NathanH

For the starting-focus nerf, I have a suspicion that the fundamental reason people were finding Ciphers ridiculously strong is that major Afflictions apply on grazes, and even with short durations are very powerful. If I'm right, this is just going to be a constant problem with casters in general, and it's only that Ciphers were best-placed to take advantage of it that they look so strong. These sorts of balance changes are likely to only scratch around the edges of this problem. Indeed, I don't see my low-level Cipher play changing that much. A good patch that balanced Ciphers would nerf their most powerful attacks without nerfing less powerful ones. The change to their starting focus doesn't really do this. Ciphers will be weaker, but I'll be casting the same spells with them. The blunderbuss nerf is a good idea, because it does achieve such a thing. The most powerful focus-generating attack has been toned down, and less powerful focus-generating attacks have been improved.

I find the almost-English words annoying. If we assume that the languages being spoken in the game world have no particular relationship with English (because why would it?) then either something should be given its in-world name (if it isn't relatable to any English word) or an English translation (if it is similar).

I enjoyed playing Baldur's Gate 2 and Icewind Dale multiplayer, so it would be a nice thing. This type of game works nicely when played with friends. It's probably too much work, though.

It oughtn't be a big deal either way, but the principle "all achievements must be physically achievable by all players" is a good one. Actually I'd probably strengthen that a bit to get rid of achievements like "post something from the game on Facebook!" which is also an annoying type of achievement. Achievements should be completely in-game in all ways, that seems reasonable.

Those gold-named NPCs are part of a secret plot to destroy the region. You must kill as many of them as you can.

I'm not sure what I'd prefer corresponds to in the voting, but I'd say: 1) Companions ought to be fully voiced, and this is the most important thing 2) The main villain should probably be fully voiced, or something close 3) I like it when you initiate a conversation with an NPC and they say a line, like in the IE games, but this isn't a big deal and the same lines can be repeated across NPCs. 4) I don't think any other voicing is necessary. If there is spare money, then villain sidekicks and plot-critical NPCs should be prioitized, obviously.

I do find it a little annoying when a scripted quest breaks the standard game-world rules. Ideally, the reputation/crime resolution system would be robust enough that killing Harond assigns you the appropriate reputation penalty simply because it's known by relevant NPCs that you were involved with him and you went to see him and then he was dead, and that this system is in place for all other situations that share these properties or some of these properties, without need for developer scripting. Sadly, such mechanization is not really ever done and further nobody seems interested in attempting to do it---essentially, RPG developers are not particularly interested in really being RPGs. Another interesting point is that tracking such things by a simple numerical representation of "reputation" is way too simplistic. You should leave that situation with the possible reputations "not suspected", "suspected murderer", "known murderer". This is rather different from the alternative reputation "blackmailed a powerful lord" that another resolution can lead to, and yet will be tracked by the game in essentially the same manner.

I have stopped for a bit, but I will probably return soon. But since I reached Twin Elms my reaction to pretty much every quest has been "meh, I don't have any particular desire to do that".

Hah, it would indeed be great fun if you could sacrifice all the companions to it. Now if only there was a way to sacrifice backer NPCs to it...

I'm glad they didn't do low-int dialogue. At least, not in the ridiculous way it was done in the mentioned games. Stupid people do not talk anything like that.

Supposedly you're having nightmares every night and barely sleeping, but you only find out about this when Eder tells you you're having nightmares every night and barely sleeping, It's pretty odd. Th game doesn't communicate this very well.

I think I'm with the badger here. When I saw a Gul for the first time I didn't think "Aha! A creature that shares many similarities with a ghoul but likely with several key lore and mechanical differences!" but rather "Aha! A ghoul that is too cool for school!"

Major debilitating attacks like dominate and paralysis should probably give lesser related effects on Graze, rather than just reduced duration. So Charm and Dominate could give say Dazed or Confused on graze (still at half duration). Paralysis might give Hobbled or Stuck, that sort of thing.

You don't really need to worry about this in PoEt. I suppose Perception and Resolve are the most common. Might is probably the most fun. But its not a big deal like it was in Planescape. Build your character attributes for combat, and the dialogues will be fun enough regardless.

I wasn't even deliberately trying to do this quest, I just opened his door and he attacked me, so I killed him. Turns out he was a baddie. Well that's fortunate.

Hmm, I'm unconvinced that the main problem is that the AIs tend to stream forwards and merrily just attack you tank. I am unconvinced because this was also broadly the approach of the AI in the IE games (the typical melee AI was "If see nearest enemy, then attack last seen") and this didn't feel at all like PoEt. I'm not yet sure what I think the main problem is. Perhaps it is that, after a few levels, your tank is very good at defending? Like, significantly better than every other character. Or perhaps it is that everything is a bit samey, I'm not sure. I'll need to play more. Maybe it's that there wasn't really such a tank/dps divide in the IE games, except for edge classes like Kensai.

It also tells you what the designers thought was a suitable number of rests per dungeon that you should be aiming for, which is helpful. I'm not super-opposed to allowing people to turn the restriction off, given that I wouldn't do this and I don't hugely care if people make their lives less fun. I would worry that they might then give feedback that the developers listened to without realizing that it was based on this style of play.

Typically what you will try to do in such a situation is build in this sort of thing into your physical simulator, with a parameter or collection of parameters governing the magnitude of changes. Then you can produce a range of forecasts based on what you think these changes might be. In a simple approach you might consider just throwing in a range of plausible parameters and reporting on the resulting range of forecasts; a more elaborate approach would be to state beliefs about what you think the changes are likely to be and then come up with some sort of probabilistic statement or something similar. In practice these sorts of models are too computationally expensive to run extensive simulations of this type (the simulator may have dozens of different parameters and take months to run at a single parameter setting, so you can't explore the parameter space anywhere near adequately), so what you can do is try to build a statistical approximation of the complex simulator's behaviour, called an emulator. The idea is to model the simulator's output using a relatively simple global trend, with a local correlation structure (so that the emulator's prediction at "nearby" parameter choices are "close"). This allows you to use a small number of simulator runs to build a prediction of simulator output everywhere that can be evaluated quickly. Naturally, this introduces additional uncertainty, but it's surprising how well this works in practice. This allows you to choose good parameters by matching to past data, and to make forecasts in scenarios where conditions change. Of course, once you've done this you've still only got a prediction for what the simulator will say at a given choice of parameters. Even if you get the "right" parameters (if that is even physically meaningful) there's still going to be a discrepancy between simulator forecasts and actual behaviour in reality, based on the collection of modelling assumptions, numerical solvers, missing processes and so on. Often this discrepancy is going to be rather large and is almost always extremely poorly understood (and rarely formally considered). There is still a lot of work to do in this area (you will often see "probabilities" being quoted in climate change reports but exactly what these probabilities really represent is rather mysterious) but improvements are being made. I see that current state-of-play as follows; until relatively recently this question wasn't really given serious consideration; now it is given serious consideration by the scientists but it's still somewhat seen as something to give to a statistician after the model has been built and run and have them just "solve" it. Whereas of course what is really needed is for these questions about the link between models and reality to be seriously considered from the first moment the investigation begins right to the end. The resources put into building and running these models currently is far more than resources put into working out what the model output really means, which is not really good. But progress is being made. Probably in my limited experience one of the big reasons scientists don't like these questions so much is that they necessarily involve subjective judgements. There are calculations you can do to assess some of the model discrepancy, but if your model is just missing a physical process and you know it's not there but you can't put it in, then necessarily you just have a make a judgement about how "wrong" this is going to make the model. There are things you can do to estimate bounds on these sorts of things, but ultimately you've got to make a judgement. Scientists often don't like that (and the public often don't like that either) but as long as you're open and honest about it, and you try to show how people with different judgements might reach different/similar conclusions, then you're not doing anything unethical.

One thing that might improve matters in general is if most AI had secondary ranged attacks that they liked to switch to if they couldn't draw a path to any enemy but had LoS on at least one.

Yes, that's fair enough. Indeed, why not restrict leaving Gilded Vale until that point anyway? So they could put Sagani on the road to the stronghold. That'd give you a full 6 member party too, although perhaps they specifically didn't want that so early.

Regardless of whether it's a troll post, it's an interesting point. It's unusual for these sorts of party-based games to give only men at the start of the game. I suppose Pallegina has to be in Defiance Bay, and both Sagani and Grieving Mother need you to be a revealed Watcher when you first meet them, and so they have to be gated behind the Stronghold.

Mental binding is crazy powerful for its level, so something should be done about that. I suspect Ciphers also start battles with a little too much power.

Perfectly congruent with real life #foreveralone I see it now...the game is suggesting that women are all reprehensible gold-diggers that only join you if you've got a Stronghold to flaunt. Disgusting sexist Obsidian. To the twittermobile!

The split is 5-3 male-female. But it is odd that the 3 female ones all take a while to reach.

I thought this whole affair was a tiresome waste of time. But it caused this post to happen, so all is forgiven.