a lot more films drop every year than rise. This suggests that, while some films do move up the list, it's much more common for them to be moved down by new entries, which also often move down too.

The variation is the wildest number for me. While IMDb have changed their formula over the years, that's clearly made the list settle somewhat. In 2016 I said that, 'Each of the last three years has seen less change in the list and the amount of change has almost halved in two years.' That was true but it's no longer the case. It's jumped around again since..

Creative Commons License
Last update: 14th January, 2018