Jump to content

Gender differences in anime ratings


BunnyAdvocate

Recommended Posts

Ffx7CJo.png

I was curious how much the anime fandom split down gender lines. What do men and women prefer? Is there a difference in how they rate things? What anime most divides them? Using the profile and rating data on MyAnimeList, I tried to find out.

I'm not too sure how to list tables in this type of forum, so I hope you'll forgive me if I link to my analysis elsewhere, here on tumblr, and here on reddit. I hope you enjoy the results (:

Link to comment
Share on other sites

No surprises on yuri... the same for yaoi.  Vampire rating surprised me though.  In the US, women are more likely to go for vampire literature than men... and there are a ridiculous number of vampire otomege out there.

Link to comment
Share on other sites

Interesting read, Bunny!

 

11 hours ago, Templarseeker said:

Whoa never thought most female viewers would prefer to watch animes relating to vampires and demons?! I kinda assumed they much prefer romance or slice of life or high school kind of animes. Surprising... In a way...

You would be surprised by the amount of fucked up stories about demons or vampires I found in Josei/Yaoi magazines... Girls must really like them (well, I'm a girl and I like).

Link to comment
Share on other sites

What I find interesting is the distribution of scores and votes are between the genres. If we compare cars with yuri, it's a no brainer that yuri is really popular. Just look at the success of Sono Hanabira ni Kuchizuke. As it turns out, that's not the case. Male gives a score of 6.65, which is 1.49 below cars and cars have more than 4 times as many votes. For females we see the same pattern with cars getting 1.65 more and cars have 10% more votes. Keep in mind that cars happens to be the genre with the highest percentage of males. In other words yuri in anime seems to lack votes/viewers and those who watch it gives it quite low scores compared to the other genres.

 

Out of the 19 genres with 65%+ male viewers, 15 of them have higher votes from the females. This means the females who watch do enjoy it a lot. However in the other end with a majority of female viewers, the same pattern isn't visible with male votes. In fact the absolutely lowest score is from males to a primarily female genre. Looking at the graphs, male have more votes in 7 and 8 than females while females have more 10 votes. Does this mean that there are more female than males, who doesn't scale their score and just gives 10 if they like it? Is that enough to give the offset where females vote a bit higher in most categories?

 

It states that many genres have more male fans because there are more male accounts. It's hardly surprisingly, but it does bring up a problem: how would it have looked if we had precisely the same number of people of each gender? One way to get it this way would be to multiply female votes by (number male accounts/number of female accounts) when calculating the "male fanbase" column. This way if 10% of male and 10% of females vote for a specific genre, then the ratio would be 50%. This number actually tells more about preferences of each gender than the absolute number we are given.

 

1 hour ago, Fred the Barber said:
  • Disappointed girls don't like Spice and Wolf more.

Me too at first, but if you look at the numbers, then you will see a different pattern. Females actually voted 7.97(I) and 8.09(II), which is actually pretty good in itself. It's not as high as the male votes, but it's still far better than "don't like". In fact it's higher than 16 of the male votes on the top 25 table for highest male vs female votes.

Link to comment
Share on other sites

2 hours ago, tymmur said:

It states that many genres have more male fans because there are more male accounts. It's hardly surprisingly, but it does bring up a problem: how would it have looked if we had precisely the same number of people of each gender? One way to get it this way would be to multiply female votes by (number male accounts/number of female accounts) when calculating the "male fanbase" column. This way if 10% of male and 10% of females vote for a specific genre, then the ratio would be 50%. This number actually tells more about preferences of each gender than the absolute number we are given.

Oh we limited the dataset so there were the same number of fans of each gender. However on average, males watched twice as many shows as females did, so they tend to be the majority of most genre's as multiple shows watched in the same genre counts multiple times. I did a breakdown here listing what ratio of each gender had ever tried an anime of that genre, which I think better captures the differences.

Thank you for your feedback though. You raise some good points.

Link to comment
Share on other sites

42 minutes ago, BunnyAdvocate said:

Oh we limited the dataset so there were the same number of fans of each gender.

I'm not sure limiting the dataset is the way to go. In fact I wondered about writing something about trustworthiness related to using a relatively small dataset. The problem is that with a reduced set, you basically made an opinion poll of the votes rather than the votes themselves. Polls usually needs at least 1000 to be trustworthy, yet many specific titles seems to have less than 100. In fact some of them are so small that adding just one 10 or 1 vote would alter the score. The only thing to do about this issue is to increase the dataset.

 

In order to get the most trustworthy scores, use all the votes for the score average. It could also be interesting to see the median score. Another option would be to get the average when the top and bottom 5% have been removed, which would remove some of the offset from people just voting 10. Remember this is useful for calculating specific scores (say female comedy), which can then be used to calculate score difference between genders, while it's horrible regarding finding male/female radio for a specific category.

 

One reason for using the full dataset would be to avoid all the issues around using a sample of a set. For instance if you only use 10% and only 5% voted for a certain title, you can by chance get only 10% of the votes for that title and they could all be from the fanclub of said title, meaning they would be the top 10% and the average score could be say 1.5 lower. It's not likely, but at the same time not impossible. You can avoid this issue by using the full dataset.

 

55 minutes ago, BunnyAdvocate said:

Oh we limited the dataset so there were the same number of fans of each gender. However on average, males watched twice as many shows as females did, so they tend to be the majority of most genre's as multiple shows watched in the same genre counts multiple times.

This in itself is interesting as it gives insights to something other than just what people watch. It does screw the results a bit. Maybe the best approach here would be to not balance for accounts, but balance for votes. Take the full set, count male and female votes in each category and then multiply female votes with (total male votes/total female votes). I think that would give the most accurate picture of what each gender watches when they watch something.

 

I know it might not be the most fun reply you can get right after finishing what seems like a huge task. I intend this to be constructive feedback and I do realize that statistical analysis is not the easiest thing in the world and most people underestimate the impact of statistical uncertainty. I do not claim to be an expert in the field, but I have passed statistical analysis at the university. What I have written here is my educated guess on how to maximize the quality of the result while minimizing the usage of complex math.

Link to comment
Share on other sites

1 hour ago, tymmur said:

I'm not sure limiting the dataset is the way to go. In fact I wondered about writing something about trustworthiness related to using a relatively small dataset. The problem is that with a reduced set, you basically made an opinion poll of the votes rather than the votes themselves. Polls usually needs at least 1000 to be trustworthy, yet many specific titles seems to have less than 100. In fact some of them are so small that adding just one 10 or 1 vote would alter the score. The only thing to do about this issue is to increase the dataset.

Yep I'd certainly like to use more data, I'm hoping for more on any follow up post I do. However we found that the results for 4k users and 6k users were pretty much identical. A few genres moves up or down a position or two, but overall the results seemed steady. So I'm fairly confident that our results provide a decent picture to the state of things, although of course they aren't definitive and could be more reliable.

1 hour ago, tymmur said:

It could also be interesting to see the median score.

Unfortunately because MAL only allows whole number ratings, the median wouldn't be very useful. You'd have all the anime clustering on 7, 8, 9 and 10.

1 hour ago, tymmur said:

This in itself is interesting as it gives insights to something other than just what people watch. It does screw the results a bit. Maybe the best approach here would be to not balance for accounts, but balance for votes. Take the full set, count male and female votes in each category and then multiply female votes with (total male votes/total female votes). I think that would give the most accurate picture of what each gender watches when they watch something.

This is what I'd have preferred to do, I'd have just weighted the male ratings a little lower to balance out how many more of them there were. However I wasn't responsible for the data collection aspect of this, my partner did that, and they cut the data down so I used what I had. I'm hoping to do a follow up post in a few weeks and I'll be doing the data collection myself, as I'll also be collecting far more data on such things as stalled and dropped anime.

1 hour ago, tymmur said:

I know it might not be the most fun reply you can get right after finishing what seems like a huge task. I intend this to be constructive feedback and I do realize that statistical analysis is not the easiest thing in the world and most people underestimate the impact of statistical uncertainty

No worries, I appreciate the feedback. It's the only way I can make sure the next post is better (:

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...