3 thoughts on the Twilight Struggle AI

Twilight Struggle, the board game that places you in the role of the USA or the USSR during the Cold War has been, by far and away, my favorite 2-player game. I don’t think I need to rehash why the knife-fight levels of intensity the game offers keeps me engaged as every decision becomes significant when it has the opportunity to snowball and when the very act of opportunism rules the day.

I loved it so much that I backed the Kickstarter for the digital edition. Although I prefer playing live games, I’d be happy to give the online version a go. Moreso, since Twilight Struggle on Steam also allows you to play single-player against the computer AI.

Now, this post isn’t going to do an in-depth review of the Steam version of Twilight Struggle. But to give a brief rundown, the online version is good if you’re looking for human opponents and bad if you’re looking to play against the AI only since it makes some head-scratchingly bizarre moves that don’t seem to give it an advantage. So as I’ve gone and done livestreams of boardgames, I figured Twilight Struggle would present a good opportunity to explain what I’m doing and why. See below for my first ever playthrough!

Here are my 3 thoughts on the AI’s performance in Twilight Struggle:

• The computer is pretty good at setting up realignments: Although the computer is pretty bad at the overall big picture of playing its cards or setting up the board, it is pretty good at figuring out realignments. I was impressed at how pesky the computer proved to be in South America, when it used a successful coup in Uruguay to push me out of Argentina. Trying to maintain my Control levels in South America required that I put in a lot of resources into that region, resources that I sometimes didn’t have and forced me to make some unfortunate tradeoffs.
• The computer is bad at defending its regions: One area in which the computer had an advantage was Africa, but it did a poor job of defending it. When I used South African Unrest to really gain a foothold in the southern African territories, the resistance the AI put up was not effective, and eventually, I was able to make Africa almost irrelevant in the Final Scoring Phase.
• The AI failed to see the value of NORAD: Now maybe I’m a bit biased, but I really, really love NORAD. It’s a card that, if you control Canada, allows you to ping any territory that you have influence in when DEFCON goes down to 2. Which happens quite often since the USSR has all the motivation in the world to coup early and often. The implications of NORAD’s ability means that it can often turn territories that are deeply Soviet into neutral territories in a very effective manner. Normally, putting 1 influence in a USSR-controlled country requires 2 Operations Points, but this gives it to you for free!. On top of giving the USSR headaches, you can use NORAD in conjunction with cards like ABM Treaty to do a double dose of NORAD. A truly strong card that, in this game, was severely under-utilized.