Matrix Games Forums

check out the AI forces on turn 6!

hey Jdane,

Yes, I am a brand new dad, Thanks very much for the congrats! Definitely have needed to maximize my time management skills!

I agree with much of what you said, and I like the multiple save game file idea. I will implement that!

I actually have only reached turn 10 and I am doing fairly well (on easy). I have most resources flowing in and have started attacking a major and conquered my first city (through a mountain range, which was tricky for me!).

I was just nervous when I saw some posts saying that they felt like they were playing pretty good and then got overwhelmed later in the game playing on easy. I tend to want close games, where player mistakes are forgivable on easy, even if I end up ultimately losing! That's why I would push for Vic to slowing implement an option for players to chose where the AI can use more human rules. That was my major point!

Distant Worlds Universe is the perfect example of massive pre-game customizable/options in an extremely complex game. It allowed me to set up easy galaxies with maximal computer help for me until I got comfortable with the game. I don't think having those options detracted from coding the game! It just made it better for all players.

Thanks again for the nice feedback and I will definitely use some of your ideas to help me get a leg up!

-Matrix forums!!![:@]-

@Mook881 : I'm glad you seem to have received my message in the spirit I intended, and wish you only the very best, both in your gaming and family life.

@YohanTM2 : There are huge gaps in their lines. You could encircle and destroy them one by one.
But those numbers are big, I would agree. Are they slavers? Which difficulty level are you playing at?

ORIGINAL: Mook881

Does that make sense about options and learning curve on easy?

what are other players thoughts on this and why the resistance to more starting options for games (as long as it doesn't take too much time away from improving the game in other ways)?

I would not resist more starting options at all. I would not resist a 'tutorial' setting where major AIs do not generate for example to allow a player to hone skill against minors.

I resisted the particular one you mentioned because it would be massively detrimental to the game and Vic's sanity (he's not some robot here for our entertainment and we can't expect him to break himself![:'(]) in my opinion.

I am all for any options that make things easier for new players to learn. I would rather that be the player opting to universally simplify starts, wider game scenarios etc. not divergent rules (obs new standalone rules could be toggleable if anyone worried about air power etc is reading). I would be for a setting that let a player start with a free (as in upkeep too) L1 or L2 truck station. This would neutralise early AI logistics advantage without changing the overall game systems and make it clear to the player exactly what bonus they are getting. When they choose to play without it, they then know exactly what is missing and how to adjust their game, though knowing is not the same as implementing.

In fact I think ideas like this are really healthy for the game, the wider genre and the players as well as Vic and Slitherine (in terms of wider appeal). I would fully support lots more starting options to tailor difficulty so long as they mostly work within existing systems and did not demand two radically different AIs to implement!

Worth noting you, of course, get a few advantages on easy, like better starting units (much better in fact; you start with a unit of makeshift tanks I think) etc. The AI difficulty bonuses (things like the rate they get their militia units and their ruthlessness in declaring war) are turned to minimum. The AI only gets the rules simplifications it needs to play the game at all well.

Side note and another tangent: I wonder if some beginner settings paradoxically make the game harder. Like are AIs less aggressive or less aggressive to the player. My big game against majors has relied on exploiting the AI overextending itself against other AI so far. If AIs do not overextend on beginner because they are universally less aggressive, then they may actually be much harder when players bring the fight to them.

Final note: Vic is looking at unifying AI and player rulesets in some areas. If this happens, it will probably occur equally everywhere, though I guess high difficulties could still give a truck station on top. The problem is ensuring the standard AI, regular, can actually keep up with players the setting is intended for using full rules. You don't want everyone on hard plus because the AI cannot play on normal anymore. You then just end up with difficulty inflation.

Also the AI will probably never use full rules. Some of the politics and story stuff will never be enabled for them to save CPU time as much as anything. Modern CPUs are good but they are not that good.

Final final note: players saying they are good does not mean they are good. Eg. Tortuga (I love you man BTW no homo and keep making awesome entertaining vids for me to nitpick) is not a very good player. He is largely competent and would be better playing off camera with more time. Tortuga has not fully groked some of my detailed explanations of his supply situation in his current series. Yet the vids literally show that he is not pushing enough supply due to a logistical error and this is partially the reason his troops have intermittent supply penalties. Hopefully, he has taken my word that the troops can probably still fight and are probably winging about empty stockpiles not empty bellies. His food production was an issue not the issue but my explanation of this swooshed. That's on me TBH.

It would be easy to look at someone like Tortuga as a very skilled player because he has a lot of hours. Hours do not translate to skill necessarily.

Post if you have a problem and listen to the consensus opinion. Implement it, even if you do not understand it, and you will get better with time.

ORIGINAL: KingHalford

Chess and other very simple board games aside, I'll challenge anybody here to name one computer strategy game where the AI is both competent AND plays by exactly the same rules the player does.

I fully agree with the point you are making, but for the latest developments in gaming AI you may want to check out Alpha Star, which is the DeepMind AI learning how to play StarCraft 2. Same AI team that made AlphaGo. They actually had to limit the Actions Per Minute of the AI player just to make it more human. Also, it plays by the exact same rules in every other regard. It can only see and manipulate what's on its screen, as opposed to issuing thousands of simultaneous orders all over the map.

Just to be clear, this is in no way representative of what your average game developer can do for AI in his own game. But the future of AI in gaming is really an intriguing concept.

MC456 see the poll thread for some interesting discussion of this. If you have any more detailed thoughts chip in over there.

ORIGINAL: MC456

ORIGINAL: KingHalford

Chess and other very simple board games aside, I'll challenge anybody here to name one computer strategy game where the AI is both competent AND plays by exactly the same rules the player does.

I fully agree with the point you are making, but for the latest developments in gaming AI you may want to check out Alpha Star, which is the DeepMind AI learning how to play StarCraft 2. Same AI team that made AlphaGo. They actually had to limit the Actions Per Minute of the AI player just to make it more human. Also, it plays by the exact same rules in every other regard. It can only see and manipulate what's on its screen, as opposed to issuing thousands of simultaneous orders all over the map.

Just to be clear, this is in no way representative of what your average game developer can do for AI in his own game. But the future of AI in gaming is really an intriguing concept.

I'm sure that will go over like a lead balloon. Deepmind AI use some cheesy tactics in Starcraft2 which beats 99% of players. I can see the complaints now how the AI is unbeatable. Playing Chess with the AI feels like playing a calculator with it's powerful search engine.

hey jdane and laiders,

Thanks again for the productive conversations!

to Laiders: I think we are actually basically on the same page and I love your suggestions.

Good luck to Vic and I can't wait to see this game evolve over the coming years!

ORIGINAL: Smidlee
I'm sure that will go over like a lead balloon. Deepmind AI use some cheesy tactics in Starcraft2 which beats 99% of players. I can see the complaints now how the AI is unbeatable. Playing Chess with the AI feels like playing a calculator with it's powerful search engine.

All is fair in love and war lol
I think a diversity in playstyle is important, but I can't blame an AI designed to win for using "cheese" tactics.

(l.e. jesus those were some large pictures, resized)

Here's the issues I have with the way the AI "deals" with logistic. With pictures for ease of explanation. Sorry about the mediocre paint drawings and the oversimplification. The AI can build free roads, and is always in supply if his unitss have access to a road that connects to a city, regardless of how many Logistic points the AI actually has or how far the truck stations are.

Let's pretend there are no resources in the area.

Can you identify which side is the player and which side is the AI?

That's right, the AI is obviously the one without any logistical buildings outside the city.

But why does that matter?
Because at the moment what the player needs to do to quickly defeat the AI (in this situation) is a double envelopment, to trap the enemy units, cut them from supply and defeat them easily.


Except he needs to do it with units all along the front.


The players can't leave an opening, and can't just cut the road, if for some reason, mountains in our example (the barely visible yellow triangles are the mountains, they looked better on the stupidly big size) he can't attack on both flanks the encirclement fails.


Why? Because if he doesn't, if he just cuts the road, the AI will immediately build a free road to keep his units in supply (technically his units should still be out of supply for 1 turn)


So what you say? The player can do that do, which is true, but he can't do it for free, meaning he can't do it forever like the AI can.



Which becomes even more of an issue when you deal with terrain that makes roads very expensive, like our mountains.


Silly isn't it?

But what about endless logistical capacity?
Well, let's go back to our original double envelopment, it's a success, the enemy forces are starved of supplies and easily killed, now all the player has to do is seize the enemy city... except he can't.
His supply lines don't go that far. Which is fine, is as it should be.


He'll just have to stop and build up some supply extensions to continue our attack.


Which again, it's fine, it's as it should... or is it? Because the AI doesn't have to do ANY of that.
All the AI has to do is get past the front line and seize the supply extension assets.


That will cripple the player in a much faster way with far less forces that the Player would need to cripple the AI.


Afterwards the AI can just seize the Player City much faster than the Player could, he even gets the bonuses from player's supply assets, which he didn't need anyway.



So the AI doesn't follow the same core rules as the player, what's the big deal?
Well as I said in previous posts, it makes the game less fun and more importantly less strategical. The Player CANNOT SEIZE THE AI's supply assets, to cripple him or help his own attack because there aren't any. Meanwhile the player must have a WW1 style entirely manned front or have troops defending key positions behind the line.

You either do a full encirclement of enemy forces, harder done when the AI will have more units and redeploy them faster, or the war ends up in a siege of the enemy city and all of the troops that got away.

Wouldn't it be fun if the AI had some logistic assets we as the player could aim a risky attack against? Wouldn't it be less tedious to seize the AI's supply assets after we've wiped out his forces to keep our assault going? Wouldn't it be better if you had to worry about reaching such an assets fast enough before the AI destroyed it to deny it to you?

Some of you will say "multipalyer is the answer to that" to which I say "How many of you have the time, discipline and patience to do a play by email game?" (because I don't and I envy you) and "You'll still have to deal with AI ignoring the rules"

Did I mentioned the AI could also do this if you let him in about 1/3 the time the player could do it (since the player would have to build supply assets)




P.S. Personally I think even dirt roads built too fast as to make seizing them not that big a deal unlike it was historically. If they took longer to build capturing or cutting them would be very important, but that's a discussion for another time.

That's just my opinion, but that's the best case in favor of having the AI play more along the same rules as the player I've seen so far.
I can only agree it would be great to have logistics assets as targets of operation.
(The presentation does a good job at conveying the point.)

ORIGINAL: Jdane

That's just my opinion, but that's the best case in favor of having the AI play more along the same rules as the player I've seen so far.
I can only agree it would be great to have logistics assets as targets of operation.
(The presentation does a good job at conveying the point.)

Well said.

I like the LIS system for the player, but the AI needs to be debuffed from roaming about and building unlimited free roads at will.


@Palora- pictures are great!

To be fair, other people talked about this very issue, but the pictures make it all the more striking.

What strategy game have you played that perform such things. Even RTS until recent times had awful pathfinding. Go back and play the original Command and Conquer to see just how bad the AI pathfinding was. While there has been some improvement it still far from anything Palora posted.

(P.S AI war sometimes tried to attack from a different front which is the best I've seen from any strategy game but of course the AI doesn't play the same rules as the player as that the main part of the game.)

I'm not saying it would be a trifle to implement, mind you, but simply that it's a well presented suggestion for improvement.
What to do with this is solely VR Designs's decision.

Of course every change can just as easily make the AI worst than better. The difference between the player from the AI is the cost of the roads. As he said a player can do the same only it cost resources. AI gets extra resources to make it challenging is a norm for strategy games.
His statement "it makes the game less fun and more importantly less strategical." even applies to Chess AI. Chess AI beats me by the brute force of it's search engine and not do to it's strategy. It what makes playing Chess with the AI more boring than playing another human. I'm basically playing Chess with a calculator.

Compare the AI roads to yours as you can clearly see the difference.

When the ai algorhytm that will inevitably be created from this chain of arguments is finished, we're all either doomed or forced to live in a perfectly run society - which for humanity nowadays = doomed.

But seriously, love the debate. The OP apparently isn't aware that game "ai" has been "cheating" since the first Space Invaders game...

If you're interested in an AI that doesn't cheat, the Galactic Civilization games make a particular selling point about how at baseline difficulties their AI doesn't cheat and is arguably the best strategy game AI in the industry. On lower difficulties the AI will occasionally call you up and say "Hey so I can SEE you massing fleets on my borders, I know what you're doing but I can't do anything about it because my brain has been handicapped at a lower level, FYI."

The thing is, the game achieves this AI performance by carefully designing the game from the ground up to be easily understood by the AI. Techs are basically all linear boosts in stats, buildings work much the same way, combat works pretty much the same from beginning to end, just with higher stats, there's no "revolutionary" techs that completely change how the game works, etc. etc. It's basically the gaming equivalent of pre-chewing food to feed the pap to a toothless old man, and the result is that gameplay tends to be pretty bland and uninteresting. Nothing like the complex web of choices and varieties in strategy present in Shadow Empire.

The thing is, the game achieves this AI performance by carefully designing the game from the ground up to be easily understood by the AI.

Exactly. As you said, the fact that SE was not designed this way is also what makes it amazing. Galactic Civ is also not that amazing in my experience. It plays well, but getting a start in a corner behind an aggressive AI is a much, much worse handicap than putting the AI on the highest level without that unlucky start.

So far, I've found the SE AI to be a lot more fun because it is consistently challenging and (except for mountains that can indeed suck) I have yet to feel like I was put into an unwinnable situation.