Average stars for ALL books on LT?

Is there a way to see this?

Mostly just idle curiosity, but prompted by the following consideration: If people tend to rate books they really like, then the average book rating might be, say 4 stars. (I know it's not 4 stars, I'm saying, hypothetically.) If that were true, then a rating of 4 stars would mean less.
OTOH, if people tend to rate books they really hate (maybe it's fun to bash them), then the average book rating might be, say 1.5 stars. If that were true, then a rating of 1.5 stars would mean less.

Even better, I'd like the zeitgeist top-rated books list to be much expanded (say, to 1000 places), so that we can see what ratings are really exceptional and what books have those ratings.

My experience is that a book rated 3.8 tends to be average, while 4 is usually pretty solid. 4.2 is really noteworthy.

The average of my own ratings seems to hover at 3.28-3.29.

But it was higher in my earlier "trial" months since to make the most of recommandations, similar libraries, etc... I tended to include only my best loved books, no more than a few stinkers or "meh" books got in, just those I read at that time.
Since I went paid I've been nibbling away at adding more or less all I've read and can remember, so the range of ratings has been wider.

Maybe others have been doing the same, even without a real agenda of preferentially rating/bashing books they loved/hated.

The point is really - well, I think - that site owner Tim doesn't really believe in ratings, so he hasn't implemented them to actually do anything. When I realized that I mostly stopped assigning them - why bother if it doesn't actually do anything on this site?

Ratings don't work - so people don't use them - so they don't work. It's a self-fulfilling prophecy.

I keep adding them in the hope that one day they /will/ do something!

Also I like it for my own reference. I've often gone into a tag in my library and sorted by rating to give people a list of the best YA books I've read with awesome heroines or my favorite urban fantasy books or what not.

For me the cutoff is around 4.5 stars. If a book's average rating is that high (and there is a significant number of people who rated it), it usually has something going for it. In that sense, the stars do something; they're somewhat informative.

I'd love to see more stats about LT average ratings, being the voyeuristic reader that I am. I rate and review both books I like and those I don't but my average is fairly high. That's because I don't read books unless I think I'll like, if not love, them and I'm not usually too disappointed. But every once in awhile, there is a book that sounds great by the description but that just doesn't work for me.

I am wondering how people rate an individual book.

If I just finished a book that I really enjoyed and look forward to the next in the series or reading more by the same author is that a 5 even though it certainly was not the greatest book I ever read?

Would the best book(s) I could think of be a 5 and then a book I really enjoy but was not exceptional be at the most a 4?

Maybe one of the problems with rating is a lack of a common scale.

>Maybe one of the problems with rating is a lack of a common scale.

We also have folks who not fully understand the rating system. I see reviews where the reviewer raves about the work only to give it a single star.

Even though I know the site (ie Tim) doesn't value the ratings, I use them all the time. I rate all the books I read and only wishlist books with a rating or 4 or more (with a minimum 100 ratings). I find it very helpful, especailly for literary fiction and non-fiction. For genre books it is less helpful as these are often rated very highly based regardless of literary value - but then, vive la difference!

>9 theapparatus: What's not to understand? The stars mean whatever you want them to mean. Perhaps those people who rave about a book and give it 1 star use the stars to indicate the book's physical location in their library or something. I agree that without a "common scale" as Tom mentioned the ratings are not particularly useful socially (my stars are useful to me), but you can't say they don't understand the rating system. They just don't use it the way you think they should.

Yeah, I just can't help but think the whole "but who knows if people are rating on a common scale???" argument is a red herring. I think enough people rate the same way that they drown out those outliers.

As long as most everyone uses 5 stars as better than 1 star then it doesn't really matter that they don't use the exact same scale since on average 5 stars will be better than 4.

Exactly. If we only had three people rating books, I'd be more concerned with that. And when you DO have very few people rating a given book, I tend to not pay much attention to the rating even if each of them went to a class called "How to Rate Books on LibraryThing."

8> I had so much trouble with my own scale that I ended up putting it on my profile, so I could check what I mean. I know what I mean by 5-star - those are the books that are exceptional. And half-a-star means really really bad. But 2.5-3.5 kept wandering up and down. Now I can check my profile and make sure I agree with myself, if no one else...

I like rating my own books - for one thing, it lets me see at a glance what I've read. Although I check the ratings for books I'm interested in, in most cases they don't affect whether or not I buy the book. I noticed one LT user's rating system was in reverse order to most - with one star being the best rating, down to the "lowest" rating of five stars.

I've also read reviews that mention major problems with a book, but still give it 5 stars.

My personal average is 3.56 with the most books getting 4 stars. I try to give stars to most books as I read them, but not the big fantasy series I keep rereading.

I do certainly agree that a lot more can and should be doen with ratings, and that many of the arguments against them are flawed. I'm not sure that a sitewide average (and deviation) would be that useful although I suppose it might be interesting to know, for calibration as above. Someone then ought to do the stats to find out how many ratings are needed for a given difference in rating value to be significant.

(ie site wide average 3.50 sd0.5) you choose a book with 100 ratings, average 4.00. Is this actually rated higher, or is it just random variation?)

Bump. Is there a way to see the LT-wide avg # of stars?

Okay, I'm still failing to see the point.

You have a Zeitgeist page with the top 25 books by rating, bottom 25 books by rating, and 25 books with the most disparate book ratings.

Each work has a scale of the star rating to the right of the primary data.

Your profile has a graph of your average star ratings.

To me, ratings, like reviews, are rather personal. Just because 10 people chose to rate a book at 5 stars, that doesn't mean I'll like it. (In fact, frequently, I disagree) Sometimes, the average rating may be 3 stars, but it doesn't say anything one way or another to anyone other than the person or people who put in a rating.

So how would a site wide aggregate of every rating mean anything?

I just read a book that had very close to an even distribution of stars over on Amazon ... within a couple of percentage points of being 20% on each rating ... found that fascinating.

I don't typically star-rate books (but I do write 1,000-3,000 word reviews of them all), as something that might be a 4 star for quality of information (oh, and FYI - I pretty much ONLY read non-fiction, so there might be an "apples & oranges" dynamic going in relation to fiction readers), could be a 2 star for writing, a 3 star for layout/design, a 5 star (or 1 star) for companion site, etc. ... making it very hard (for me) to determine an over-all star rating.

Of the 2,328 books I have logged into LT, it appears that I have only rated six, and most (possibly all) of those were 5-star ratings for books that just blew me away, and I was still "in the heat of the moment" when entering them into my collection.

>21 BTRIPP: Actually, that adds an extra dynamic I forgot to consider.
The distinction between Fiction and Non-fiction would also alter the site wide aggregate because different criteria go into each rating.

So how would a site wide aggregate of every rating mean anything?


Normalization.

If the sitewide average is 4.5 stars, it would mean that a book with an average rating of 3.75 stars is less well-liked than average. This doesn't mean YOU won't like it, but I don't think anyone expects it to mean that.

If the sitewide average is 3 stars, it would mean that a book with an average rating of 3.75 stars is more well-liked than average. This doesn't mean YOU would like it, but I don't think anyone expects it to mean that.

It just helps set the scale, that's all. Population effects mean it's not as meaningful as you might want - different sub-populations may rate higher or lower overall (people who have been conditioned by Amazon to think three stars means AWFUL may rate higher overall than those who haven't), and those sub-populations may systematically rate different types of books, but that doesn't mean there's no point.

>23 lorax: For me, the issue is that I know people who do not use the star ratings to mean what I mean. And unlike Goodreads, where there are descriptions of what each star rating means (3=I liked it, 4=I really liked it, 5=I loved it), there's no real way to get or keep everyone's ratings on the same page. Which isn't a complaint on my part, just so we're clear. But it does mean that aggregated ratings don't mean a whole lot.

>24 rosalita:

I'm not saying they mean a lot. I'm saying they are not entirely meaningless. And I'd argue that rigidly enforced normalization of ratings as on Goodreads would render an overall average less useful, not more - if you already know that 3 stars is going to be the overall average, then you don't need that to be computed.

>25 lorax: Yeah, I agree on not wanting the star values to be pre-determined like Goodreads does, either. And I certainly have no objections to this feature being implemented. I probably wouldn't use it much, but it's not going to hurt me to have it exist.

I notice ratings now and again (though most of the books I look up don't have many ratings) and whilst to me they suggest nothing whatsoever about a book they sometimes suggest a great deal about a member's tastes/knowledge--4 stars to this? has this person specialist knowledge that enabled them to properly appreciate something that I was too ignorant to do? 1 star to that? good lord, I so hope we own no books in common. Etc. Which lead me to profiles, some of which lead to libraries I want to investigate & some of them that have me sighing in relief (no books in common) or, once or twice, wondering whether a member's assigning 1 star to each & every book is a feeble and I'd think unfulfilling way of expressing anger about matters unbooky.

>23 lorax: Normalization.
If the sitewide average is 4.5 stars, it would mean that a book with an average rating of 3.75 stars is less well-liked than average. This doesn't mean YOU won't like it, but I don't think anyone expects it to mean that.

Precisely.

>28 Thresher:

So in other words, nothing.

A non fiction rating isn't always about liked. It may be about content. It may be about presentation. It may be about addition to the field. It could also factor in all three.

Fiction books, depending on the genre, could also come in various reasons. Literary merit, addition to the genre, quality...
Normalization says all reasons are bubkiss, and we're going to say that 3 is book is okay, 4 is like, five is love... Regardless of what ONE person wants their star to mean.

>29 gilroy: Oh, nonsense.

Do you think many people rate books deliberately higher when they're full of spelling and grammar mistakes? When they have major gaps in logical reasoning? When there are continuity errors?

There are certain characteristics that will pull down a rating despite everyone's different preferences.

>23 lorax: by way of >28 Thresher: If the sitewide average is 4.5 stars, it would mean that a book with an average rating of 3.75 stars is less well-liked than average. This doesn't mean YOU won't like it, but I don't think anyone expects it to mean that.

Yet if I only rate my favorite books and give them each 5 stars then my average is 5 stars but that doesn't mean that I thought the books were average. Similarly if the site-wide average is high it's probably because people try to choose books that they think they will like and most of the time they're right. So if the site-wide average is 4.5 then a book with an average of 3.75 stars then that does not mean any individuals think that book is below average, most of them probably think the book is pretty good.

>30 _Zoe_:

For a single book, that's fine.
For a single Author, yeah, still pretty decent.
For a single genre, it starts to look a little off, but I'm still seeing it as on the edge of okay.
For a single tag, this gets iffy. Because a tag like "history" will start crossing fiction and non fiction

For the entire site, that wouldn't make any difference. Which is what I'm trying to say.

We have more than 111 million books catalogued, and it keeps growing. If only 1 percent of those had star ratings, we're still talking over 1 million books to average together over type (fiction versus nonfiction) and genre. One book, with a low rating, isn't going to effect the overall site average. Even 10 percent of that number, so over 100,000 books, isn't going to draw that average down a full star because they are a half star to every other 3 star and above.

And as >31 jjwilson61: points out, many people (definitely not all) will be cataloguing higher rated books because they like them. Me, I rate based on my reading. So I have a pretty decent bell curve to my star ratings. Not everyone will.

>32 gilroy:

many people (definitely not all) will be cataloguing higher rated books because they like them.

Sure, which is why we might expect a sitewide average of 4 stars rather than 3. Doesn't mean that 4 is not still greater than 3.

>33 lorax: But it does mean that knowing the overall site rating average does nothing to inform you about the meaning of an average for a particular book.

I don't really think there's a point in continuing this discussion; I think that individual vagaries of rating will most likely average out overall so that of the people who rated it, people liked a book rated four stars more than one rated three stars, even if there are no raters in common; you obviously don't, and since this isn't an area where we have data we're just going to pointlessly argue with each other.

>29 gilroy: So in other words, nothing.

“Nothing”? Seriously, nothing!? This average would convey no information at all?

A non fiction rating isn't always about liked. It may be about content. It may be about presentation. It may be about addition to the field... Fiction books... Literary merit, addition to the genre, quality...

If I can't interpret your richly complex star-assignment algorithm with the site-wide average in my head, what would let me interpret it without the site-wide average in my head? In the latter case I have less information.

If stars are meaningless with that info, they’re certainly meaningless without that info!

(If you’re seriously arguing that star ratings are completely meaningless to anyone other than the rater, then, er... props to you for your internal consistency, I guess. But I don’t agree.)

Normalization says all reasons are bubkiss, and we're going to say that 3 is book is okay, 4 is like, five is love...

? Where did that come from?

>34 jjwilson61: But it does mean that knowing the overall site rating average does nothing to inform you about the meaning of an average for a particular book.

If I can't interpret a number of stars with the site-wide average in my head, what would let me interpret it without that info?

Honestly, you two. Plainly it makes a difference if the site-wide average is 4.5 versus 1.5.

What would be even more interesting would be a plot of all the stars for a work so that you could see whether they follow a normal distribution. Pony time! Informative? I don't know. I would file it under games like the "Will you like it?" algorithm.

>37 krazy4katz: The data is already shown https://www.librarything.com/work/60626/members/134415764 (in the All members section). So all we need is the numbers to be printed and plotted..

If the site-wide average were 1.5 all it would mean is that most people are reading books they don't like. But it doesn't make a book rated 3 any better thought of than it would be if the site-wide average were 4.5.

>37 krazy4katz:, >38 AnnieMod: It is shown as a graph on the right of the work pages. I look at it all the time.

The data is not just there, it is plotted. Normally things do not show a normal distribution because a lot of people don't use half stars. But once you adjust for that, many books have a cluster with most ratings around the same point. Others show more variation.

http://www.librarything.com/work/13805455.

>36 Thresher: I find it interesting that the person arguing so vehemently for this feature has star rated just two books in 8 years. *shrug*

Also, having looked at most of the people commenting in this thread, your average star rating, JUST FOR THIS THREAD, is around 3.71. (Had to rerun my numbers, sorry)

What does that tell you?

>32 gilroy: " So I have a pretty decent bell curve to my star ratings."

Centered around what value though? It would only be 2.5/3 if you were picking books completely at random and had an equal chance of liking them or not. I suspect this isn't the case. And is much less likely to be the case on LT as a whole. It's one of Tim's main rational for not doing anything with ratings, - like incorporating them into recommendations which would otherwise seem to be a logical step - people choose books that they think they'll like, and that has more data than whether or not they actually like them. I'm not sure I completely agree, but I understand the viewpoint.

Hence I think it would be of use to know what the LT average is - I'm sure it's higher than 3 - so that for any suggested popular book ( more than 20?? ratings) You can see whether or not it's widely liked in general.

I think the ratings graph used to be higher up on the page, so that it was visible without scrolling.

>42 reading_fox: Actually, you are perfectly right. It centers around 3. Because I have been reading more at random, trying new authors, and trying to expand my overall reading footprint.

When I checked on several other profiles, their curve is more like a wave or snail shell, with it centering on 4.

My curve certainly centers on 4, although my average is 3.48 Which feels about right. I aim at reading books I really enjoy. So most books (over half) get a four star rating if I rate them at all. A few get more, but they have to really stand out. So there is a lot of room below for various degrees of 'This book didn't really do it for me.'

>38 AnnieMod: >39 jjwilson61: Sorry! I meant to be talking about the site-wide average that the original poster wants. My mistake (and definitely a pony).

I do look at the individual book star distribution sometimes. They are almost never a normal distribution.

>32 gilroy: "We have more than 111 million books catalogued, and it keeps growing. If only 1 percent of those had star ratings, we're still talking over 1 million books to average together over type (fiction versus nonfiction) and genre. One book, with a low rating, isn't going to effect the overall site average. Even 10 percent of that number, so over 100,000 books, isn't going to draw that average down a full star because they are a half star to every other 3 star and above."

Yes, the average of a lot of numbers is not substantially affected by one number. Why does that mean the average shouldn't be calculated?

>36 Thresher: "'Nothing'? Seriously, nothing!? This average would convey no information at all?"

Yet, somehow, Tim's "No" review of a book is packed with information. Ya gotta love it!

>39 jjwilson61: If the site-wide average were 1.5 all it would mean is that most people are reading books they don't like. But it doesn't make a book rated 3 any better thought of than it would be if the site-wide average were 4.5.

Whoa, Nelly! I don't agree with that at all!

>41 gilroy: I find it interesting that the person arguing so vehemently for this feature has star rated just two books in 8 years.

Yeah, so? I find it interesting that the person who is saying stars convey no information has star-rated 30 out of the 30 books on the first page of their reviews (that's all I looked at). if they're meaningless, why bother? Come on, gilroy!

>49 Thresher: Yet you don't give any reasons.

>50 Thresher: Actually, if you read my statements, I said on a per book basis, they do make sense. Because that is the average for EACH BOOK.

Site wide would be useless, because it wouldn't really tell you anything.

In fact, you ignored my question in >41 gilroy:.
Your average star rating for the 24 people commenting in this thread is around 3.73. What does this tell you?

>52 gilroy:

But if you don't know what an average for a book means, you don't know whether 4.5 average rating for a book means "great" or "acceptable". You need some baseline for normalizing that; you can either just assume an average rating is 3, or 4, or whatever, or you can use the available data.

>53 lorax: It isn't valid to use the overall average for normalizing the averages on individual books because the ratings that make up each person's average is almost certainly a biased sample of all the books they have ever read. It makes more sense to assume that a 3 means that their experience of reading that book was around an average experience for most people than it does to assume most people think that 3.5 is such an average experience just because that's the site-wide average.

>54 jjwilson61:

If each individual person has a 3 as their average (which is what I try to do), then why on Earth wouldn't the overall site-wide average reflect that? I'm talking about averaging ALL ratings, not just averaging the average.

I really don't think that many people adjust their rankings to try to make their average hit any particular mark. I think that most people just rate their books using an internal scale from bad to good, and because they aren't choosing books to read at random and because, at least in many cases, they will abandon a book they don't like without finishing it, and without rating it, those ratings will skew to the high side.

However, I don't think that changes what they mean by the rating values. That is, I think that a 3 still means average to me even though the average of all my ratings is 3.74. And that still applies when you apply it to across all of LT.

That is, I think that a 3 still means average to me even though the average of all my ratings is 3.74

But...but...that is not what average means. This is not Lake Wobegon, not all of your books can be above average. They can all be "better than a book selected at random off the shelf", but that's not what you're measuring the average of.

I don't choose to read books at random, either. So for me, an average book is one that I enjoy reading. But that's *average*. Three stars.

Whilst I don't give stars to many books, I can see that my average is distinctly skewed by the fact that I almost always get rid of books I've given or would have given fewer than 3 stars to. I wonder if this is true for many members.

>58 bluepiano: Looking at a lot of the profile graphs, yup. People tend to catalogue and keep only the books they like or that they rate above 3 stars. Many graphs look like a snail or wave, large at one end and flat on the other.

>57 lorax: Why can't all of the books that you choose to read be above average? There are a ridiculous number of books published each year; it's impossible to read anywhere near all of them. If I read an average book (meaning middle-of-the-road in terms of all the books available to me), then that's a disappointment because I could have spent the time reading something better.

"Average" has many meanings. If someone wants to compare the books that they read to the standard of a "a book selected at random off the shelf", that's their prerogative.

>60 _Zoe_:

Well, sure, and someone can also use the star ratings as a count of how many times they've reread a book. But I'd sure be confused by someone saying "Everything I read is above average" when what they mean is "I am not totally incompetent at selecting books that I enjoy reading". Am I really the only one?

>57 lorax: But...but...that is not what average means.

I see I'm guilty of using average in two different ways. When I'm talking about an individual rating an average rating would be one that is in between worst ever and best ever. It's the average of possible ratings, not the average of my actual ratings.

>61 lorax:

I think the books I read tend to be above average in some objective sense. I think Orthodoxy and The Brothers Karamazov and Great Expectations and Middlemarch and Godel, Escher, and Bach and To Kill a Mockingbird and A Grief Observed and The Lord of the Rings are objectively better than the vast majority of books published.

Of course, I could be wrong. Hence, the value of averages.

>57 lorax: This is not Lake Wobegon, not all of your books can be above average.

To put it another way, it depends what your averaging. All of my books can indeed be above the average of the whole universe of books. Just as all the students in a school for gifted children could be above the average of all students.

>64 jjwilson61:

Well, sure, but I when you were saying "average" twice in the same sentence but applying it to different universes I was a bit baffled. (Sort of like the time one of my profs used "rho" twice in the same equation but referring to different things. Either one, in isolation, was valid. But not both at the same time!)

A brief digression, where do you find your personal ratings average? If I ever knew, I've forgotten.

>66 PhaedraB: Profile, Stats/Memes, right hand side at the top. You have a graph of all your stars and a number with your actual average rating.

>67 gilroy: Oh, goodness. I must have looked right at it without seeing it.

Thanks.

I've hesitated to poke my head up, here, because it's been interesting to read, but I didn't have much to add. I just wanted to point out that my average is 4.5 stars, but that still conforms to what you'd expect. Most of the books I've added to LT are the remains of a much larger library (from before I retired). In addition, they're the tiniest fraction of books I've read over the years. In the beginning, I gave stars to books that I'd bought *just* for the cover art (but have tried to go back and remove them where I notice). The majority of books I own that have between 4 and 5 stars are ones that I've had for many years, and that I kept because I loved them.

Anything I don't love was gone, long ago. I have read some things since I retired that were very worthy, and I've rated them accordingly. I note that I have only rated 85% of my library, and that's probably still higher than I expected (which means I still have some out there that probably need to be unstarred). I've reviewed less than I've starred (416), and I think that's more meaningful (about two thirds of my total books, which is 657 unique works, give or take).

I'm probably more generous than I should be; I tend to give 3.5 when it ought to be 2.5, for example (and I don't know why, since the author of said book is not around to notice). There are a LOT of books in my library that have 5 stars, and some of them should probably be 4 or 4.5, but it's hard to separate love from value.

I do have to agree with a larger point most have made here, though, which is that a site-wide average doesn't really show much in the value of a book. I'm sure that those Harry Potter books rate very high, but I'm not going to rush out and read one on that recommendation (actually, I tried the first one, and was glad it was from the library, and not something I'd given money for).

My average is 3.35. I only give star ratings to books that I've received as reviewer copies. Those star ratings reflect my general level of satisfaction with the book, and I give them as part of my "service" as a reviewer. But in general, I don't star-rate books in my catalog.

There is no LT-wide guideline for stars even as basic as the one I gave above ("my general level of satisfaction with the book"), and there is thus no genuine basis for aggregating star data across different user accounts. On a game cataloging site that I use, item ratings are on a ten-point scale, with clear guidelines that are oriented around the criterion "How likely are you to want to play this game?" People still rate idiosyncratically, and a small minority even willfully abuse the rating system. But at least it has a designated meaning.

My average rating for my catalog is a 3.8. This is partly because I try to select only good books and partly because if a book is really bad (in my opinion) it goes into the "officially unfinished collection and I don't rate it. That is why I think any site-wide average would be more for fun that for real information.

>71 krazy4katz: I make a point of rating the books that are so bad I can't finish them, just to maybe warn other prospective readers. I usually also add a review to say why.

My average is 3.49, my most frequent rating is 4. I expect the books I read to be better than 'average'. The only books I don't usually rate when reading them are the series I reread again and again.

I’ve been reading this with some interest, too, mostly because I think it’s a futile attempt. One thing is that if someone doesn’t know how to use a five-point system by now, nothing is going to teach them. Also that I think many of us choose books we think we will enjoy, starting off the ratings to an advantage. Then there’s the DNF. Sometimes I rate them, but usually not, but I sometimes keep them in a collection where they can’t infect anything else with their cooties. If they're a library book though, they might not be in my catalog at all. Of the books I do finish and don’t like, I do rate those, but they sometimes end up in the same Kicked to the Curb collection. Books I enjoy I keep and rate, usually generously. So with all of that, it’s not a surprise that many of my books are 3 stars or higher.

jjwilson61 at 39: If the site-wide average were 1.5 all it would mean is that most people are reading books they don't like. But it doesn't make a book rated 3 any better thought of than it would be if the site-wide average were 4.5.

Me at 49: Whoa, Nelly! I don't agree with that at all!

jjwilson61 at 51: Yet you don't give any reasons.

I don’t have to give reasons for saying that 1.5 conveys different information from 4.5; it’s obvious. You have to give reasons for saying 1.5 conveys the same information as 4.5.

Indeed, you admitted this when you said, If the site-wide average were 1.5 all it would mean is that most people are reading books they don’t like.

That would indeed be a reasonable inference to draw. And 4.5 would mean something different.

>54 jjwilson61: the ratings that make up each person's average is almost certainly a biased sample of all the books they have ever read

Quite. And if so, that would be useful to know.

>52 gilroy: Actually, if you read my statements, I said on a per book basis, they do make sense. Because that is the average for EACH BOOK.

Site wide would be useless, because it wouldn't really tell you anything.

lorax at 52 suffices as a response to this.

In fact, you ignored my question in >41 gilroy: gilroy:.
Your average star rating for the 24 people commenting in this thread is around 3.73. What does this tell you?

I ignored it to be nice to you, because the answer is so devastatingly obvious. But since you insist on being taken to school...

It tells me nothing of interest to me, which is why I’m interested in the site-wide number.

In fact, this debate has convinced me of the opposite of what some have argued: I'd now like even MORE information than I originally mentioned, not less. Specifically, I’d like to see the entire distribution of star ratings, not just the mean. That could tell us, e.g., whether a 4.5-avg book was in the top 50%, top 20%, top 10%, or top 1% or whatever.

>74 Thresher: I never said that there isn't a difference between a site-wide average of 1.5 and 4.5. What I said was that it doesn't make a difference in how to interpret a particular books average. And I have given plenty of reasons. If you want to argue further you need to say why my reasons are wrong.

>75 Thresher: It tells me nothing of interest to me

Okay, so if an average star rating of the 24 people discussing this idea means nothing of interest to you, why do you think a site wide average would be any better?

>76 Thresher: Uh, go look at the Zeitgeist page. Most of what you ask for is there already.

I'm interested in this discussion, as I was trying to find if LT has a place where I could search by Genre &/or Author & then by rating. It would be helpful for when I go to the Library looking for books. I tend to like Series, Historical Fiction, Mystery & light reading. I've been using LT for year for keeping track of my books and would now like to figure out how to use it to help me find new authors and books of interest. I agree Ratings are subjective and some author write a good series, but then their other series are not up to the same level, so it is hard to rate. But I feel it would be helpful. Any help you can give me to help figure out how I can make LT more useful to me, would be greatly appreciated. Thank you. ps..duh...didn't realize what the Zeitgeist was or what it was for, so never looked at it. It may be a bit of help. Thanks for clarifying.

>79 GailL: You can see the ratings of an author's books. Go to the author page. (for example Neville Shute) On the left is a link for rating statistics. You see that for someone like Shute this doesn't work well. There are too many anthologies and abridged versions for the sorting to be really helpful.

>79 GailL: I've been using LT for year for keeping track of my books and would now like to figure out how to use it to help me find new authors and books of interest.

The user recommendations feature is fairly helpful, in my experience.

>78 gilroy: if an average star rating of the 24 people discussing this idea means nothing of interest to you, why do you think a site wide average would be any better?

“If a poll of 24 people isn’t good enough, why would a poll of 100,000 people be any better?” Seriously?

Uh, go look at the Zeitgeist page. Most of what you ask for is there already.

Good, if true, but I couldn’t find any of it: the site-wide avg stars as well as the detailed breakdown by 0.5 stars, 1 star, 1.5 stars, etc.
Where is it specifically?

>77 jjwilson61: And I have given plenty of reasons. If you want to argue further you need to say why my reasons are wrong.

Where? What reasons (especially, that haven't been addressed)?

And your attempt to reverse the burden of proof is noticed. I don't have to argue that not everyone uses 3 to mean average; you have to argue that they do.

But it's too late anyway, because there's now an example that not everyone does. See next post.

>39 jjwilson61: If the site-wide average were 1.5 all it would mean is that most people are reading books they don't like. But it doesn't make a book rated 3 any better thought of than it would be if the site-wide average were 4.5.

Talk about asserting something without giving reasons for it! I can just as easily assert the opposite, and with a rationale:

People don’t necessarily use 3 out of 5 as the benchmark for “an average book.” One could argue that it should be 2.5 out of 5, obviously. Indeed, look at post 15:
I know what I mean by 5-star - those are the books that are exceptional... But 2.5-3.5 kept wandering up and down. Now I can check my profile and make sure I agree with myself, if no one else...

Even back before grade inflation, an average student was a C student, which was not a test average of 50%, more like 75%. At Harvard, apparently the average grade now is an A-. And this has changed over time.

So no, it is not axiomatically obvious that everyone uses C to mean average grade or student, or 3 to mean average book.

>84 Thresher: At Harvard, apparently the average grade now is an A-. And this has changed over time.

Oh I'm sorry, but I demand citation for that fact.

>82 Thresher: So your request in >76 Thresher: was for : Specifically, I’d like to see the entire distribution of star ratings, not just the mean. That could tell us, e.g., whether a 4.5-avg book was in the top 50%, top 20%, top 10%, or top 1% or whatever.

Which gets answered here:
https://www.librarything.com/zeitgeist/books

Top 25 books by rating
Bottom 25 books by rating
Top 25 books with greatest deviation.

If a poll of 24 people isn’t good enough, why would a poll of 100,000 people be any better?” Seriously?

YES Seriously. You have no base question; you have no theory, no hypothesis, you have no science to tell you WHAT the answer to the poll means. All you have is the answer.
What does that mean? 100,000 will be no better than 24.

Just to put in another perspective, my 3 rating is often for a book I enjoyed but isn't great literature. Also, there is nothing anywhere that specifies this has to be a linear scale. Could be log2. My mind certainly isn't linear when it comes to rating books but maybe my brain was miswired during development.

>85 gilroy:

Mean reported GPA for class of 2015 is 3.64, just below an A-: http://features.thecrimson.com/2015/senior-survey/

>86 gilroy: Which gets answered here...

No, it doesn't. I wasn't asking about the 25 lowest-rated books, etc.

"If a poll of 24 people isn’t good enough, why would a poll of 100,000 people be any better?” Seriously?

YES Seriously.

(Facepalm.)

>88 lilithcat: Mean reported GPA for class of 2015...

Thank you lilithcat.

>89 Thresher: (Facepalm.)

Ah. Finally seeing the failure in your own arguments...

>90 gilroy:

You've said many thousands of data points is no better than dozens of data points. Back to Remedial Statistics with you!

I can't help you.

>91 Thresher:

No. I said having no question to go with many thousands of data points is no better than having no question and dozens of data points.

Google “summary statistics” and/or “descriptive statistics,” gilroy. I’m sorry if this is inconvenient for you, but this is an actual thing.

From https://en.wikipedia.org/wiki/Summary_statistics :

In descriptive statistics, summary statistics are used to summarize a set of observations, in order to communicate the largest amount of information as simply as possible. Statisticians commonly try to describe the observations in

a measure of location, or central tendency, such as the arithmetic mean
a measure of statistical dispersion like the standard deviation...(etc.)

And https://en.wikipedia.org/wiki/Descriptive_statistics says,

Descriptive statistics provide simple summaries about the sample and about the observations that have been made. Such summaries may be either quantitative, i.e. summary statistics, or visual, i.e. simple-to-understand graphs. These summaries may either form the basis of the initial description of the data as part of a more extensive statistical analysis, or they may be sufficient in and of themselves for a particular investigation.

And http://www.statisticshowto.com/summary-statistics says,

“Summary statistics summarize and provide information about your sample data. It tells you something about the values in your data set.”

*sigh* You are mixing up details. When they pull Summary and Descriptive Statistics, these are all based on like data and related subjects.

You would be pulling from differing subjects and not necessarily like data.

So that's a fail.

I'm new to this discussion and want to stay out of what the conversation has become. I'd like to weigh in on the original post.

I too am curious about the star distribution for all books on LT. I spent the morning looking for a discussion topic (apparently I can't see), went to a meeting, came back and posted my question - and immediately after I posted - saw this topic...go, figure.

I do understand that there may or may not be utility in this information, based on the subjectivity of the rating system. But I am still of the opinion that the information would be useful, at most, and interesting, at least.

I have noticed on Amazon that just about every book I looked at had a star rating of 4 or above. I have also noticed that when someone writes a review that might have a lot of merit, but be a 3 star or lower rating, it tends to get voted as not helpful. So, that got me thinking about the reliability of Amazon's rating system. I wanted to check it out. Since I already keep a database of my own for the books I read, it was pretty easy for me to add the Amazon ratings for the 36 books I read this year. I also thought that I would add LT and Goodreads ratings for those books as well. For the January and February 2017, the average rating of the books I read was 3.80. Of those same books, the average rating on Amazon was 4.25, on LT 3.73 and on Goodreads 3.96.

Since the rating on Amazon was so much higher than mine, that got me thinking about whether there were any books on Amazon that had a low rating. Since I was on the Kindle page, I looked at the star ratings that could be used to find books - it appears that 77.72% of books have a rating on Amazon of 4 & above, 15.95% from 3-4, 3.48% from 2-3, and 2.85% 1-2. (Since I couldn't get the data, I just assumed that there was nothing below a 1 - which I know is not correct, but...) So it still seems as though books are rated higher on Amazon. They certainly don't form a bell curve. I decided to look at my own books because I suspected that mine would also not follow a bell curve either. Since the Amazon data was 4 & up, I added all of my 4.0, 4.5 and 5.0 rated books. Since the Amazon data started at 1, I added all of my 0.5, 1.0 and 1.5 books as 1.0. So my layers were basically 1-2, 2-3, 3-4, 4-5. The 1-2 layer and the 4-5 layer include 3 tiers of ratings, while the 2-3 and 3-4 only included 2 tiers. Amazingly, Amazon's average rating for those 1,874,374 books was a 3.69 - my books averaged a 3.52. 66.28% of my books fell in the 4& above layer, 30.64% 3-4, 2.95% 2-3 and 0.14% from 1-2.

The Amazon rating for those Kindle books was lower than I had expected it to be - especially since I don't think I've ever seen any books rated below a 3 on Amazon, and rarely any books rated below a 4. I'm not sure what all of this tells me... I know that I typically search for books I expect I will like so it wouldn't surprise me that the something like 97% of my books are above a 3 - and I think a 3 is an okay book. I would also imagine that the same applies to most books - people are going to look for books that they think they enjoy, so the ratings would be more likely go up, instead of look like a bell curve (especially since the 4 and 5 tiers are not segregated.) I guess one thing it does tell me is that if I see a book on Amazon that's rated below a 3, I will run like the wind away from it.

So, I would be really curious to see how LT's catalog compare. In just the books I read this year alone, it would appear that a rating on LT MIGHT be a more accurate reflection of my own tastes.

Anyway, that's quite enough of my babbling. I vote that this is something that would be useful AND interesting.

>95 rretzler: Just as something to point out, Amazon and GoodReads have their rating system gamed by authors, especially the self published who don't know any better, to make their book look shiny. Anne Rice has even been caught ripping apart reviewers who rate her stuff below 4.

>96 gilroy: Thanks, good to know. I'm sure every site has their reviews gamed to some extent. Several authors that I read have a very small library here on LT, which I suspect may be to be able to bump up their books stats. It wouldn't be that difficult to do really, and would likely be difficult to stop. The whole real names issue on Amazon recently points to all sorts of problems inherent in having the public be able to review anything anonymously. I think that's a given on any type of social media where reviews are allowed. That's why I think that being able to take a look at the composition of the ratings MIGHT be helpful.

Honestly, I think Amazon's "cure" for fake reviews is far worse than the actual problem ever was. There's been a noticeable decline in their overall review quality since they started hiding the majority of the reviews and focusing on verified purchases only. Meanwhile, concealing the voting results from our own reviews (at least on our profile pages) has eliminated a major incentive for people to review in the first place. The whole thing is just sad.

>98 _Zoe_: I totally agree that focusing on the verified purchases is a big problem. I feel like I want to take a picture of myself with the book and say "See...I read it; I have it; I just didn't buy it here."

And don't get me started on the whole voting on reviews...once many years ago, my son brought home a book from an author day at his elementary school, where the author visited, signed books and discussed the book with the children. We read it at home, and we all agreed that some of the book didn't make sense. My son communicated what the author told him about that particular part, and we just didn't "get it" based on the context of the book, even after we had the author's explanation. I was actually one of the first reviewers to review the book on Amazon, and I gave it a 3 star, but said that it was not for my family or me, explained why, etc. but indicated that I could see how many people might enjoy the book (and tried to explain that as well.) You would not believe the number of people who said that my review was unhelpful. For a long time afterward, I did not review a book on Amazon. It seems that the only reviews that people thought were helpful were those that they agreed with, regardless of whether there were any helpful comments or not. To this day, if I think I wrote a fair review, but it gets a lot of unhelpful votes, I will just delete it. It has only happened once or twice including that review. Frankly, I thought that a review that explained why someone did not like a book, but why others might, would be pretty darn helpful. But that's just my opinion, apparently.

The whole public review system just leaves a lot to be desired.

It seems every site has something that makes star ratings problematical. So, I've concluded that star ratings are problematical. I know what people want from star rating data, but it does not appear that star ratings can provide it.

>100 PhaedraB: Speak for yourself; I regularly get what I want from star ratings. They remain a valuable comparative tool even without having some absolute fixed meaning.

>99 rretzler: Yes - on Amazon, for books and for other things, I go to the low-rating reviews first. They're the most helpful to me. Either "this is a bad X and this is why" or "I don't know how to use/read/set up X so I've rated it low" (my analysis). The high ratings are often interesting, on books, but not nearly as useful - and high-rated book reviews are most interesting if I've already read the book, or at least have read books by the author.

Unfortunately, on Amazon particularly and elsewhere as well, reviews/ratings are seen primarily as a popularity contest, so authors/sellers will try all sorts of things to get good reviews/high ratings. On Amazon, it will actually affect the sales - Amazon believes the popularity contest and makes high-rated stuff easier to find, therefore better-selling (a vicious, or virtuous, cycle - which one depends on whether you think the book or whatever deserves its high rating). LT _shouldn't_ be like that, and isn't (much) now, but Tim's disdain for ratings is part of what keeps it that way. So while I'd love to see more info in/about the ratings, I don't want to upset the applecart either.

>94 gilroy: You are mixing up details. When they pull Summary and Descriptive Statistics, these are all based on like data and related subjects.
You would be pulling from differing subjects and not necessarily like data.
So that's a fail.

"You don't understand, Thresher, these are magic data. Descriptive statistics don't apply to them!"

There definitely is a failed argument here.

>103 Thresher: Yeah, I knew you'd go there. So let me boil it down to simplistic terms you aren't catching otherwise:

Dataset: All things cataloged on LT (This includes dresses, perfumes, movies, games, etc.)
Subset 1: All things that have stars (You still are including all things that aren't books)
Subset 2: All books (Some narrowing, but still huge chunk of data)
Connection between two subsets: All books with stars.

To most scientists (especially when you talk with polling places) this is still too large a subset to review for relevance. Otherwise, polls would be done to cover 1/10 the population of the US in order to get a proper accuracy. No, this set is just too big. So it must get smaller.

So we have to get a subset of the connection: Subset of all fiction books. Subset of all nonfiction books. (Oh wait. These definitions can't be agreed upon, as proven in a few other threads.) Subset of all romance novels... using which definition? Originally, all books not considered non fiction were categorized as romance. It was more recent times that more and more genre definitions appeared. And even those lines tend to blur.

Star ratings of all legacy libraries? Don't know the owner's views for that, necessarily.

Give me a proper subset of the primary data collection other than just all books with a star rating, and I'll consider this useful.

>104 gilroy: I'm not at all convinced by your argument that a few ambiguous edge cases makes the entire classification useless. For the vast majority of books, there's no disagreement about the fiction/non-fiction classification, or about whether they're romance novels.

>105 _Zoe_: You take one portion of the argument and say it's all wrong?
Okay, fine, even if you take all Fiction or all Non fiction, that's still an especially LARGE subset, which most statistical researchers would want to still narrow.

>106 gilroy: Yes, this is how arguments work. If just one part is wrong, the whole thing fails.

But I can question other parts of your argument if you prefer. For example, you talk about polling and the necessity to look at narrow subsets of the population. Yet it's generally accepted that actual election results are better than any poll.

Can you provide an external link that explains the point you're making about the necessity of narrow subsets?

>107 _Zoe_: Has anyone else been asked to provide the same?

http://www.aapor.org/Education-Resources/Election-Polling-Resources/Sampling-Met...
A link

>108 gilroy: Thanks. But the basic premise of that page is that you need to take a sample because "It’s impractical to poll an entire population—say, all 145 million registered voters in the United States." I don't see where it says that smaller samples are actually *desirable*.

As to your question, I don't think it's particularly productive for the site to judge feature requests based on whether people can provide some sort of objective proof of their value. If there's interest in a feature, and it wouldn't require much effort to provide, and it wouldn't harm other people's experience of the site, that should be enough. What's gained—for you, or for anyone—by insisting that a datapoint that people are interested in doesn't actually have any value?

>109 _Zoe_:
Well, if we really want developers creating something and not working on fixing things that need to work, like Touchstones, go ahead, make this.

>110 gilroy: I don't think the scale of those two projects is remotely similar.

And I do think there's some value to interspersing minor improvements among the major ones, since it increases the likelihood of an individual user getting at least *something* that they care about.

Since I've already posted and at the risk of going down in flames - I thought I would weigh-in on what seems to be an ongoing argument, and possibly defend my position as to why I still think ratings add value. I'm not going to back up what I say with proof, these are MY opinions, and I've found throughout the years that what that means is that most people will likely NOT agree with me. I can live with that.

Since I've already said that ratings are both useful and interesting to me, I'll mention it again. Yes, the rating systems leave a lot to be desired, however, even when others are trying to game the system, I think that a person who understands all of this can turn the ratings into a useful feature. When I look at a rating from anywhere, I consider a couple of things:

First, where did the rating come from? I had the sense that ratings from Amazon were a little high, and I think I proved that to myself to my satisfaction. If I can compare overall site rating of books that I have read with my rating of the book, that helps me. In general, the more data I have about ratings on a particular site, the more I could satisfy myself that I am likely to find that, for example, LT provides me with ratings that are generally closer to my views.

Second, how many people rated the item in question. If only ten people rate something, it is going to be subject to the effects of gaming the system more than an item that has 500 or 5,000 or 50,000, etc. The more people rate something, the more valuable the rating becomes IMO. The larger the pool, the less likely the outliers are going to affect the overall rating.

Third, I realize that ratings are VERY subjective. Even I might not rate the same book the same way on different days. I give you the fact that I recently rated a classic sci-fi book that won the Hugo, the same four-star rating as a "cutesy" mystery book that has no literary merit only because I thought it was a fun book. I enjoyed both books, but for different reasons. I also gave a James Joyce book a one-star rating because it was narcissistic and unreadable when many others agree that it is a literary classic.

Last, I read the reviews. Like another poster, I actually read the bad reviews first, and then I read some mediocre reviews and then some good ones. A review that just says "I enjoyed this book" or "I hated this book" is not very helpful to me unless I know the reviewer. But I can glean some helpful things from the reviews - grammar is a big issue for me, so I'll stay away from a book if more than a few reviewers complain about the grammar. Likewise, if there are a lot of complaints that something doesn't make sense, I'll probably steer clear. I love a review that tells me clearly what the reviewer liked and disliked about the book; unfortunately, I, personally, have difficulty writing reviews that convey that type of information.

Any rating system is going to be totally subjective. For instance, someone who may primarily read only romances might read a sci-fi book that was recommended to them by a friend. They may not like it because it isn't their usual genre, however, for a sci-fi reader, it might be a great book. Someone's dog may have died recently, and they read a book about a dog getting killed, which may have turned them off the book. The list could go on. Everyone's ideal rating system is going to be different. The perfect rating system for me would be to have several million of my clones read and rate a book - then and only then would I know I had a rating that I could rely on! ;-)

I may give people too much credit, but I think, for the most part, people understand that ratings are subjective and to be regarded with caution, but that they still get utility out of a rating system.

So, I would like to be able to see the overall average ratings for a site - because I think it will tell me more about the general overall ratings on LT. Plus, it's just interesting!

And I do agree that LT should work on little enhancements along with much-needed improvements for many reasons. It gives the people what they want (or maybe some of the people), it gives the programmers a break from big tasks, it adds a new feature to the site, etc.

I find that for any given book, the distribution of ratings tells me more than the average. I think it would be interesting not just to know that the site average is x.x, but also that that includes so many of each particular rating.

>113 MarthaJeanne:

The zeitgeist section "25 Books People Can't Agree On" gives the site-wide winners for highest standard deviation in ratings. It's among the few ratings-based data that I find interesting.

>114 paradoxosalpha: but that is just 25 books, and the number given doesn't mean anything to me.

>113 MarthaJeanne: Look down the right side on the main book page, right under Popular Covers - the box gives both the average and how it's derived (how many at what star level). The top data just gives the average. And yes, that's an interesting bit of data for a particular book.

>104 gilroy: Yeah, I knew you'd go there. So let me boil it down to simplistic terms...

Let’s review the moving of the goalposts.

In 29 you bizarrely said, Normalization says all reasons are bubkiss, and we're going to say that 3 is book is okay, 4 is like, five is love...

despite the fact that no one is saying that, and indeed we don’t have any basis for saying that or rejecting it without seeing the info, which you’re objecting to seeing.

In 58 bluepiano said, my average is distinctly skewed by the fact that I almost always get rid of books I've given or would have given fewer than 3 stars to. I wonder if this is true for many members.

In 59 you responded, Looking at a lot of the profile graphs, yup. People tend to catalogue and keep only the books they like or that they rate above 3 stars. Many graphs look like a snail or wave, large at one end and flat on the other.

So you made a claim - people tend to discard books they give below-3 ratings to - based on the distributional properties of people’s reviews... exactly the sort of info I’m interested in (see 76).
So you yourself made an assertion you couldn’t have made without the kind of info you want to NOT be made more easily available.

On one hand, you say
So how would a site wide aggregate of every rating mean anything? (20)
Site wide would be useless, because it wouldn't really tell you anything. (52)

In 29 you say,
A non fiction rating isn't always about liked. It may be about content. It may be about presentation. It may be about addition to the field...
Fiction books, depending on the genre, could also come in various reasons. Literary merit, addition to the genre, quality...

But in 52 you said, on a per book basis, they (star ratings) do make sense. Because that is the average for EACH BOOK.

...which contradicts your own statement from 29: If you really believe that users’ different criteria makes ratings meaningless to anyone but the rater, then your statement that ratings make sense for each book, is false.

Your objections aren’t even internally consistent.

Then, in 78, you said the real objection was that these summary statistics already exist anyway, on the Zeitgeist page.

When pressed for where exactly they were, you again asserted that the info existed already, and provided a link to the Zeitgeist page, where the information does not, in fact, exist.

Worse, in 110 - without any acknowledgement of the foregoing - your objection becomes that the stats don’t exist, and it’s not worth the site staff’s time to create them.

In 78 you also said the millions of ratings site-wide wouldn’t any more informative than a sample of 24 people’s ratings.

When pressed on this bizarre assertion you actually doubled down.

Then, you tripled down in 108, with a link that doesn’t support your assertion at all - it doesn’t say small samples are statistically better; it just mentions cost of sampling, as Zoe noted.

In 86, you said that the statistics would be meaningless because: You have no base question; you have no theory, no hypothesis...

You reiterated this in 92.

Then, after I refuted that by establishing that summary statistics/descriptive statistics even without a hypothesis are standard practice among statisticians, you tried to revert to your original position that the stats would be meaningless because the items are incomparable.

Whew! It’s fine to have two or more mutually consistent reasons for a position. But I’m getting exhausted chasing these shifting arguments, ad hoc rationalizations, mutually contradictory points, and constantly moving goal posts!