So what do you use mostly for feedback? How are you collecting data then?

On this, we used QA and playtests, and I made a series of questionnaires, and we watched the videos back. We spotted the elements that were causing frustration or that didn't score well on the questionnaire. And so many times we went back and forth between: the Idris is too easy, the Idris is too difficult, the Idris is too easy. Because what we'd be doing is identifying, okay, why is it too difficult? It's because it's firing all of its turrets at one thing at once and just blowing it up. Okay, fix that. Now it's too easy. Great. Okay, so what do we need to do? Well, let's up its shields. Okay, now it's too difficult. Okay, what do we need to do? Oh, we need to make torpedoes do more damage. Okay, now it's too easy. Right, we need to slow torpedoes down. Back and forth until finally you end up with something everyone's happy with. And by that point, you've got all that stuff I said: AI firing at turrets, AI much more intelligently, or maybe less intelligently, using their countermeasures, because too intelligent just felt cheap. If every time I fire a size 9, it just fires off its countermeasures, it's really frustrating. That's why we were saying fire a small missile, make it fire its countermeasures. It has a little bit of a cooldown before it can, so you can sneak in a size 9 missile while those flares are dying down, before it will fire another one. And that was just satisfying that way. Otherwise, it's frustrating.

And I know, besides the internal QA and stuff, you're a continual watcher of streams. I've seen you in various Twitch streams, watching how these events actually play out in the players' hands. After Prisons came out, it seemed like every stream I went to, there was Luke Presley watching and getting that raw feedback directly from our players.

And if you didn't see me, I was lurking.