Machine Empire screenshot:
Well, Martin has talked about how the tile system is really the last system that needs to be re-done from the ground up because it’s simply holding back too much other design effort.
I actually tend to believe him, not so much out of blind optimism, but rather it doesn’t make much sense that they have spent so little effort on AI (when there is at least one mod that makes the AI decisively superior to vanilla already existing) unless you put it in the context of “it would be a wasted effort to teach the AI to use the tiles better if we are dumping the tiles entirely”.
I find the tile system to be the largest offender when it comes to creating busywork. Early on it’s fine, there isn’t all that much to be doing with your time anyways. It’s kind of a shame because one of EUIV’s greatest strengths is that literally before you unpause the game you have made some pretty major decisions with what to do with your first diplomats.
But if they can actually make the management of planets more interesting while simultaneously making it less tedious, I’ll be impressed. I’d settle for either more interesting or less tedious, it’s getting both right that is tricky.
But seeing as getting rid of the Tile system will almost certainly improve optimization of the late game, and hopefully the AI can be taught to be at least small child-level at playing the game without cheating, I suspect I’ll end up really enjoying this patch.
That is a good summary but I got a couple nitpicks:
-
The impact of the tile system was due to the very basic AI Paradox had was polling every tile on every settled planet very often. The mod you mention, which introduced a straightforward routine that effectively managed the tiles for you, avoided that impact by keeping some extra data that allowed to evaluate in one single operation whether or not to check the tiles in a planet. All this at the scripting level, without access to the engine.
-
We have discussed on this thread in the past that PDX seems to have very little programming muscle in comparison with the complexity of the systems in the game.
-
PDX is hiring at least one c++ programmer.
-
Nothing in PDX track record makes me think we are going to see an AI that doesn’t rely on cheating .
Nothing in the entire industry should make you expect to see AI that doesn’t rely on cheating for a challenge. Not sure why PDX is being singled out there. :)
I only realized this the other day, it looks like Clasusewitz can’t spread processing between ticks, and it didn’t matter too much until Stellaris. Ops.
I do wonder how many people don’t play enough of a game (like me), or don’t want to do work at min-maxing (that’s work… and a lot of game time) for it to matter, except when it completely falls on it’s face. It doesn’t seem to matter for sales.
There’s a bit of variation depending on the genre, but yes, the ROI there isn’t great for strategy games whose rules change very often. That is pretty much Sysyphus’ job.
Because we’re talking about a PDX game. If this was a thread on Total War Empire I would have put CA rather than PDX in the sentence :)
on what data set is this assumption based? You first would need a game like Stellaris which has actually great AI and look how that works out.
I also think that people have often the wrong idea about what “better AI” really means. In most cases the discussion ends up being about difficulty while in reality that’s really not the foundation of 90% of AI complaints.
What really matters (even if people don’t realise it themselves) is to have a competent AI that creates a good experience and doesn’t disrupt your immersion.
Developers often try to play exactly this hand by saying that people don’t really want a “good” AI because then the game would be too hard or that people just don’t care enough but to me that’s really ignorant because they must know how important the AI is for the whole EXPERIENCE of the game.
So one of the major problems of current “AI” is not only the fact that it is often pretty incompetent at handling the various game systems but that it is also a bad opponent, it doesn’t create good stories (gameplay situations) for the player.
There is so much unused potential in this area that I honestly believe that we curently don’t have any idea what we are missing.
I can only hope that the rise of machine learning will have an effect on the games industry in 5-10 years because this seems the only way for things to move forward in this area (compare the advancements in graphics to those in AI in games, it’s quite laughable though it’s kinda ironic that the demand for better game graphics and thus our modern graphic cards lead to the new possibilities in ML which might result in better AI for games some day).
Well, that’s my game AI rant for today but I honestly believe that we just underestimate the impact really good (immersive) AI can have in games like this.
I tried to write in a way that didn’t contradict that, because that’s what I feel as well, but have no real data. One good anecdotal point, though, is how the fairly grognardy 3MA panels have come to praise games while admitting they’re not good strategy games. The subjective justification is that they’re doing new and interesting things instead of the same old resource optimization problem of MoO/Civ2.
When an AI flounders as visibly as in Stellaris or modern Civ, though, immersion hits a brick wall, because it’s too clear you’re just playing the system - very fun for some, and I enjoy it to an extent, but it’s not as satisfying in general. Thankfully, sometimes, it’s just a mod or 10 away.
ML isn’t a silver bullet, there’s too many variables in a constantly shifting game balance, but some generalized insights can and will be coming from those techniques, just not a fully optimized gameplay. It can’t help much with immersion, though, unless you can define a way to measure and model it, which will most likely take a long while before it’s possible.
Not that I’m disagreeing with you when you say that really good (immersive) AI can have a huge impact on game, it will be exciting to see it done.
Paradox games in general tend to just have too many systems interacting for the AI to make good synergistic use of them. Interestingly, out of all their games I feel like EUIV and CK2 tend to tell the best emergent stories. Stellaris and HOI4 seem far easier to create a “good” (read as: difficult for a player to defeat while still not cheating) AI for, because in either of those games I could literally hand-write a checklist of stuff for the AI to do and beat most players that would work in most scenarios with a handy series of IF/ELSE statements.
But as has been pointed out, while that may on some level be “good AI”, as in it could beat people with decent though not fantastic game sense, and it wouldn’t need to cheat to do so, it also does no good on a storytelling level EXCEPT that absence of an actual existential threat in 4X games is itself a tension-killer, and most people’s favorite memories of various 4X will be about moments they thought they actually might lose.
I feel like Stellaris could really be vastly improved by a number of non war-related decisions interacting with the AI. Right now if you want to interact with the AI you can trade small amounts of resources with it, you can declare war with it, or you can get tied down in some Federation that leaves much to be desired. That’s about it. With a bunch ways to influence galactic events that aren’t solely “here are my war goals, enjoy”, I feel like we’d get more interesting stories. While a pack like Distant Stars was kind of nice for those of us who have seen every event a hundred times, and I actually think the game could use for more staggered pacing of various Crisis, just adding a bunch of new anomalies doesn’t add much in the terms of storytelling beyond that first few handful of games. Then it’s just bonuses and penalties.
Thanks for posting that! I’m even more encouraged that Martin said that the point of pop happiness will not be boosting resource generation but new mechanics. Here’s to keeping your pops happy so they don’t undermine you or put your head on a spike…
Sounding better and better with each teaser.
Thanks KevinC! My thought exactly. Not only do they get rid of tile busywork, but you allow planets with high populations, distinctions between different strata of society and a distinction between resource-gathering worlds and industrial centers.
“Planets with high infrastructure are more efficient at research and other specialized jobs. Many sparsely populated worlds would yield a lot of raw resources but likely make you reliant on the market to import manufactured resources.” ~Wiz
Oh and also a market and what sounds like a real trade system.
This indeed looks very promising. Hope it exceeds expectations.
Any speculation on when we might see these changes? End of Q3 for them to scoop some DLC revenue before a quarterly report?
My heart says next week. My brain says next year.
We haven’t even begun the designer diary series of the upcoming patch. We’re talking months.
My speculation is that Dev Diaries start up in the beginning of August and they shoot for a release by end of year (Late November through December). If there is slippage, I’d expect it early 2019.
The last DLC was a small one and released at the end of May. Given that and the screenshots that Martin is showing, it looks like they’ve already gotten a good start on the expansion.
It seems like this rework would touch on so many aspects of the game, balance issues, etc. Plus whatever is in the actual expansion. I’d love for it to be November/December but I’d worry that a lot of things would be out of whack. Maybe that’s inevitable.
Actually meaningful planet specials would be nice.