Please implement a better and more sophisticated rating system

LordCrash That's a great question. Many of the highest performing videos are VR hardcore debuts after all, so it's important to be able to identify promising newbies. Luckily there are techniques for that as well.

1) Were they well received in softcore porn?
If they've previously shot softcore porn, how popular were they compared to other performers shot by the same studio? E.g. Agatha Vega and Eve Sweet both shot softcore videos prior to their SLR Original scenes. From those videos, there were already clear indications that their hardcore debuts were going to be hits.

2) Were they well received in 2D porn?
Have they previously shot 2D porn? How well received were they compared to other performers shot by the same studio?

3) How well received is their look in general?
You can use the method I previously outlined to get a sense of the characteristics the SLR audience likes. Teens vs milfs, blondes vs brunettes, natural vs enhanced etc. You can also look at the top performing videos and the types of performers featured. Then you make a judgement call - you take a calculated risk based their similarities to the best performing pornstars and general preferences of the SLR audience. After you release the video, you'll get data to confirm whether you made the right call.

My main issue with the current likes are as discussed in a previous thread more that studios look at it and reject any criticism in the comments, because many people like it and few comment/complain so majority must be right. A single like-metric may give you the mainstream popular results but it also means it's the lowest common denominator result. If you want to target that it's going to be similar to TV where everything becomes shitty realityTV because that's what stats said is watched the most.

At least some better split-up way of giving a rating for different aspects of a scene would be great. At the basic minimum split up the regular likes for a scene and likes for each pornstar in a scene so popular models won't undeservingly cause high amount of likes as much on technically badly shot scenes.

Also I still believe a visual marker for having downloaded/watched a scene so people won't have to use the like-button for that would improve results. Pretty sure this is already tracked anyway.

And I agree with others here that some sort of additional feedback would be nice for scene specific things like story/direction, camera angles, lighting. Perhaps that way regular critique in comments will be taken more seriously as well. This is more to let studios know where to improve upon, not to determine what's the most mainstream popular thing.

dango789 Your system works for huge sites like Youtube, Facebook, etc. with a much larger dataset and not so well in a smaller environment where some girls only have a few scenes or if you don't want a bad director to film 20-30 scenes before you have enough data to find out that he does a bad job.

Another problem is that this only shows you the taste of the mainstream, but doesn't work to find what works for smaller groups or fetishes, where the dataset is much smaller. For example I hate typical US-Porn and pick my subscriptions by authenticity. If a studio/site wants to keep me as a user they have to provide that kind of content regularly and not just the stuff the mainstream likes. You won't find out about people like me with some standard database queries, that only check for overall popularity, but I would tell you what I like with a better rating system. You may find out about users groups you didn't even think about when you created your standard queries.

LordCrash The dangers are:

1) We may end up making decisions based on unreliable data - a skewed data set with a small sample size.
To illustrate my point, look at the number of likes compared to the number of comments below:

Even the less popular videos consistently get around a hundred likes or more. But half the videos don't even have a single comment. The more complicated you make it for users to give feedback (e.g. by asking users to type a comment or by asking for 9 different scores per video), the fewer responses you will have. Your responses will also likely be less representative of the SLR community as a whole. It will gravitate towards those who spend the most time on SLR, or very likely mainly a few hardcore users.

And there lies the risk - and it applies even more so to new models with just one video. What if you cast a milf and you only get 1 response from somebody who's not into milfs? Or what if one hardcore user hates your storyline/costume/camera angle but most other people couldn't be bothered rating? Now you are pressured to make decisions based on a very limited sample size.

The alternative is feedback from hundreds or thousands of users. Better yet, if you have access to total time watched, you'll be able to see how well the new model/storyline/costume/camera angle performed across every single person who viewed your video.

This is especially a concern for smaller groups/fetishes. E.g. @someoneX isn't into typical US porn but instead prefers studios that shoot more authentically. But users who prefer typical US porn are going to rate someoneX's favourite scenes poorly, even though they're not the niche that the studios are targeting. So now the niche studios get poor scores and are pressured to move away from shooting the scenes that their audience likes.

Instead of relying on feedback from a few users who may not be into your niche, you can effectively filter out those who aren't your target audience using likes/total time watched. Those who aren't into foot fetish simply won't like/watch the video. But you'll get hundreds of people who do.

Then you can check how your videos compare to others that serve the same niche. If others consistently perform better, what are they doing differently?

2) It could clutter up the UI

Final thoughts
Anyway, you clearly have strong opinions about this as do I. I don't think there's much point discussing this further. As long as we take the reviews with a grain of salt and the UI doesn't get cluttered, then it's not a problem.

LordCrash Those solutions help with the UI issue, but they still don't solve the core problems:

1) You cast a new model: a milf. It gets an average casting score of 3 out of 10. A grand total of 10 people voted - 7 of which aren't into milfs at all. The three others exclusively watch milf videos and were ecstatic. The studio decides to never cast the model again.

This can also easily apply to other model niches.

2) You are a niche studio that caters for a specific audience. Maybe you shoot realistic porn, like the ones that @someoneX likes. 20 people give detailed votes - 15 of which prefers typical US porn. The niche studio decides to stop shooting in the style that their original audience liked, and becomes more of a mainstream studio.

spacepirate The problems are:

• 10 voters cannot accurately represent tens of thousands of viewers and their preferences. Imagine if you went on the streets and asked 10 people if they preferred chocolate, strawberry or vanilla ice cream. You get 5 who prefer strawberry, 3 who prefer chocolate and 2 who prefer vanilla. The next day, you go out again but you get a completely different result.
  It is dangerous to draw any sort of conclusions on such a small sample size. The scores are simply unreliable and will entirely depend on who just so happens to be voting on the video.

• Studios wouldn't be happy with low scores. It would negatively affect their brand to consistently score 3/10 even if their target audience was happy. So they will likely adopt more mainstream tastes (or whatever the tastes of the few who spend the most time on SLR are) and alienate their original audiences.

Anyway, I feel that we've reached a dead end here. I've made my point clear and don't want to prolong this any further. I won't be replying to this thread any longer.