Flash of Steel header image 2

The Napoleon Complex

February 11th, 2007 by Troy Goodfellow · 7 Comments · Design, Gamers

When large groups of troops collide in Forge of Freedom, you have three options – autoresolve, tactical combat or quick combat. The first is self-explanatory. Tactical combat lets you control a battle like a traditional wargame and the quick model is all about placing troops in general stances on a grid and letting the thing play out. Pretending you are Stonewall Jackson and fighting all the major battles takes a lot of time and the simplified model isn’t very satisfactory. So, most of the time, it’s all about the auto-resolution. And everyone knows that auto-resolved battles are a quick way to lose a lot of troops.

I noted this in passing in my Medieval 2 review. The AI had a tendency to build silly armies composed of nothing but catapults, crossbowmen and the occasional militia. If you fought the battle yourself, this was an easy combination to beat – a well-balanced infantry army with cavalry doing long flanking maneuvers to take out the artillery. If you auto-resolved against this configuration, you would usually lose half of your army. Automated generals are idiots.

The big difference is that the Total War series gives you strong incentives to control important battles. The combat never takes very long, is visually spectacular and can be decisive in a campaign. The Western Civilization games give you the last one, but to make the battle count, you often need to play a really long battle against a dim-witted opponent. Since the core of the game is the strategic map, there is nothing really interesting to see in the battles most of the time.

Imperialism may have been the first historical grand strategy game that gave you this choice. Would you be Napoleon, winning your own battles as you controlled your empire, or Bismarck, ruling from the capital and trusting your generals? As the technology advanced, the battles in Imperialism became artillery assaults followed by cautious advances and sacrificing skirmishers to draw the fire of defenders. And they were always less satisfying than the main game of commercial exploitation. Just like Total War, the naval combat was abstracted altogether and the capital knockout rules made the grand military strategy game silly, but the heart of Imperialism was the bean counting while you slowly prepared for the inevitable World War.

Then there is the third way – letting players watch battles that they have no control over. Dominions and Galactic Civilizations have taken this route. In some ways it is satisfying because the point of the battles is draw lessons for future combat – not demonstrate your own prowess. As every gamer knows from experience, a human brain can take low tech forces and wipe the map against better equipped or more numerous AI opponents. The third way reinforces the importance of research and resource management; the battles reflect decisions you have already made. No do-overs, take-backs or Teutoberg Forests. Watch and learn.

There is no right way, of course. But it does pose a design challenge. Take GalCiv. It follows in the wake of the legendary Master of Orion series which made a big point of letting the player get involved in fleet combat. Stardock decided against it and a thousand forum threads were born. Knights of Honor went for a real time version of Medieval Total War and let you fight battles while the clock was ticking in the strategy world. There will always be a solid contingent of a fan base that demands complete control over everything. Strategy games make them desktop gods, and not distant deities sitting on cloud – meddling omnipotent deities who need to know that the computer isn’t making mad decisions for them.

This is why auto-resolve is such a tempting button for developers. They can give all this control to people who want it, but tell the rest of us that we can press the button if we like. No one is forcing us to play the battles, so why take away from those who do want to? Part of the problem is that the disparity in outcomes between auto-resolve and manual control is huge. When I fight evenly matched battles in most of these games, I win 75 to 80 per cent of the time. CompuGeneral wins maybe half the time, often less depending on the type of battle. Every gamer, therefore, becomes Napoleon or Lee or Marius, the towering genius of the battlefield.

This is certainly flattering. I hate losing, and controlling all potentially close matches keeps me from losing. But it also feeds into those complainers (often the same ones who relish total control) who find “today’s games” too easy.

So should there be better auto-resolve algorithms? Should design teams be divided into wargame/strategy game parts so that each side is compelling? Is the third way going to become a more attractive way? Any and all thoughts are welcome.

EDIT: Just as I was finishing this post, I noticed that Michael Akinde has decided to drop the tactical battle component of Imperium: Rise of Rome, his eternally in development game about the Eternal City. Since he comments here from time to time, I hope he has something to add.


7 Comments so far ↓

  • Bruce

    When I asked the question about tactical combat preferences in my most recent CGW column, I expected the responses to be one-sidedly for controlling everything in tactical combat. Instead, the responses have favored whatever makes for “the best” game, which seems to mean a combination of most historically accurate and most challenging. Having a an incompetent AI control battles seems to be a big problem for suspension of disbelief.

    The quick combat in Forge of Freedom is anything but quick, but it does seem to give reasonable results for the forces involved.

    By the way, in Dominions, you can’t say you’re watching a battle you have “no control over.” You have a ton of control at the beginning, with the setup, orders, and spell scripts. This has a huge impact on the outcome.

  • Troy

    First, it’s GFW now. Didn’t you get the memo? This was a reader survey, I assume? I look forward to reading it, as always.

    “No control” is Dom 3 a gross simplification, sure, but the things you cite are like general stance orders and change based on your understanding of your troops. But once the battle begins, you are stuck with your plan and hopes that your intelligence on opposing forces was accurate. You have no control over your men once the blood starts.

    An interesting middle ground is the early Slitherine engine, which gave you no control once the battle begins, but perfect information on terrain, opponent forces and opponent deployment right before the battle. You then made a battle plan (hold, charge, etc.) based on your forces and the layout. With practice you could do some pretty amazing things. This is actually a pretty good representation of ancient warfare. Legion Arena, though, moved to the Total War model with supposedly limited control based on a general’s power, but it didn’t take long until you had nearly perfect command of everything in your army on the fly.

  • JonathanStrange

    If a game relies primarily on tactical combat for its entertainment value then I think most of us would then expect the AI’s control to at least appear to be competent in handling its troops. It helps also if there’s some interesting to watch: colorful troops, exotic equipment, great animations. It’s really fascinating to see what might otherwise be a cut-and-dried computer calculation of “Spanish arquebusiers fire at men-at-arms. Hit. 5 casualties.” etc.

    I rarely feel like a honest-to-god Napoleon as our command and control is far beyond what the low-tech generals had available to them. And they could forget about knowing what was what once the battle commenced – although I should note that the gamer is not just the general, he’s also the colonels, majors and captains.

    Maybe a realistic depiction of battle as experienced by a Roman consul wouldn’t be as much fun: limited knowledge of enemy troop dispositions, poor ability to transmit commands, and battles that were probably more of a long bloody shoving match that ended abruptly when one side had had enough.

    Rather than say auto-resolve or complete control is best, I would say that for me having a suite of interesting choices is optimal: cool graphics here, interesting strategic choices there, a light RPG touch here with a beautiful map there, and so on. Sometimes if a game has one really aspect – like Mount & Blade’s depiction of mounted medieval-era style combat – then I can forget its inaccuracies or shallowness in its other gameplay elements.
    That’s rarely the case for me though, I’d rather have (like M2TW) all kinds of interesting stuff: awesome battles, strategic alliances, province development, economic effects, religious authority,etc.

  • Michael A.

    In the case of Imperium, I didn’t really drop the battle module because I couldn’t find a good “solution”; I worked through several, and think that at least a couple of them would have worked out quite OK. But the work effort going into this was simple not proportional to the “benefit” gained from a part of the game people were going to be auto-resolving most of the time anyway.

    I don’t think auto-resolution is really a problem per-se; technically, it (should) be easy simply to resolve battles as regular battles but with two AI generals instead of one with results similar to that of a “full” battle. The problem, of course, comes if the AI handles one’s troops incompetently. I think that is a deal-breaker, every time. As Bruce notes – incompetent AI makes it hard to suspend disbelief. And almost without fail, the AI plays the part of Varro, while the player plays the part of Hannibal.

    The only battle engine I’ve played that I think really worked is the one in Conquest of the New World (http://www.mobygames.com/game/windows/conquest-of-the-new-world-deluxe-edition/screenshots/gameShotId,22397/) which was about as abstract as they come. But it worked, IMO, because battles between evenly matched armies were always tense affairs that could swing either way. It was also very, very quick – and the quality of your general was of extreme importance (basically, he decided how many attacks you got each turn). The system _did_ have a problem with elite all-artillery armies being way too powerful, but at least it took a while to create those.

    I think Slitherine’s “Legion” engine was very interesting; the problem I had with it was always that of sufficient interesting decisions though. Once one learnt the Rock-Paper-Scissors mechanism (and realized which 3 units were of any use), deploying for battle became almost trivial. I could definitely have seen this engine evolving into something unique, though, if Slitherine had continued the development of the engine in this direction.

    The big problem I see with every tactical/strategic mix-up game, though, is that a weakness in either components lets the whole game down. If the strategic game doesn’t provide exciting battles (because the strategic AI can’t put together a reasonable army), it makes the tactical games boring. And vice-versa if the tactical AI can only win battles with 4-1 odds in its favor, the strategic game becomes one of beating off innumerable hordes with one hand tied behind the back.

    Then again, lots of strategy games have the same problem, even without the distraction of battle engines… ;)

  • JonathanStrange

    I was thinking of Conquest of the New World’s battle system myself when I was trying to remember a battle engine that seemed to give reasonable results – the AI didn’t need overwhelming numbers and usually fielded a rational mix of troops – and although its chesslike turn-based combat seemed highly abstract, it did have the virtue of allowing you to quickly make some interesting tactical decisions so auto-resolving through them wasn’t necessary. Oddly enough, these simplified combats can be just as fun as the visually exciting Rome Total War battles.

    I liked Slitherine’s Legion combat system – I thought the pre-battle deployment and limited in-battle control more like what the actual commanders might have had – and I wished this sort of simple combat had been available in other games (liked the city builders, e.g. Caesar III, that I’d played) which didn’t emphasize tactical combat much. If the rest of Slitherine’s gameplay was a bit more exciting I might have played them much more; as it was their Chariots of War, for example, was not as involving all round. However, I felt their streamlined concepts made for interesting and fluid play but maybe many more gamers thought their favorite elements had been pared away.

  • Alan

    It sort of makes sense that auto-resolve would turn out worse than a manually-resolved battle; why would anyone opt to control the battle personally if auto-resolve generated a better outcome? Of course, the other problem is that auto-resolve is based on what should happen, not what actually happens when a human player finds a way to exploit quirky AI behaviors.

    I suppose one solution is to make the AI smarter. Of course, this has a couple of problems as well. The first problem is that, from a game design perspective, you want the player to win more often than lose. The second problem is that you don’t want the AI to use “annoying” tactics or worse, exploits. I suppose one way around this would be to have the AI use efficient but (relatively) inflexible strategies.

    Quick combat is a nice compromise, such that a good player has an advantage but is still bound by the intended combat mechanics. This also has the advantage of making outcomes transparent and plausible, which are probably the two biggest problems with auto-resolve.

  • Carnival of Gamers #22 – March 2007 « VirginWorlds.com

    [...] the RTS analysis category, Troy at Flash of Steel examines the design elements of some of our favorite divide-and-conquer games. Tough topics like autoresolve, tactical combat and quick combat are analyzed. He leaves some [...]