Life lessons learned from XCOM Impossible Ironman

Dmytro Mishkin
4 min readJan 6, 2021

--

Recently I have twice beaten XCOM in the hardest mode — Impossible Ironman. It took me a lot of time to learn things, which helped me to beat the game. Ironically, most of the things I learned are not related to the game. Are you ready? Let’s go.

In case if you are not familiar with XCOM — it is a modern reboot of the classical tactical game from 1993, where you command a small number of soldiers in order to protect the Earth from the Alien Invasion. “Ironman” mode means that you play with a single save, which is overwritten when you make any action. Impossible level of difficulty means that aliens can one-shot any of your soldiers, while you mostly cannot. And there are more aliens than your soldiers on the map. And things get worse every minute.

To be outnumbered is a typical situation in XCOM.
  1. Mindset is your most powerful weapon, but not in the popular “you can do it” way. “I can do it” is bullshit. Well, not bullshit, actually, but is only a part of the solution. The full and correct version is “I can do it, let’s find the proper way”. You cannot win just because you believe in yourself. You have to experiment and find things, which work and those, which do not. The easier difficulty, the more strategies works. The harder — the less. Things, which allowed you to win easily now lead to disaster. There is even a DeepMind paper about it.
  2. You don’t play when tired or angry. Once I have an argument in a cafe and end up upset. I have opened the game (I play on mobile) to distract myself. Two turns — and half of my squad was wiped out. Luckily, I have found the only winning move — close the app and go for a walk. The problem is that when you are tired or angry, or upset, your attention is flawed and you make mistakes without realizing it. XCOM I/I is unforgiving.
  3. Discipline means not to play the lottery. Ever. Imagine you take a risky decision and got lucky. Good for you. Actually, not good. If you get away with it once…you will be tempted to do it again. And once it will kill you. This takes us to the next point:
  4. Bad and risky decisions lead to even worse and riskier decisions. What bad can happen, if you just reposition that soldier one tile away, right? Well, for example, alerting three Chrysalids. Now, to get away from it you HAVE to get the rest of the shots to hit. Which they didn’t. But…that were 80% and 90% shots, how can it be?
  5. Humans are bad at interpreting probability, expressed in numbers. How likely is it to miss two 90% hit shots in a row? The answer is — it happens every single day. I have read Taleb’s Antifragility, but only now I am starting to understand what he meant.
    You don’t operate in probabilities, you operate in possible outcomes. “First, avoid ruin” — great wisdom from Taleb.

6. However, sometimes you are screwed from the beginning. You can do everything correctly and just get unlucky. That happens. Now, it is a time to risk — when you have nothing to lose when you have already lost. Then take that 10% shot and pray. Miracles happen, sometimes.

7. With all the luck involved, most often it is not the current situation when you lost the game, but the subtle decision you made a (game) month ago. Remember when you decided to build an extra power generator, which delayed the laser weapon research? That is the real reason, why you lose this fight and not the bad tactics now.

8. You cannot win on the hardest difficulty without overfitting. Just “being good at tactics” is not enough. You need to know all the weapon and alien stats. Calculate the base construction month ahead. Otherwise, see point (7). But it is never ONLY overfitting which leads to victory.

Yet, overfitting is the most visible thing to the 3rd party observer thing. Let me give you an unrelated example — Kaggle. People, who don’t like it, often point out“unrealistic” things like leaderboard probing and similar stuff, which is done by Masters and GMs.
However, they miss the point — when you compete on the so high level, overfitting and leaderboard probing alone bring you nowhere. You have to be both the data scientist of the highest category AND then know “overfitting” tricks to be the first.

9. You can never come back to the easier levels. That’s actually a funny thing. Once I finished my first I/I run, I decided to relax and go for C/I — Ironman, but on the easier — Classic (3/4) difficulty. I played for an hour and got bored. Things, which before the I/I were challenging, now are boring and easy.

--

--

Dmytro Mishkin
Dmytro Mishkin

Written by Dmytro Mishkin

Computer Vision researcher and consultant. Co-founder of Ukrainian Research group “Szkocka”.