If you’ve been following the series up to this point, you’ve undoubtedly been waiting patiently for something which up until now I’ve avoided doing, and I don’t mean taking unsolicited cheap shots at Tom Chick. That comes later. I mean talking about the AI.
After the jump, Gary’s AI Adversary, or GAIA (they really do call it that)
I’ve avoided making any real comments about the AI because I’m only eight turns into one scenario. The Operation Barbarossa scenario doesn’t give the Russians any real chance to counterattack, so the AI is limited to forming line after line of defenses for the Germans to breach with their panzers. So far, it had been doing a fairly good job of this, at least from the perspective of a human player on his first attempt at this scenario. It’s impossible to stop the German armor entirely, and the AI has, for the most part, done a good job wearing my units down without making unnecessary stands. A lot of Soviet units are going to get encircled in the early going, but that’s History’s fault, not the computer’s.
Until just now, which is the first turn where I think the computer made a clear mistake. It made two major decisions in its half of Turn 7, which were to abandon both Smolensk and Kiev. In the former case, I think it was probably the right decision. I had Panzer Group 3 driving north of the city and Panzer Group 2 driving south, and it was completely screened by infantry and about to be isolated. It probably wouldn’t have lasted more than another turn, and the casualties incurred keeping it for that turn were probably not worth the one extra victory point it would deny me. So good move, Comrade Computerski.
The Kiev decision is more complicated. I drove 48th Panzer Corps north on Turn 6 after crossing the lower Dnepr to threaten Kiev from the east. Here’s a cool screenshot of it.
I was pretty pleased with the move when I thought of it. As you can see, there are virtually no Soviet defenses east of Kiev north of the Dnepr. You can’t see it, but there actually isn’t much south of the Dnepr, either. Army Group South has broken out and is threatening to race for Rostov. It all looks very dangerous and blitzkrieg-ish. In theory.
The problem is that completing that encirclement would have exposed 48th Panzer Corps to its own encirclement, and I’m not sure I could have kept that from turning into a major debacle, had I pressed the attack. In order to do it, I had to secure a bridgehead over the lower Dnepr and threaten a breakout to the east, which spread me pretty thin. And I couldn’t abandon the drive southeast through Kirovograd on the other bank, so I was driving three different directions with limited troops and a shoestring supply train. Attacking westward toward Kiev was probably not going to work. But it sure looked dangerous, in a theoretical sort of way.
And the computer bought it.
This, citizens, is why the quest for “human AI” is doomed to failure, no matter how many Jon Shafers and Soren Johnsons we throw at the problem, for the simple reason that in the end, I know it’s an AI, and not a human. No matter how well it plays, the fact that I know it’s artificial freaking intelligence colors my reaction so dramatically that I’m not sure we as a species can ever solve this.
If this had been Tom vs. Bruce: War in the East, I would have screamed, turned the volume up on The History Channel, sent out a taunting Twitter message, and then gotten into my Tardis for the long trip back from whatever crazy dimension I was in that had Tom Chick still playing hardcore wargames. Because the threat to Kiev was 90% illusion and 10% reality, the decision to abandon the city was not militarily “correct”. Psychologically, though, it probably was. I’d just chalk that up to out-bluffing Tom Chick, think up a witty joke to drive the point home with style, and then send the whole thing to Jeff Green. Talk about paradise lost.
Instead, because I know that the AI isn’t afraid, or uncertain, or any other psychological adjective you can come up with, I’m left with the feeling that whoever Gary Grigsby assigned to write the AI while he concentrated on making the interface as inscrutable as possible just wasn’t up to the job. After all, it should be able to just think Three Moves Ahead, figure out that I’m not going to be able to carry out attacks on three separate axes with the limited forces in Army Group South, and move accordingly. Against a human, it’s a big win for me. Against the computer, it’s like finding the crack in the drywall in your new house. I paid for the thing, so it should work, right?
And if it worked, what would it do? Make the “correct” decision every time, like some kind of robot? Sadly, yes. Because these games are supposed to be historical, part of playing them is about finding out whether a different historical outcome is possible against a properly played opponent. Kind of like a counter-factual history simulator. And when it makes mistakes, the illusion falls apart.
So how is the game going? It’s currently 7 August 1941 (turn 8, of 24 total), and Army Group North is at the gates of Leningrad.
If you look at the screenshot, you’ll see the fortification multipliers on the defending Soviet units approach 100. If you recall the explanation of the combat system from last time, this means a stack of units with a combat value of 13 will have an actual strength of between three and 30,000,000. All I know is I am going to have to rest and refit 4th Panzer Group if I am going to have any chance to taking the city. The actual victory point hexes are marked with red stars. As you can see, it will take a while to reduce that fortress.
In the center, 3rd Panzer Group is a scant 11 hexes from Moscow. However, supplies, fuel, and ammunition are low, and I can’t imagine being able to take that city without some sort of rest and refit period, either.
Just south of that, 2nd Panzer Group (Guderian) is a similar distance from Tula. The drive on Moscow will start in earnest when (if) 2nd Panzer takes Tula and turns north.
The south is, like I said at the outset, pulling troops in three different directions. I will have to reorient the offensive, drive through Kharkov (top arrow), and link up for the drive to Rostov (way off that map to the southeast).
This is the tenth of these almost-daily game diaries for War in the East. I was originally going to do a week, and then it became two, and I don’t want to stop because I’m enjoying it a lot. But the fact is that I don’t have the time and can’t keep it up at this pace. So my compromise solution is this: I’ll try to post weekly updates until I finish the whole scenario. That should be a few weeks yet. Stay tuned until then. I’ll leave you with some Glantz to tide you over.
Army Group Center’s advance to Smolensk and encirclement of the Western Front’s 16th, 19th, and 20th Armies in the region northwest, north, and northeast of the city proved to be a textbook exercise in the conduct of an encirclement operation. […] Thus, on 23 July Army Group Center was justifiably proud of what it had achieved in the course of one short week. The question was, could the army group digest the encircled forces without losing the high offensive momentum it had generated and maintained since 22 June? Much of this depended on the actions of the Stavka in Moscow and Timoshenko’s defeated and shattered but still not liquidated Western Front.
Click here for the previous War in the East entry.