Probably worth noting that I was here assuming a hard threshold-based system of Speech checks, as in New Vegas. That may or may not be the case for PE, but I think at least some of this is still salient regardless. Oh, and Fallout obviously has a classless skill system and PE doesn't, so.
I'm also going to censor some of the curse words, as I'm not yet sure of this forum's policy regarding them.
Regarding the skill check debate:
I've thought a lot about this, because it's one of those eternal problems with cRPGs that nobody's ever managed to solve for both the people who want hard thresholds and the people who want skill check rolls every single time. Furthermore, I'd argue that neither type of interaction is entirely satisfying even for the proponents of each.
Skill thresholds are a clear attempt to make the mechanics behind dialogue checks less obtuse for people without a PnP background, and in this, they succeed. However, in making the mechanics transparent and fixed, they render said mechanics dull as dishwater.
If I have 75 Speech in New Vegas, and I talk to somebody and see a [Speech 50] dialogue option right there, I know immediately that A) I am going to pass that check, B) that the check's result will always be a positive one for my PC, and C) that I'm not going to regret passing the check in any way later.
That system is, I think, a better one for video games specifically, because every player instinctively understands why he or she failed and how to remedy said failure. If someone's never played a PnP game before, they're going to be angry at the game when they arbitrarily miss a check, because all they understand is what's on the screen. If their Speech is 90, and they fail a Speech check, they don't understand it as a roll that went askew, because there's no die rolled in front of them and no DM to argue with. It simply reads as the game deciding to f**k them over. And even if you are a player who, at the least, understands intellectually that dice are your only measure of success, I'd argue that it's still aggravating as hell, because all other video games train you to expect your failures to be your fault. In an action game, when you appear to connect and don't, you blame the game for being cheap and call it bad design.
(Sidebar, skip if tl;dr: This is why I'm always preaching the gospel of good visual feedback (and let's face it, most old-school RPGs had crap visual feedback, even if you understood intellectually what was happening); if the die roll is hidden, it might as well be imaginary, and that means the game ends up feeling unfair, even when you know it isn't. That, more than anything, is the barrier to new publisher-funded "old-school cRPGs." You can argue all you like that it doesn't matter, but when everything else on the market is designed to be grokked in a WYSIWYG manner, and the genre we're all fans of works on a principle that is inherently not WYSIWYG-friendly, the fact that publishers want it to be a WYSIWYG thing makes complete sense. My feeling is that they're right about WYSIWYG, but wrong that the only way to achieve WYSIWYG is through real-time action-game mechanics. If you have visuals that clearly communicate success, failure, and why the success or the failure occurred, the player won't feel cheated. It's the lack of same that makes even many PnP players long for real-time or RTwP combat in their cRPGs.)
ANYWAY, the threshold system eliminates all of these problems, but it introduces a new and, depending on the player, equally aggravating problem; there's no sense that you've actually overcome a challenge, because there's never any risk involved. You don't have the required points, you don't succeed. You have the required points, you succeed. In New Vegas specifically, you can't even have the chance to succeed taken away from you, because failing the check nets you little more than a goofy "That dog won't hunt, fella!" line from an NPC, and you can retry it whenever you want. This renders diplomacy naturally inferior to combat as a means of conflict resolution, because combat has emergent possibilities and diplomacy doesn't.
Some would argue that this is an inherent problem with dialogue trees, and those people are right, to a certain extent. However, Wasteland 2 will presumably have a dialogue tree mechanic, and rightfully so, I say, because most interactions in an RPG work best when handled through a big, diverse dialogue tree. We'll also have a text parser, but I think that's best used for passwords and name-dropping. Most complex interactions should be handled through dialogue trees, just because of the granularity of character definition you get with them.
Let's assume for the sake of argument that not one bit of my long-winded analysis above is wrong or even particularly disagreeable. I'm sure the truth on both counts is exactly the opposite, but please grant this lunatic his myriad hypotheses for the moment.
So, how do we confer the advantages of thresholds and the advantages of random skill checks while eliminating or minimizing the disadvantages of each approach? Well, I can't say with any certainty that these solutions are remotely definitive, because I am an armchair designer who is crap at math and middling at logic, but here are a few tentative suggestions:
1) Assuming there are hard thresholds, in all instances where a threshold exists, allow us to roll a skill check that has a percentage of success that goes up as we level up in that skill until it hits one hundred percent. Let's use New Vegas' transparent thresholds as an example: when your skill wasn't as high as the threshold you needed to pass, it showed a fraction like this next to the goofy little skill failure line of dialogue:
That is converted easily enough into a percentage out of one hundred, and that percentage could be used to determine our chance of success. Using the example given, that would mean we would have a ninety percent chance to pass the skill check despite not having the required skill. This is easy to understand, it makes every rank meaningful, and patently ridiculous bulls**t like someone with 100 Speech arbitrarily failing the speech check doesn't happen.
2) The result of a successful speech check should always be successful for whoever is speaking, but they don't have to lead to an optimal outcome for everyone involved. For example, say you are attempting to convince a, I dunno, bandit to let the hostage they're holding go. In New Vegas' model, a failed check would mean the hostage would not be released. In my proposed system, even if the check itself is an automatic pass, what if you then have to negotiate the terms of the hostage's release? Just because the bandit agrees to let the hostage go, that doesn't mean that he's not planning to shoot the hostage in the back. This is where something like Spot Lie comes in handy, and also some ingenuity on the part of the player. Negotiations should not begin and end in a single dialogue tree anyway.
Having multiple different skill checks in a single negotiation that all have a chance of succeeding makes for a much more intense game without compromising the roleplaying mechanics or player agency, and it requires not one bit of twitch gameplay to work.
3) The possibility of extra rewards that SniperHF [a guy on the W2 boards] mentioned could be accounted for by using the PnP mechanics of Nat 20 and Nat 1. To wit, you might automatically pass the hard threshold of success whenever you try convincing someone to lower his gun, but there's still a roll in the background to determine the outcome. So if you roll a natural twenty on that roll, the guy gives you his gun, but if you roll a natural one, the guy's gun accidentally goes off as he's lowering it and hits a party member in the leg. There's no need to make any of this clear to the player who doesn't read the manual, because for all anyone knows, that's what was supposed to happen anyway.
4) Hide the non-combat skill thresholds in the game unless the player uses an "Analyze" ability that requires some sort of Perception check, maybe PER + INT. This would be an abstraction of a PC's deductive reasoning abilities ("I don't think he means to let that girl go, fellas. See how his hand's at his side, ready to draw?").
5) This one might be somewhat controversial, but I think it could be interesting. I'm not beholden to it, but it could make for an enjoyable mechanic if done correctly. Have to-hit chance work on hidden skill thresholds like the non-combat skill checks. However, those should be capped at a ninety-five to ninety-nine percent success rate, Fallout-style, to allow for the occasional miss. That way, you get the verisimilitude of the occasional missed shot in every encounter, but you aren't missing a f**king rat you've killed thousands of and have long since ceased giving a s**t about, as in Fallout. The threshold percentage could be modifed by distance and maybe degree of cover.
There, those are a few ideas for a threshold-based system that I believe addresses a lot of the problems people raise, but confers a lot of the advantages of a threshold-based system. I could, of course, be wrong.
Feel free to use or discard any of this as needed, inXile [or, in this case, Obsidian]. Apologies for the tl;dr nature of this post. I didn't mean to witter on forever, but, you know, I never do...
Now, obviously, this isn't something I think PE should adopt wholesale, but there are some interesting ideas in there that could be adapted to PE's systems. I would never claim it's perfect, but it's a good starting point.
What say you, PE forumgoers? I'm open to any reasonable criticisms.