VNDB Length correction?

[Okami]

Well I noticed that I always need more time to finish all routes in VN then it is suggested on VNDB and I thought that VNDB length is based on how much time you need to finish only one route in witch case it would be mostly correct. But then I noticed that some people around here are also saying that lenth suggested on VNDB isn't correct but in the oposite way, that is they are suggesting that VNDB lenth is  exaggerated and that you don't need that many hours to finish VN as is suggested.

 

 

So what from your point of viev, is lenth of VN suggested on VNDB correct, exaggerated or decreased?

 

 

Lets discover mystery behind VNDB suggested lenght and on what it is based on.

[Nosebleed]

Length in vndb is calculated by how much time the vn in its entirety takes to complete in the default auto mode.

Everyone has different reading speeds though so it's a VERY relative number.

At least that's what i've been told.

[Chocolat]

I don't really look at the length of the visual novels in vndb, but it takes me approximately the time suggested to finish some games, or sometimes I finish it with more time. All depends of the visual novel I've choose.

[Tenkuru]

Well it really depends on the person who's reading. Like how fast a person reads and whether the person waits for the characters to finish speaking before clicking. Personally, I'm a fast reader and I don't even wait for the text to show up, let alone the voices. 

[Okami]

Length in vndb is calculated by how much time the vn in its entirety takes to complete in the default auto mode.

Everyone has different reading speeds though so it's a VERY relative number.

At least that's what i've been told.

 

I doubt that, as auto mode is pretty slow nobady would then need more time then suggested.

[DoodlyNyan]

I think they're based on the auto-mode at the VNs, so they aren't really accurate.

[Funnerific]

I doubt that, as auto mode is pretty slow nobady would then need more time then suggested.

Some VNs have pretty fast auto-modes by default and you won't be able to keep up without paying attention. Also, count in the time you spend taking pauses and breaks.

[Nosebleed]

I doubt that, as auto mode is pretty slow nobady would then need more time then suggested.

I read pretty slowly and i've taken more time than suggested. Auto mode is meant for an average reading speed so it makes sense that's how they calculate it. Number of words and then how many words per minute.

[monkeysrumble]

This pretty much all depends on reading speed. Personally I usually take shorter or the correct amount of time that it says it takes on VNDB. They seem to be somewhat accurate and they don't give exact numbers but they give numbers in a certain range.

[Tenkuru]

Some VNs have pretty fast auto-modes by default and you won't be able to keep up without paying attention. Also, count in the time you spend taking pauses and breaks.

So far, all the VNs that I played have a reaaaallly slow auto-mode (at least to me.) And it gets very annoying when the option for changing the speed is bugged, like in Grisaia. Also, I doubt people take into consideration the time spent on taking breaks when calculating their play-time. 

[Nosebleed]

My best guess personally is they use an average reading speed, a constant, let's say 200 words per minute (below average adult, middle/high school level) and then by knowing how many words are in the vn they calculate the length it'd take to read which would be equal to: number of words divided by wpm.

The reason why you take less time often is because the average reading speed is lower than the average adult's reading speed.

That's my personal theory.

Auto mode uses a set wpm speed, hence why it's not wrong to say they use it as well.

[Darklord Rooke]

Length in vndb is calculated by how much time the vn in its entirety takes to complete in the default auto mode.

 

That would be the only reasonable way to calculate it, so sounds about right.

[Narcosis]

People pay too much attention to trivial things. Read in your own pace and stop pondering how long will it take you to finish; reading should be a joy, not a punishment.

[cryofrzd]

I take the length thing on VNDB with a pinch of salt.

I can't find how for example Clover Day's is regarded as Very Long and then Daitoshokan no Hitsujikai is only Long.

 

Wouldn't it be faster and more accurate to measure the length of the text scripts than using the auto-function, or is it a problem with separating the text from the game? The file size should be a good approximation if same encoding is used, if the game doesn't use stupid character-caching, I've seriously only seen one VN doing that shit.

[Pabloc]

Sigh...

Ok, I wouldn't expect this from the OP. But seriously, 13 replies, and still not a single person got the idea to actually check the VNDB guidelines, where length and all other """""mysteries""""" are explained? -.-'

Length

Very rough estimate of the time required to finish all endings of the visual novel. To determine the length of a game, it's often better to ignore this time indication and instead compare it with other games you've played. It's all relative, after all.

In other words, it works just like that:

"Let's see, "VN A" took me roughly the same amount of time to read as Planetarian or Saya no Uta, so I'll mark it as short." That's it. No deep philosophy, advanced calculations or quantum physics. A slowpoke will need 10x more time than a fast reader to finish the same VN. I need 2-3x more time to read the same VN in Japanese vs English. Does anyone seriously expect those roughly estimated times to actually match everyone's personal reading speed? -.-'

 

And where did that auto-mode idea come from? Some VNs don't have auto-mode and the majority of others let you freely adjust the text speed, how would that work? -.-'

[Nosebleed]

I take the length thing on VNDB with a pinch of salt.

I can't find how for example Clover Day's is regarded as Very Long and then Daitoshokan no Hitsujikai is only Long.

 

Wouldn't it be faster and more accurate to measure the length of the text scripts than using the auto-function, or is it a problem with separating the text from the game? The file size should be a good approximation if same encoding is used, if the game doesn't use stupid character-caching, I've seriously only seen one VN doing that shit.

Using the auto-mode doesn't exactly mean they actually sit through the whole thing on auto and see how long it takes. That'd jusst be really stupid and non efficient.

 

Using auto-mode means checking what's the speed at which the words show up and disappear. Auto-mode gives you the words per minute which allows you to calculate how long the vn would be at that reading speed by diving the total number of words in the vn by the words per minute. 

And by auto-mode I was referring to the default one where you don't change its settings.

 

But that won't apply to everything as Pabloc mentioned there's vns without auto-mode

 

That's why my best bet is they simply use a constant speed, a reasonable one would be 200 wpm. And then just apply the same formula to every single vn. It's quick and accurate for the most part (not as accurate if the vn has gameplay).

 

And in the end this is just a generic formula. Everyone has different reading speeds. I rarely pay attention to the length of a vn because it's irrelevant to my choice on wether or not I want to read it.

[cryofrzd]

Well it's good to have some reference about the length, I prefer medium and longer because I honestly feel many shorter ones are either rushed or contain too little story and development. Of course there are exceptions to this.

 

I don't use vndb that much for choosing VNs, but it's good as reference once you find something interesting, primarily the tags.

 

When I see Very Long I used to expect a length like a VN with long common-route and more than five long character-routes.

With medium I used to expect like an average linear-plot one with around five side routes. Or a long common route with shorter character routes.

 

Today I don't expect anything from it.

[Okami]

Using the auto-mode doesn't exactly mean they actually sit through the whole thing on auto and see how long it takes. That'd jusst be really stupid and non efficient.

 

Using auto-mode means checking what's the speed at which the words show up and disappear. Auto-mode gives you the words per minute which allows you to calculate how long the vn would be at that reading speed by diving the total number of words in the vn by the words per minute. 

And by auto-mode I was referring to the default one where you don't change its settings.

 

But that won't apply to everything as Pabloc mentioned there's vns without auto-mode

 

That's why my best bet is they simply use a constant speed, a reasonable one would be 200 wpm. And then just apply the same formula to every single vn. It's quick and accurate for the most part (not as accurate if the vn has gameplay).

 

And in the end this is just a generic formula. Everyone has different reading speeds. I rarely pay attention to the length of a vn because it's irrelevant to my choice on wether or not I want to read it.

 

Yea I think that way would work the best, take most average reading speed/number of worlds and get the time. The only problem would be if you actually can get info about number of worlds in a VN.

 

And yea I also agree that lenght of VN can influence your choice about if to read it or not.

[Nosebleed]

Hm so basically it's all relative according to vndb.

Well not that shocking if i dare say so myself though I think they could use a more accurate measuring method. I like accuracy~

 

This is why I say the length is irrelevant since everyone has different reading speeds. It's a very rough estimate, that's why the intervals are in the 10/20 hour range because you'd never be able to get a more accurate time without using math.

 

Oh well.

[cryofrzd]

Why use number of words? number of characters and weighted with some language factor seems easier to calculate and more applicable to Japanese. If they now want to make some reliable length estimate.

[cryofrzd]

Perhaps they should do like they do with the Bayesian rating. Let people write their length-estimations and those that on average rates the same VNs as longer than others will have their length-estimates weighted downwards.

 

This is an well-behaving statistical estimate that is widely used, as long as enough users write length-estimates it will work without using even more statistical "magic".

[Okami]

Perhaps they should do like they do with the Bayesian rating. Let people write their length-estimations and those that on average rates the same VNs as longer than others will have their length-estimates weighted downwards.

 

This is an well-behaving statistical estimate that is widely used, as long as enough users write length-estimates it will work without using even more statistical "magic".

 

That would work too, Tags are decided that way allredy so it might not be a bad idea at all to decide lenght this way too, and it would be more acurete than the way it is now.

[Zakamutt]

Edit: rebuttal not coming, nothing much comes up, cry wins. I wonder if that many people would be arsed to fill in lengths, but maybe...

[Pabloc]

Why are you trying to fix something that isn't broken? -.-'

This is just a rough estimate, not some magically objective length, and it will never work like that no matter what you do.

 

The so called "more accurate" method would require a crapton of time to implement (almost 15k titles in the database, have fun with word/character-counts), and it wouldn't really improve anything. Let's say that you calculate a VN's length as 32h. So what? Fast reader will finish it in 15h, slow will need 50h. All they would get is the general idea how long the VN is. That is... pretty much what they get with the current, rather effortless system.

Another issue is that calculated time estimates will never be accurate. Muramasa will take much more time to read than a simple moege even if both have similar amount of words/characters/whatever. That's why the relative method is superior here - it shows how much time someone spent on the title in comparison with other titles he/she finished with his own pace.

 

And weighted estimates wouldn't improve anything either - again, after "weighting", the VN gets 32h length. And again, fast readers need less, slowpokes need more. Plus, what Zakamutt said about small sample size - that would screw up the majority of untranslated titles (with 1-3 votes) big time.

Besides, the relative system is already "weighted" in a sense - as I mentioned, people editing entries aren't supposed to stick to suggested times, but compare the time they needed for one VN with others that they have finished and classify it accordingly.

Now that I think about it, length shoud be only described as very short/short/medium/long/very long descriptions, without any suggested times - that would be sufficient and less confusing.

 

And finally - you think something has wrong length? Then just fix it. VNDB is edited by users after all. I have changed the lengths of a few VNs that I have read myself. That's especially the case with untranslated titles, those available in English have pretty accurate lengths (I'm not talking about the time, but the short/medium/long classification).

[cryofrzd]

Perhaps I misused the bayesian terminology, sorry.

 

I meant that every user has a bias weight calculated from his average deviation on the length estimates.

And then the resulting estimate is the average of vote*weight.

 

Of course it won't correct bad estimates made on just one user making that estimate (no other estimates by that user and no one else estimating that VN).

But this problem exists already today, since the length estimate currently is the estimate one person makes.

 

Zakamutt has a good point that we'll get more unexperienced estimators if it becomes easier to estimate. Perhaps the weighted length estimate can live parallel with the old estimate, like the ratings.