Civilization VI Prelaunch AI: Disappointing or Promising?

By Darren Henderson (DizzyPW)


The Civilization fanatics have been in an uproar since Wednesday’s brazen AI Battle Royal stream put on by Firaxis. Following unimpressive performance in various press demo let’s plays, the community went in with high expectations that Firaxis would not be showcasing their AI a couple days before launch without confidence in its performance. The fact the livestream was not scripted but straight live only fed in as kindling to the fire. So when 475 turns of dull behavior, burning improvements, idle workers, useless settlers, asinine city placement, and nary a pillaged tile being repaired, it’s no surprise quite a few Twitch commenters were ready to cancel their pre-orders on the spot. Now surely the Civ fanboys are rushing to the defense and tilted at reading this introduction, but I say leave your comments till the end. The purpose of this article is to examine what we have seen to inform our readers of what they may be buying into, not to trash the game as a whole as a failure. Sure if you have a group of dedicated friends that only play multiplayer with you, poor AI means nothing. But for the vast majority of my acquaintances that have been playing Civ since the second edition, AI is a make or break point in the enjoyment. Let’s dive in to what we saw, both good and bad, from Wednesday’s livestream.

Before I get into full rant mode, a couple disclaimers are necessary. The demonstration on Wednesday leaves some key elements out in judging the AI as a whole. For starters, the resources, both luxury and strategic, were invisible to us. This was explained away as a key point for the presence of archers and warriors well into the late 1900s, since the civilizations lacked the strategic resources to upgrade these units down their path. Plus these outdated units are said to have minimal to non-existent upkeep costs, making disbanding them a zero gain for the civilizations. With this of note, there’s nothing wrong with the AI keeping useless warriors around. Plus it explains why the Civilization seemed to disband more advanced units once they became obsolete. Overall this doesn’t feel like ideal gameplay mechanics for making the AI challenging, but at least we can forgive Firaxis’ AI design for not being part of the problem.

Another piece of the puzzle we were blind to was the negotiations, partnerships, and war weariness going on behind the scenes. While the overall game was FAR too peaceful for the likes of the average Civ enthusiast, there may be more to blame than simply the AI being passive. One issue came in the form of putting a heavily religious focused civilization on each of the two major continents. Both continents fell in line quite quickly and for the majority of the game to a single religion on each continent, minus a couple rebel civilizations that had other issues going on making them into real non-factors on the map. With shared religions and much land to expand to throughout the game, two of the primary motivators for warfare were removed. And while sure, there was clearly the potential for a continent versus continent world war, with no clear dominating civilization on either continent, such a war would feel inefficient for all sides involved.



The Bad

Russia Rushing Round the World

Now that the disclaimers and potential excuses are out of the way, we need to dive into the behavior that the AI has no excuse for. The game was barely underway when Russia decided its second settler absolutely had to found a city to feed into its tundra supported religious victory. But alas, there was little good tundra nearby its southern continent starting point due to the narrowing of the continent at the tip. The north of course had plenty of room, but the AI would never move a unit over the course of twenty-five plus turns to settle their second city on the other side of two burgeoning world powers… right? Peter of course said nay to logic and reason, and “forward settled” the world’s most powerful militaristic civilization, The Aztecs, with what might as well have been a second capital for a new Russia.

This behavior continued throughout the game as Russia continued to defy logic or sound strategy by sending settlers on voyages that would make Christopher Columbus’ stomach churn. By the end of the game, Russia held roughly seven cities around the world, with only the cultures of two of them actually touching. Considering Russia’s instant culture explosion upon settling cities, this made their overall strategy more about annoy and deny rather than achieving victory. Even the pacifist nations of the AI Battle Royale eventually took two of their cities to punish their insolence.


I should mention that post 2020AD, Brazil decided to found a city in a pure southern snow region with zero yields beyond the three coastal blocks not locked in by ice… but it left such a bad taste in my mouth that I will just move on.



To War… Eventually!

Now again and again I talk about how pacifist the nations were. That isn’t fully accurate. Especially in the early game when war penalties are low and religion unity was yet to be present, the AI and their patron city-states went to war constantly. At one point in the early AD, Rome amassed a mighty ten+ unit legion of their unique unit, legionaries, supported by a spattering of catapults and archers, and declared war on Japan. Japan was truly hurting from their artic start, suffering in science and already battered from the never-ending assault of arctic barbarians. With how much easier cities are to take in Civilization VI versus past versions, Rome should have easily been able to march south, starve out, and end the Japanese civilization in forty-five to sixty turns, and that’s assuming the AI is as inefficient at war as a dog barking with peanut butter filling its mouth.

Alas, such an AI would be preferred. Rome declared war and then let its legionaries sit there, perhaps stuck in a traffic jam and unable to decipher how to move its hoard through the narrow mountain pass separating itself from Japan. After twenty turns of inaction, Rome gave up and let Japan slide. Japan went on to nearly catch-up from its slow start to win the space race, while Rome lost its last chance of relevancy. As time went on, the legionaries became obsolete and were disbanded, never to be replaced by any standing army of note for the rest of the game. Rome itself fell into obscurity, barely limping into the atomic age before the Aztec victory in 2045. This was all the more heart breaking as Rome was on the weaker of the two continents, and easily could have transitioned a win over Japan into continental conquest to eventually win it all.



Let it Burn

Perhaps the saving grace of the AI was their complete pacifism. For as the late game arctic ISIS barbarians eventually proved, the AI in the Wednesday demo had no idea how to micromanage workers to repair pillaged improvements. Any improvement pillaged continued to burn for the remainder of time, completely crippling Spain from a very possible front runner position in religious victory race. Even the mighty Aztecs, who had enough workers to build a human pyramid from their early unique unit captures, watched their recently conquered Russian cities burn instead of using a single worker to maximize their city production. One of the streamers commented ‘Clearly their workers had more important things to address.’ Yes standing around on tax payer dollars is a time honored tradition of Azteca.



Stopping the Aztecs Breaks Immersion

Something controversial but in the right direction was the central design decision in Civilization V to make the AI think more like players and less like historical figures. Unfortunately Civilization V took it to the extreme, resulting in dull figures that lost the magic that Civilization III and especially IV captured so well. Now with VI the pendulum has swung back, and it seems to have swung back entirely too far, abandoning the Civilization V steps in the process. The Aztec victory was in no way a surprise occurrence. It was a very slow cultural victory built up over centuries. Even if this wasn’t the case, the Aztecs were going to win at any time from a science victory. All forces of the world interested in winning should have focused on the Aztecs to buy themselves more time.

In particular, England, who was in position to win the science victory themselves, was positioned as southwestern neighbors of the Aztecs. They had an information age army of tanks creating one of the largest standing armies I saw all game, along with plenty of battleships to bombard the coastline. Meanwhile the Aztec military was on the march chasing down the variety of oddly placed Russian cities, leaving their core space focused and world wonder cities nearly unprotected. The Civilization V AI would have torn Montezuma a new asshole. Instead an incredibly boring whimper of an ending was what the livestream viewers were treated to.

When angrily called out, the representatives simply stated that the war penalty for a surprise attack on Montezuma, combined with past partnerships and dealings with leader agendas, were enough to keep England peaceful and let Montezuma win. This excuse is poorly thought out, boring in practice, and depressing because it seems to hint the devs are happy with the current state.



The Good

While I could and probably should go on bashing the various ills, both perceived and theorized, shown in countless Youtube let’s plays and this pre-launch Battle Royale, I will call it quits here. Instead let’s put on our optimistic hat and look at what went right!



Infrastructure is King!

The cities that were not ravaged and left to burn for eternity were monstrous feats of impressive AI planning. Every workable tile around every city was optimized and improved upon in ways that showed the AI has an appreciable understanding of many of the new features being introduced in Civilization VI. The AI even seemed to plan ahead for district placements so as to not waste builder charges on improvements that they intended to soon replace. Perhaps the most mind blowing talent of the AI when on king difficulty or higher, is that they even recognize eureka bonuses and put clear effort towards pursuing them. This was a clear factor leading into my next compliment.



Tech is Queen!

Perhaps as an inevitable result of immense pacifism, impressive infrastructure, and constantly chasing eureka bonuses, the AI across the world blew threw the tech tree at such an alarming rate that I’m unsure whether to compliment the AI or fault the design team for making such a short tech tree. We saw Greece and Brazil blowing into the Medieval Era not long past 1000BC for crying out loud! The atomic super powers were forming earlier than the real world American Civil War! Modern era fighter jets were circling Aztec cities before the Wright Brothers even got their glider off the ground. If the AI properly prioritized their infrastructure and tech towards a science victory, England easily could have claimed victory fifty+ turns earlier than the Aztec culture win.



Espionage is Joker!

While it is hard to confirm much of the espionage that occurs in an AI only game, everything we saw combined with the compliments from the AI lead developer confirms that the AI knows how to play fuckarounds with each other in the spy V spy game. Stealing great works of art seems to be a glorious game of cat and mouse that should have caused enough international incidents to have broken the eternal cold war. While full details of the espionage system are still unknown to me, I have faith that this is a well addressed part of the AI going into launch.



In God We Trust!

The victory condition I was sincerely shocked to have not been met was the religious victory. The AI like Japan, Spain, and Russia focused on religious wins went hard to the paint from the first quarter. All three rushed their religions and spent immense resources and time focusing on converting every sentient creature that walked their respective continents. At one point I saw at least four Russian Eastern Orthodox missionaries and an inquisitor moving across their continent in formation. Heck we probably saw more religious battles than military battles over the course of the game if you exclude barbarians. On that note…



Barbarians Get It

Depressing as the reality is, the barbarians were the MVPs of waging successful warfare. They captured countless workers, even a couple unescorted settlers, and wrecked chaos across the world from beginning to end. With little support from the pacifist nations, they proceeded to ruin any chances of Japan and Spain from winning. They properly pillaged the landscape and attacked cities directly when viable, keeping nearly every city-state on the map in a constant state of fear. In the very late game when partisan forces began to appear, the Information era barbarians made ISIS look tame as they sieged more cities at one time than any other civilization sieged from start to finish. I sincerely hope the AI team looks at the barbarians and figure out where they went right with their AI.



Closing Thoughts

By the time you’re reading this, people across the Internet are undoubtedly testing the game for themselves, having already dropped the price of entry to Firaxis, and posting a slew of positive and negative feedback on the AI. But I will share my crystal ball based on my experience with nearly every Civilization game and expansion in existence.

The AI team working on Civilization VI seems to have the best tools and best minds of any Civilization title thus far. The fact that they were confident and open enough to show the public the battle royale is proof of that. Unlike some of the more PR oriented people that hosted the stream, the AI lead was candid and honest about the current situation with the AI. He understands what’s wrong and did not try to sugar coat that there is still much work to be done.

That said, the real villain here clearly is not the AI team, but the financial decision makers that choose to push the game live so soon. Clearly the preview reveals building hype over the last month were not timed to align with this launch, but rather timed to showcase what was actually in a completed state as the Civ VI team went into hyperdrive to finish on time. With so many new and altered features altering any AI groundwork laid via Civ V, it’s no wonder the finer points of the launch state of the Civ VI AI is going to be garbage.

The features the AI needed to learn to address likely were only finished a few months ago or maybe less. You can’t test and improve on AI until you have the actual features in-game to see how your current code utilizes them. With features being rushed and finalized constantly, it’s the equivalent of testing a control group versus test group of monkeys in separate cages but with a walkway between the cages!

So crystal ball, how will this all play out? Ah… yes that makes sense. You zealous people that have bought into Civilization VI have just paid full price to be beta testers of this poor AI. Now fear not, you are still heroes even if you have played the fool. Your funds have ensured the full support necessary to create competent AI for what is sure to be the most complex Civilization game ever created. The game breaking issues will likely be resolved by the end of November, and five to seven months from now we will be receiving news of the first expansion. By then, Civilization VI will begin to approach the levels consumers expect of a legendary 4x strategy game franchise.

Addendum: It seems that at launch, a last minute, 4 GB patch went out, and players are already seeing a far more aggressive AI than what we saw in the Battle Royale. Here’s hoping the team listened to community feedback.


Social Media :