Jon Shafer's At The Gates

I know Digital Legends uses GOAP for their work (public info since they did a presentation on that). There are others. In general it’s a good tool for MP bots (limited set of actions, easy world state definition, clear goal most of the time…).

I certainly don’t expect any game engine to come with it by default, though. It is harder to implement and do the initial setup for than for a behavior tree before you can start defining actions (world state definition is a non-trivial problem). It’s not a generic tool.

GOAP is a tree search when deciding on the course of action, but on debugging/executing the chains of action in a real time game it is a state machine (you execute one action/state until you get the go-ahead to advance to the next state or reevaluate the search). For designing (deciding actions to insert to facilitate certain behaviors you want) it helps me visualizing it as a complex state machine to which I’m adding states but have no real control over the connections. You can also visualize it as the search creating a state machine for the game to run until it’s no longer viable. But I’m not an expert, so I certainly am looking at it from a weird perspective, it’s just a tool I use, not something I develop.

By definition, GOAP can’t decide on the ultimate goal of the search. In chess or in a shooter that goal is easy to decide previously and inject (win, kill the player, get a board state evaluation over a certain level…) but in a more complex strategy game, or in simulations when the agent does not need to “win” at all, just be part of a world, the decision of which goal/world state to go for in any given moment is less clear and requires of a higher level evaluation algorithm (which granted, could be done with a more abstracted GOAP set with a goal of “winning”, but it’s probably overkill).

That’s very cool about modelling probable external change into the search. I am not aware of any game that takes it to that extent.

Years ago, when Atomic Games was doing wargames, I did some testing for them. Some of it consisted of playing the game (the one I remember most was I think Crusader) over and over, against the AI in its current state, and recording my moves so that the programmers could get a bunch of data about, well, playing the game, decisions, strategies, whatever. Now, I’m a terrible player, so I’m not sure how much this helped, but the idea was kind of cool–see how humans approach it and figure out a way to make the machine’s actions at least approximate the sort of thing a person does.

Of course, if there are horrible defects in the AI for that game, it’s probably because it’s emulating my terrible generalship…

Are you the reasons tanks can’t drive straight in any Close Combat game??? ;)

Fwiw, I have 2139 hrs in AoW3 according to Steam (in reality there a couple of hundred more that never got registered), was a beta tester, and played the preceding games to death, and sometimes dream of a resurgent Azrac Empire.

I also organised and hosted 2 tournaments, and played a shit tonne of multiplayer, and exposed and corrected (not just me to be fair) loads of exploits before the game got released.

And sometimes I still lose at Emperor.

There are 2 “tricks” to winning at Emperor:

*scout aggressively so you know when the enemy is coming, and then attack them before they gang up.

  • Abuse your superiority in tactical combat.

You can also essentially abandon your cities and just loot the AI and move onto the next city.

Review up at the4xplorer (this is the guy that has been streaming it and recently interviewed @Jon_Shafer iirc)

Gah!

" And how about the fact that the AI is basically non-existent…

Yeah, let’s talk about that.

The AI: I understand that the basic structure of this game is meant to mostly be about surviving the harsh wilderness and learning how to balance progression with simply being able to exist, but the AI for both the other Germanic tribes and the Roman Empire is basically brain-dead.

The other tribes do not expand. They do not compete for the same resources that you do. They simply exist for you to eventually declare war on (because allying them currently is literally impossible) and overwhelm with superior armies."

Yeah, the good sounds really good, and the bad sounds suitably bad.

However, while I’m going to buy this today to support @Jon_Shafer I’m going to hold off and play it after it’s had some time to cook, @KristiGaines would be proud. :)

Plus, there are tons of other things that need my attention, not just gaming (Slay the Spire just hit 1.0 today, Sythetik is amazing, and Pillars of Eternity 2 gets a freaking turn-based mode tomorrow) but also getting ready for a Gloomhaven/MtG weekend by way of picking up the house and preparing for company. Also Punisher Season 2. So much going on, it’s easy for me to hold off on this one for a bit.

Yeah, this was my experience as well.

Man this game has been in development for so long that
I signed up for the kickstarter using my old spam-filled yahoo email. I just check the ATG updates are there but no key. Anybody know or when we get our keys?

It doesn’t look like the game goes live on Steam for another hour, so my guess is just before or around then, since they probably wouldn’t want to give the keys out and have them not work yet, but that’s just speculation on my part.

My review, FWIW. I concur generally with what’s been said. I’ve definitely enjoyed myself, but the lack of content in certain areas are noticeable. I think the only thing I was truly ‘disappointed’ in was the lack of an aggressive faction AI. That, and the late game needs more stuff to draw your attention and resources (considering you’re likely to be swimming in everything once you’ve found a good spot).

That’s not to say you won’t be fighting though - those bandits WILL wipe the floor with you if you don’t pay attention to them and where their camps are. They feel a lot more “in your face” than they do in other 4X games

The keys are there at humblebundle.com now, but the game is still locked.

Weird, I don’t have At the Gates in my humblebundle keys. I have no memory of ever choosing humblebundle as my kickstarter choice of keys either.

I dont remember having a choice, but I might have bought it through his website afterwards. Do you see it listed under purchases?

Thanks, got the key but the game is still not released.

I can justify buying this today if I had reason to believe that JS will eventually sort out the AI. But…it’s been seven years in development and to release a single player game with a non functioning AI seems so odd. I want to be supportive of this studio but should I hold out hope that a fundamental element of the game will be patched in at some date soon?

I have complete sympathy for Jon. At the same time, it appears to me like publishing his ordeal (something I’m very glad he did) was in some way an apology for a game that isn’t where he wanted it to be.

The implementation of everything missing will now depend on sales figures, and given how niche this game is, you can probably assume it won’t happen. Especially considering that the missing pieces are AI and diplomacy, 2 very tough nuts to crack.

He spoke candidly about adding in those very features and more post-release. I don’t think he will let low sales stop him, because fixing these issues would net him additional sales.

Unless he has a serious cash cushion, I don’t see how that’s realistic. These aren’t minor tweaks.

Looks like it might be live now.