Skip to main content
J

3 Messages

 • 

90 Points

Sat, Nov 21, 2020 1:04 PM

Unusual Disparity Weighted Average Vote, Arithmetic Mean and Median

Hello, 

I do understand there is a difference between the Weighted Average Vote, the Arithmetic one, and the Median. 

However, this title had an average varying from 7 to 8 for months.

Link: https://www.imdb.com/title/tt8829176/ratings?ref_=tt_ov_rt

Its weighted average recently dropped from 7.5 to 3.9

Weighted Average: 3,9

Arithmetic: 7,5

Median: 7

Out of 115 votes:

-there are 8 votes at 1/10 and 1 vote at 3/10

-All the other votes are above 7

-There are 108 between 7 and 10

Regardless of the algorithm's easy justification, how can one mathematically/rationally justify that? 

Do the 8 votes at 1/10 weigh 10 times more than all the 7,8 and 9?

Why did it go from 7 to 3 without any vote below 2 added over the past few months? 

No need to send me back to the FAQ, I read it all.

It just doesn't add up for this particular case. 

8 bad votes can't cancel 100 good ones, gradually added over the last two years. 

Thanks! 

N.C.C

Responses

Karen_P

1.4K Messages

 • 

20.1K Points

11 days ago

John.

Hi

The weighted average formula does not actually get used "Until" a title gets to 100 votes.

That is why it had the sharp drop.

So there is no "Unusual Disparity" It would be nice if the formula was used starting with the first few votes. But I've read here that it won't function, or it will give really high or low numbers if used any earlier than 100.

Thanks.

Champion

 • 

3.2K Messages

 • 

99.8K Points

None of the demographic categories have 100 votes yet.

So the 70 non-US users might be a good illustration of what happens when the algorithm is used before 100:

Only 4 of the 70 votes are below 7, and the median/mean is 7/7.1. Yet the weighted average is 1.9.

So something in the algorithm is putting aside votes from a lot of these users.

But among the males 18-29, the weighted average is 8.9, while the median/mean is 7/7.7, because there are no votes under 6. So the algorithm is _also_ setting aside votes in this subcategory, but instead it _increases_ the weighted average.

To IMDb: I wonder if the minimum number of votes should be higher than 100 to avoid these extremes.

3 Messages

 • 

90 Points

Thanks for your reply. 

Still, a movie received 98% of good ratings. 2% of bad ones. 

The displayed rating, affecting the movie, its cast and crew, is incredibly low.

What is the justification, how can it be fixed and on what mathematical ground a few bad grates take on numerous excellent ones?

Don't you agree that this algorithm doesn't reflect the reality, which is:

a movie got excellent reviews and ratings yet displays a terrible grade thus discouraging people from checking it and giving a false representation of its quality and appreciation? 

Would you agree with that? 

Or am I totally lost? 

Champion

 • 

3.2K Messages

 • 

99.8K Points

IMDb will not reveal its algorithm, which is aimed at muting those who try to pull the rating to one extreme or the other. I can only imagine how difficult that must be, and that it will be imperfect for some titles. 

I doubt that just math is used. I would guess that the voter accounts are also analyzed. I just opened each of the reviewer accounts, and almost all of them have only one or two reviews. Some accounts are old, but most are new. If the same pattern holds for the votes, I can imagine those don't count toward the weighted average.

The best we can expect is that IMDb might take this title to their developers and re-examine the algorithm.

3 Messages

 • 

90 Points

Thanks. But how can you ask that?

Champion

 • 

3.2K Messages

 • 

99.8K Points

John, not sure what you're asking.

This forum is monitored by IMDb employees.

I'm hoping one will read this and take the case internally.

If we don't get a reaction from an employee soon-ish, then you could send a message to the Help desk, asking them to read this thread.

https://help.imdb.com/contact