VNStats, a cool way to analyze your reading habits

[Decay]

Noslyl from Reddit made VNStats, a cool site that hooks into VNDB's api to analyze its data. It gives some broad data on VNDB's overall voting habits and then it allows you to dig deeper and see data on individual VNs, tags, character traits, and users of VNDB. My favorite feature of the site is the way it gives you an analysis on the VNs you've voted on and your own reading habits.

Here's my profile. I have Homicide, Murder, Accident, Death, Cruel, Shotgun, and Blade all in the top ten traits . Meaning I rate games containing characters with those traits highly. How about that? It also delivers some recommendations based on this data. I suppose I should finish Muv-Luv Alternative one of these days, and get into the Infinity series. Also, supposedly @douggle is the third most similar user to me on all of VNDB, hey there fellow murder-loving buddy! Actually, it seems that Douggle has a military uniform fetish. I can empathize.

In order to get someone's profile, just stick their VNDB user id number into the end of this url: https://vnstat.net/user/

What does the rest of your profiles look like, everyone?

[Dergonu]

3 minutes ago, Decay said:

hey there fellow murder-loving buddy!

K-Kowai

 

Well, my recommendations seem to be pretty accurate based off my ratings and tags. Fate/Stay night, muv-luv, g-senjou no maou, rewrite, little busters.

 

(Guh, my ratings on yuri titles are too low!

I .. MUST ... RATE THEM ALL... PERFECT 10´s! )

[Yuuko]

Have my profile set to private and seems like the database is updated only once in a day  

[Deep Blue]

it's a shame that it only counts the vns that you rated and not finished too  

also davidlu9837 .... that dude O.o no wonder I've seen his vote all the time he read more than 3191 novels   

[Arcadeotic]

I saw this today and it seems really intruiging

The Novels it records is a great feature, and it let's me see how objectively bad taste I have to the norm. It shows how much my rating is higher or less from the average, which is interesting to say the least.

The Traits and Tags gives an accurate representation on what might be enjoyable to you in the future.

Developers is just a nice add-on, which gives insight on what developers you should keep an eye on (but of course most of us already know our favorite companies).

Then the Recommendations and Similar Users. These ones are great. They give you great new VNs to try out, which gives even more VNs, and the cycle repeats itself for infinite backlog. Then they can give you  chance to meet people with similar tastes to yours.

All in all, a great new website to keep in your bookmarks.

 

And apparently Segai is the only one here with similar tastes to mine

The more you know

[Ike]

Very interesting data, for sure.

I definitely tend to go for the longest and highest rated VNs out there. My average VN takes a full 29 hours, and I read slower and less often than other people. No wonder my list has only 19 entrys since I started reading VNs.

[Narcosis]

>data based off VNDB's ratings

Sounds legit.

[Guest]

This is awesome. In didn't know someone had already been doing that !

I'm not being sarcastic this time. This is an incredible tool. It even gives you recommendations based on a weighing system. I've been planning on doing one of my own. I guess it's not needed anymore.

Mine not updated

Get a like my friend.

This should be stickied or something.

[Vokoca]

Witch, Demon, Blade, Servant, Shotgun... Evening Gloves.... uh... I don't know what to make out of those traits.

That said, the first three tags seem to be Violence, Metafiction and Villainous Heroine, so it seems they're getting at least something right.

[NickFreja]

Wow cool. I got homicide, murder, and mecha pilot in my top 3 traits
I love how it compare my rate to average rate

[Kawasumi]

Drug Use, Bad Endings with Story, Suicide, Bad Ending(s)

Seems legit  

[Satsuki]

I have "Death" sitting on top of my trait list, and I call bs on that. And 9/10 of the recommended vns are in the list of "VNs that I will never touch". This site is trash  

[Narcosis]

You know how distorted that data is, all due to how VNDB works? Everything's based off ratings and there's nothing more misleading than ratings from VNDB nowadays. Majority of titles, especially the popular ones remain greatly overrated, while rare, decent titles struggle hard to get any sort of exposition.

There is nothing inherently wrong with the voting system itself, the problem lies within users. In the current system, 5 is treated as "average" by default. Please, tell me, how many 5's have you given out lately? Have you given out any 3's or 4's? 2's, perhaps? It's what vast majority of titles deserves at most. All I ever saw were ratings starting from 7 and up above, cause god forbid voting on my favourites any lower! Tenths of 10's given out by single users, where the original idea was to give out only a single one at most. Of course, we don't want our favourite vn's to have low ratings, at least not any lower than LBEX, therefore voting lower than 8 is out of the question! That's the problem. VNDB gives way too much freedom to users and doesn't self-regulate itself; such data will always be unfit for any sort of analysis.

Meanwhile, VNStats collects data only based on users, who voted. Those that haven't, aren't even analyzed. A lot of users don't vote; some users removed their votes. That website simply continues the trend to expose the least useful and most misleading data out of VNDB.

[Decay]

As much as it would help, you can't use data that doesn't exist. I don't know what you want out of this site, but whatever it is I'm pretty sure it's not possible.

[Toranth]

52 minutes ago, Narcosis said:

There is nothing inherently wrong with the voting system itself, the problem lies within users. In the current system, 5 is treated as "average" by default. Please, tell me, how many 5's have you given out lately? Have you given out any 3's or 4's? 2's, perhaps? It's what vast majority of titles deserves at most.

To comment on just this point, it isn't exactly accurate.  If the first few users to play a certain game all give it 1s and tell everyone it was complete crap, it won't get many votes at all.  A few that hate it, a few that love it, but most people will never even give it a chance.  On the other hand, if the first few users to play that game give it 10s and tell everyone it is totally awesome, many people will play it. 

So it is likely that most of the votes in any given system with voluntary participation will be above "average", if just because people actively avoid the 'bad' stuff to play the 'good' stuff.

[Nosebleed]

Top trait:
>Baseball

Top tag:
>Onee-san heroine

First recommended VN:
>Chaos;Head

10/10 would use again  

The graphs are cool though.

[Arcadeotic]

Top 3 traits:

- Ribbon Tie

- Hosome

- Senpai

Top 3 tags:

- Teasing

- Amnesia

- Class President Heroine

 

I... don't even know what to say... 

[Tyrael]

Top trait: Death

 

Oh.....

[KuroganeHomura]

I'd put off voting on most of the VNs on my list, so I'll have to wait a day to get the updated list. So far the top 3 tags are

- Group Sex of one Female and Several Males

- Philosophy

- Doggy

and the top 3 traits

- Student Club Member

- Popular

- Stalking

Bit of a curious result considering the VNs I had already voted on haha. The recommendations though are mostly things I've already been planning to read, so that part seems to be working well.

 

Narcosis: For the average of votes to be 5, users would have to either choose the VNs they read completely randomly without taking their own tastes and recommendations by others into consideration, or adjust their votes so their average becomes 5. The former is obviously not happening, the latter is actively discouraged by the labeling next to the voting options. If you put some time into picking the VNs you read and unless you spend a real lot of time reading, the average VN you read won't be "so-so", certainly not "weak" and definitely not "bad" or "awful". It's gonna be at least "decent" or "good" as far as your subjective enjoyment is concerned. For that matter, how likely is someone to finish a longer novel that they'd rate "bad" or "awful"? I'd say there's a decent chance they'd simply drop it. Which means that unless they also vote on VNs they never actually finished, their opinion won't even be reflected in the overall rating.

Thing is, all of that doesn't matter for VNStat's similar users function and, depending on how it's implemented, not much for the recommendations, either.

[VirginSmasher]

3 hours ago, Satsuki said:

I have "Death" sitting on top of my trait list, and I call bs on that. And 9/10 of the recommended vns are in the list of "VNs that I will never touch". This site is trash  

It is VNDB after all 

[Kaguya]

Average rating- 6.57. I really should go back and re-rate stuff, since I rated everything too highly, given how most VNs are trash.

Apparently I spent  ~7,791 hours  with VNs. Huh.

 

Top 10 tags, removing route stuff, which is pointless-

1- Duel

2- Desperation

3- Genetic Research (completely unexpected, tbh)

4- Villainous Heroine (Well, of course)

5- Revenge 

6- Politics

7- Pistol Combat 

8- Intrigue

9- Homicide

10- Utsuge

 

Nothing too surprising up there, aside from Duel being as high as it is and Genetic Research.

 

Top 10 developers

1- Light (all heil Masada)

2- Nitroplus (I guess I do love all their older works)

3- Propeller (Chuuni strong in the blood)

4- 07th Expansion (As much as I shittalk Umineko and Higurashi, I did enjoy both... Higanbana is nice and Rose Guns Days is one of my favorites, so it's fitting, I guess)

5- Xuse (Ayyy, lmao. I only remember playing one Xuse game- Aselia- and gave it a 9.5. Probably because of that )

6- Front Wing (Their games are nice, though their score is only as high as it is because I'm a fan of the Grisaia series)

7- Eushully (They should be higher, since I played some of their games that I haven't added to my VNDB yet)

8- Akatsuki Works (Should be higher, but I haven't played a lot from them yet  )

9- Type-Moon (Sheesh, time to adjust my ratings. TM stuff is higher than it should be, for the most part.)

10- Liarsoft (I am a fan of their games in general, after all. And not only the steampunk series.)

 

Recommendations- Senshikan, Ruitomo, Baldr, Vermillion, Sona-Nyl, Muramasa... Pretty legit, for the most part. 

[sanahtlig]

2 hours ago, Toranth said:

To comment on just this point, it isn't exactly accurate.  If the first few users to play a certain game all give it 1s and tell everyone it was complete crap, it won't get many votes at all.  A few that hate it, a few that love it, but most people will never even give it a chance.  On the other hand, if the first few users to play that game give it 10s and tell everyone it is totally awesome, many people will play it. 

So it is likely that most of the votes in any given system with voluntary participation will be above "average", if just because people actively avoid the 'bad' stuff to play the 'good' stuff.

That's not as big of a problem as you make it out to be.  What you speak of is an exposure bias.  Different levels of marketing leads to exposure bias too.  What exposure bias does not necessarily impact is the average rating, which is what most people look at.  It does impact the Bayesian rating however, and VNStats uses the Bayesian rating (and it should).

But ironically, the most extreme exposure bias is with untranslated titles, which is why VNDB ratings are not a good tool for judging Japanese titles.  At extreme levels, exposure bias leads to selective voting--only hardcore fans of that genre or developer will play lesser known games and vote, leading to a positive bias on ratings.  Add to that the language barrier, where players might vote on a gameplay title and rate only its gameplay since they don't understand the story.  This bias can be observed in action with Raidy 3, where the VNDB rating was highly inflated before the English release, after which it plummeted to the series average.

I don't see any major issues with using VNDB ratings to find English titles you might enjoy, especially within a given genre.  The key is to compare ratings between similar types of games

[Fred the Barber]

Pretty fascinating, and I'm coming around on thinking the recommendations are useful. At first I thought it was basically just recommending me highly-rated things (which I think it clearly is, to some extent; MLA and Steins;Gate are probably recommended to everyone who hasn't read them...), but Sharin no Kuni and Shikoku no Sharnoth are both quite a bit below 8 on average rating (while there are many, many 8+ rated VNs on vndb that I haven't read yet), yet are recommended to me. Sharnoth is an easy one given my substantially-higher-than-average rating of both Inganock and Gahkthun, but I'm honestly not sure why Sharin is there... but I'm pretty sure it's rightfully there, and I'm very much looking forward to reading it eventually.

Tags are astonishingly spot-on, at least for the top three: Enforced Playing Order, Mentor Supporting Character, and Protagonist with Partial Voice Acting. Meanwhile, top 3 traits are just weird: Avoidable Death, Baseball, Pleated Skirt. Preferably all three at the same time, please

In fairness, I'm pretty sure my sample size of VNs just isn't large enough to work with to provide useful analysis, for the most part. Time to buckle down and get back to reading G-Senjou, I think...

[Eclipsed]

Yep, I totally went on a death spree late last year  get a bunch of high skoo moege stats then BAM, Death!@#



~~
Obviously 'cos of my DR & Zero Escape spamming lately I get recommended shit like Ace Attorney and Ever 17

[Dergonu]

35 minutes ago, Eclipsed said:

Yep, I totally went on a death spree late last year  get a bunch of high skoo moege stats then BAM, Death!@#



~~
Obviously 'cos of my DR & Zero Escape spamming lately I get recommended shit like Ace Attorney and Ever 17

"Aww, look at those cute tags! Yeah, Eclipsed sure does read a lot of moege and..."

DEATH

...