They aren't taking random samples from all the different kinds of players in all the different ways the game is played. They are sampling only those people who go online and want to partake in the survey. Self-selected passionate surveyors don't represent the bulk of their players who are causal. They are bungling it here.
And they know that? Remember when we talked before about how to account for biases? And they know that the passionate players are the ones who help influence the casuals.
This one is decent. I don't think the questions themselves have a lot of bias. They're just being used incorrectly to get a resulting percentage that is inherently off base.
No they aren't.
Yes. They are bungling this one as well. there are sometimes 12-15 of the same exact question over and over. Feat #1, Feat #1, Feat #3..... Feat #15. Spell #1...
Which I admitted was likely a mistake. But also... again this was literally never a point you were making until just now.
Aware of outliers isn't what it's talking about man. It doesn't matter if they are well aware of the outliers if they are ignoring them. It's saying that the outliers can deliver the most important information and WotC has said that they ignore the outliers every time. We know this because if something hits 80%, the outliers be damned. Same with 70%. They are messing this one up.
Except they aren't ignoring them. They listen to those outliers, that's why the read the comments and follow the social media community.
They just don't let the outliers shape every aspect of the game... because they are outliers. Read the article, it is about "are you getting the knowledge" such as knowing why those outliers think your product is too similiar to your competitor. Do you honestly think WoTC isn't hearing ANY of the commentary from people who say "this sucks, Pathfinder does it better"?
You can't just assume they are ignoring those voices simply because their general process means seeking 70% consensus.
You have to be joking. Crawford said straight out that they were doing this. He went on camera and said that they are putting in questions knowing or having a very good idea of how it will turn out before they ask the question and were just looking for confirmation.
And many times Crawford has straight up said "wow, this surprised us!". Also, guess what, Crawford isn't the guy reading the raw data and making conclusions. Even if they expect a certain answer, that doesn't mean they twist the data to get it. You can have expectations, built from knowing the community, and NOT have confirmation bias.
No. That's the error you are making in assuming that the surveys are the reason the game is doing well.
And still has nothing to do with the surveys. Also, seriously? The thing the team credits to their success isn't the thing that is contributing to the success of the game? How do you square that circle? WoTC is too dumb to know what is or is not working for them? They don't know that they used survey data and got good results?
Yes, Correlation isn't Causation, but that doesn't mean you can deny causation when it happens just because you don't like it.
No. They made multiple mistakes.
Nope
Are you deliberately acting this way? Because playtests is a plural word. How about you go look at all 7 and see if things are remaining stable?
So, you want to compare seven static snapshots of a slow moving process to a political campaign, where minute by minute new information and interference from dozens of sources are all working on an active system. And say those are the same thing?
By the time a political pollster has collected data and spent a week going over it, every single thing they were looking at could have changed. Crimes revealed, gaffs made, successful rallys, new funding from dark money, foreign interference, a massive societal upheaval like a mass shooting.
By the time WoTc takes their data from playtest 7 and spend a week going over it... nothing much has likely changed for the playtest. It is all still pretty much how it was a week ago.
Then do it. Show me how you can get an exact rating of 80% satisfaction by doing the math of 1000 people said they were very satisfied, 583 people said they were satisfied, 1200 said they were unsatisfied, and 188 said they were very unsatisfied. Do that math and come up with an exact satisfaction percentage that includes all of those voters.
I'm betting that you're going to ignore the subjectivity in ratings such that you could be satisfied by a rating of 66% but I might be satisfied at 61%.
You want me to take data that is nearly evenly split, and get an 80% approval from it? Like, you want me to take data, and make it fit a percentage point, to prove that they can take data and find the real percentage point...
Do you even hear yourself right now? We both know that me forcing data to match a value you choose is nothing like what they are doing.
Now, using a basic weighting method, I can get a value of ~70%, but that doesn't mean that I used the same weighting method they did.
I'll wait for you to show me above that it is possible.
Okay, I took your numbers and got a ~70% approval rating. Now what?
If they are asking direct questions that don't include reworking the ability, and you are correct that they are not indirectly asking that, then they can't possibly get a correct answer for when to rework an ability, because they don't have any information at all from us saying that we want it reworked. They are just making unfounded assumptions.
No. They are making business decisions based on their studied consideration of the metrics they want to hit.
Again, I started this conversation with this, but they'd be idiots to ask a question like "do you want us to improve this feature" because the answer is obviously "Yes!". Who would ever say "No, don't improve this"? But they have determined that if a feature has tested below a certain threshold, it isn't worth the time and effort to reiterate on that feature instead of defaulting to what they have and what works.
Because, here's a thought, how do you know that most of the features in the game from 2014 weren't testing in the 60 to 70% range? Maybe they know, from their market research, that simply going back to 2014 is acceptable, so they don't see a need to risk 50%'s when they know that means it is doing worse than what they had prior.