1812

I’m actually not that impressed with the AI. I mean, it works. But it’s not pretty. :(

It’s also shown that the AI will still fall apart at a certain point. I’ve gone from being really happy with the AI in Old World to a growing sense of dismay as I play into late games with most of the map unfogged by agents. To Mohawk’s credit, they’re showing you everything. Which is how I know the AI struggles with how to use its units and where to move them. And, as I said, it’s not pretty.

There’s a ton of suboptimal moves going on. Units marching back and forth for no reason, units loitering unused near the front, while orders are squandered bringing units from farther away. Missed opportunities for taking advantage of terrain, and abilities, and so forth. Also, it seems like the AI flat-out doesn’t know how to use ranged siege units. I don’t know if it’s the unlimbering or the extreme range, but I’ve see swarms of onagers and mangonels go to waste when they would have decisively ended a war. Conversely, I’ve seen the AI march lone onagers directly up to cities to pillage resources. And don’t even get me started on how it has no idea what to do against heavily defended chokepoints.

Hopefully these are issues Mohawk is working on, because right now what makes the AI competitive is its considerable resource advantage, which lets it produce a ton of units that it can’t use very well. In terms of the big picture, as long as it has more units than you, it can wear you down. But down in the weeds, at the level of individual units on individual hexes, it’s basically Civ V with more transparency, and – more importantly – with more thought given to how the design and the AI interact (as you guys are discussing upthread).

I mean, it’s far better than Civ V, but it still seems like a deeply flawed AI that can’t leverage its military advantages very well.

-Tom

1813

I don’t think there are many easy improvements to be made to the AI - it has to be able to handle a very wide variety of situations well enough, which sometimes comes at the cost of specific situations being handled worse - but it is hugely useful if you post saves where you run across specific identifiable dumb AI actions, such as moving an Onager right up to a city, or placing infantry in exposed terrain right next to forested terrain.

(Basically, we can look at things from the AI perspective and figure out why it takes a certain action, but that depends on having saves where’s there’s a good action to look at)

One thing you mention I want to point out is that units marching back and forth may well have a reason. Since the AI doesn’t cheat with visibility, shifting units in border areas is useful to reveal if anyone is approaching.

1814

Oh, I get that the AI is working with some of the same limitations as a human player. I’ve spent enough years watching AIs in strategy games to know that there are a lot of unknowns at work, so I shouldn’t make too many assumptions about what I’m seeing. For instance, I know the AI will harvest resources regularly, so I’ve seen it running loops between resource nodes. It looks pointless, but it’s not once you realize that it’s getting a benefit from what it’s doing. I’ve also seen it do things that look nonsensical until you realize that it’s probably dealing with fog of war that you can’t see.

But that’s not what I’m talking about. I’m talking about the AI’s inability to win encounters that a human player would easily win. That, to me, is a fundamental failure of the AI. When I get home later today, I’ll dig up a saved game of the onager situation I was talking about and post it here.

-Tom

1815

I’ve also seen the AI position hillfighters in the hills, rangers in the forest and maximize other bonuses to annihilate troops. That’s a lot more than a lot of strategy games. It’s happened too often and too thoroughly to be random I would think, but it’s far from perfect. Of course, there’s plenty of times I’m not taking advantage of terrain bonuses either (due to other circumstances, not out of laziness) so who am I to judge? ;)

Are resource bonuses spelled out anywhere? I checked the manual and didn’t see it there, and unfortunately I don’t my PC in front of me. I don’t recall seeing bonuses listed on the difficulty tooltips, but I do know they crank out an ocean of units.

1816

There are no resource bonuses for the AI. It’s all the starting development - the AI has several cities to begin with, some improvements, and of course a bunch of units, so their army is bigger than yours for a while. But there is no discount for anything, the AI is building units at full cost.

1817

Thanks, Solver. I recall you saying that previously, so I wanted to double check. I know about the Development setting giving them starting cities but thought I might have misremembered your prior posts and there was a resource bonus too, as is typically the case in the genre. I do know that when I’ve had agents watching the cities they didn’t seem to be building them any faster than I could, but damn can they really carpet the map with them.

I wonder if I still have a 2v2 save a friend and I were playing (I take it you guys fixed the MP issues by the way, haven’t had problems finding his game the past few times!). I had a wall of swordsmen and Mangonels keeping a tsunami of enemy units at bay as I moved towards a point victory. Those Mangonels are why I might not play Egypt so much anymore. While I love the Landowner family, that game really made me value Artisans and their bonuses to siege even more…

1818

That gave me an idea. @Solver: could we have some sort of overlay (on a held button perhaps) to indicate where the current unit gets defensive / offensive bonuses? The existing one to point out good structures to idle on is really helpful.

1819

I’ve noticed that since I’m playing Rome as conqueror instead of my normal builders as Egypt, or Babylon, my opinion of the tactical AI has decreased also. I’ve seen more shuffling of units than I think makes sense. Some of it’s clearly good, pull those almost dead units out of the line and go heal. But some of its seemly pointless sail an Axe unit from north city to south city, followed by sail a different axe unit from south city to north city. Now while I suppose, they are positioning them in different hexes. It is not clear that a 25% max bonus they get is worth the 4 orders or so they are using. Especially when they have a large number of unused army units near the battle.

I have yet to be attacked by a siege unit also.

The AI still provides a challenging game. In fact, I’d recommend those looking to play at higher levels make a couple of changes to the default conditions. (Map setting/Advanced settings)
First, turn center bias off. This means instead of starting in the center, the player will start on the edge. This enables you to grab 2-4 cities from empty city sites and barbarians, without having to attack a minor or major countries.
The second change is to turn play to win ON. This gives increasingly negative foreign relation malus, (I saw a -180 one game) . This basically means you’ll be fighting a 2 or 3 front war as you get close to victory.

I like having a steadily increasing difficulty.YMMV.

1820

The AI gets a head start with multiple cities. The AI in Old World is kind of like the AI in a racing game where you start in last place and you have a certain number of laps to get up to first place. I think even at the default difficulty, the AI starts with two developed cities. As you push the difficulty level, it starts with even more of them.

-Tom

1821

There honestly aren’t that many. Basically, woods and urban give you a defensive bonus from ranged, and urban gives you a defensive bonus from melee. Otherwise, melee attacks across rivers take a 50% hit. Other than that, it’s just a matter of Improvements and city centers. Am I missing anything?

-Tom

1822

Not necessarily. It just means an endgame penalty creeps into other nation’s opinion, which you can offset just as easily as any other penalty. I think it starts at -40 when you hit your sixth ambition and caps at -160 once you’re about to finish your 10th ambition. Honestly, it’s pretty underwhelming for how easily you can manage it. Granted, it takes resources to manage like anything else in the game, but it doesn’t substantially change the difficulty so much as force you to keep pouring resources into diplomacy if you want to avoid wars.

Really, it’s a weird setting, and it strikes me as odd that it’s turned off by default. But I trust that Mohawk knows what their player base wants. Soren has talked before about how people play videogames to win, and the Play To Win toggle is obviously Mohawk’s reaction to that phenomenon. Basically, they’re listening to players who complain about being penalized for winning by having the AI push back against their victory. :)

-Tom

1823

Having better explanation of combat bonuses, or better combat preview, is on my wishlist, but there isn’t really much to show for “general bonuses”. Aside from Infantry in Urban, there are no bonuses that always apply, so it depends on the enemy. Forests help vs ranged, rivers help vs. melee, that sort of stuff, but those are all conditional on who the attacker is.

1824

Tom, the opinion malus is just part of what Play to Win does. You can offset it, but Play to Win also massively boosts the previously discussed AI war chance once you’re close to victory, by as much as 50% towards the end, and it revokes restrictions like the AI not declaring war at Friendly, or when too far (except through events).

Basically, when you’re at 8-9 ambitions completed with Play to Win, every AI should be likely to attack you.

The history of such options in general is interesting, it goes back all the way to Civ4: BtS, where the previously existing Aggressive AI option became something more like “Ruthless AI”, and I myself added some AI behaviors for that option only, which some players would consider frustrating or “cheap”.

It’s the discussion of Good AI vs Fun AI that’s been ongoing since forever. Soren is more on the Fun AI side, he believes the AI is there to provide a good experience, which isn’t always the same as the AI playing optimally. I lean more towards the Good AI school of thought, in my mind the AI is most fun when it’s trying to win itself, not just along for the ride, and my favorite playthroughs are always ones where it’s a neck-in-neck race between me and the AI.

You have to strike a balance when designing an AI, but some kind of ruthless toggle is a great way to better satisfy a certain segment of players, without much development effort.

1825

It can be managed, but diplomacy cost resources and particularly orders. I was getting close to a double victory (about my 6th ambition) and It addition to all the other penalties for closeness, wrong religion. I was -300 which pretty much is automatic war, and really hard to get a truce. I think I’m playing at harder level than you and that matters also.

It is clearly situational and despite fighting two wars, it ultimately didn’t matter and I did get my double victory, but I had to work a bit harder to defend against a war. I couldn’t just hit next turn.

You are correct lots of people hated the gang up on the winner. So giving the players what the want is good idea.

The problem, that I’m trying to solve, and listening to Soren, I think he is in a similar space, is how do we make the end game more interesting for players?

The basic problem with giving AI a big head start is that game is basically over in the first dozen turns.
I spent a long day beating my head against the AI at Magnificent level. Turn 1 assign my leader (for extra speed) to my slinger/warrior. turn 2 attack a barb city, turn 3 AI grabs the city. Race my leader over to another barb city, turn 5-6 another nation takes it. Settle my free city site on turn 8, build a second settler. If I’m lucky maybe grab a barb city. I end up with 2 or 3 cities and no place to expand.

To me, that’s more frustrating than fun. After a dozen restarts, I finally got 4 cities, with a minor tribe in my rear that I could grab another 3 cities. At which point, I turtled used my huge treasury to buy all the resources to build all the wonders as soon as my cities hit legendary status and won that way. Also not fun. Cause it literally took more than a dozen starts before i got lucky.

Exactly this.
To borrow a certain math description. Strollen’s parabola is less steep and much longer than most 4x games provide. In an ideal world, yes the AI would be smarter, and play the game better and Brad Pitt would play me in SNL.

I still haven’t had a competitive game pass turn 100, but I think Play to Win will help me get there.

1826

Ah, that would have been good to know. That’s part of why I was asking about war declarations in the other thread. I wanted to know what sort of under-the-hood calculations might be going on. If the Play to Win toggle changes the rules and math for war declarations, that’s kind of important to know. Because all I can see are the listed opinion penalties.

Also, to my mind, there’s no difference between Fun AI and Good AI. I understand the debate, but I think phrasing it that way obscures the issue. If you’re going to make an AI an adjustable part of the difficulty level, that’s great. I fully get that adjustable difficulty is part of the calculus of “fun”. But the first priority should be an AI that can play the game you’ve designed. No one should get a pass for that because they claim “we want our AI to be fun!”. For the most part, I’m pleased with Old World’s AI, because it’s working so much better than other games in the genre. But I’m also disappointed to discover how often the Old World AI fails to play the game you guys have designed.

You asked before, so here’s a late-game save from my very first game of Old World, which I shouldn’t have won, because I was still figuring out how the dang thing worked. But it was at the default difficulty level, and even with Play to Win enabled, there wasn’t really much pushback once I’d gotten going. In fact, the last thirty or so turns were just me hitting End Turn, End Turn, End Turn, etc. But before the game ended with my 10th ambition, I put everything on hold to see how the AI fought wars.

So this is a saved game with all my workers sleeped and all my cities just running Council projects. Download the linked save and drop it into “Users/YOURNAME/Documents/My_Games/OldWorld/Saves” and then load “Babylonian onagers” from the main menu. You can immediately hit End Turn and watch things unfold. You can see that Babylon actually has three onager/mangonels unlimbered near the battle (it took them freakin’ forever to get them there). But it’s got tons more onagers and mangonels just wandering around, including about six of them less than one turn away, shuffling aimlessly. I should have lost that battle a long time ago. I definitely would have against a human player.

In this saved game, if you don’t give any orders or bring in any new units, Babylon will eventually overwhelm my defenses. But otherwise, there is no danger it’s going to take my city, because it’s not managing its units very well. It looks to me like it simply doesn’t know how to use ranged siege weapons, which effectively means ranged siege weapons give human players a huge advantage over the AI. To me, that’s neither Fun AI or Good AI.

-Tom

1827

Absolutely. People playing at different difficulty levels might as well be playing different games when it comes to discussion about balance, tuning, and so forth. And for the record, I’ve played at multiple difficulties. Currently, I’ve just dropped back to Noble from Glorious. I was having too much difficulty getting any traction against an AI that starts with four developed cities. :(

I think most strategy games are pretty good at adjusting to players of different difficulty levels. The heart of the problem is how to introduce new players to the experience – Old World’s failures on this front have already been discussed – not how to keep the game engaging for experienced players. And I can tell you that the answer to this issue is never “ineffectual AI”.

-Tom

1828

I think I’ve seen that as well, but I’ve also learned not to make assumptions. :) Can we have @SorenJohnson, @Solver, or @Velociryx confirm that the AI does know about and use the terrain bonuses units get with traits?

-Tom

1829

It’s the middle of the night of where I am so I’m logging off, but the save is fascinating. I haven’t touched the AI really, it’s Alex’s handiwork, but that’s another thing I want to get more familiar with, and I’m definitely seeing some things in the save that should be fixed - I’ll investigate more and summarize for Alex.

The AI isn’t quite that bad, if I end the turn then Babylon immediately kills 4 of your units, including 2 by routing, but of course it isn’t fighting optimally, and the stuff siege weapons are doing is weird. One Mangnonel decided to go help defend another city on the other end of the map. If I were Babylon, I’d be wiping out 6 or 7 units on my turn, leaving your two cities in the area with no hope of defending on the next turn.

1830

Yep, that’s exactly my point. I should not be winning that encounter. :(

I do appreciate you taking a look, and if it’s of any help to anyone at Mohawk, that’s great. But it’s nothing that I haven’t seen in my other games, so I suspect it’s nothing they don’t already know. And – hopefully – are already working on.

-Tom

1831

I was meaning for units with promotions. The common bonuses are easy to remember, but it’s more difficult (for me at least) to notice that the current unit has Highlander or Fierce or something and so is better on hills / against melee than it would normally be. You can hover over various target hexes and see a damage preview for a hint, but it doesn’t cover everything.

So I was thinking a highlight on hill hexes for a highlander, on the enemy melee units for a fierce unit.