Du Malone writes: In my early days on Football Manager, I was manged Stoke. The third season proved unexpectedly tough. By the end of the season, we were staring relegation in the face.
In the closing minutes of the final fixture — I think it might have been in stoppage time — we needed to score to stay up. And then, miracle of miracles, we did.
And then West Brom went straight up the other end and, with the final kick of the season, scored, thereby consigning us to relegation.
It was the final kick of that save too: the board fired me and I never do journeyman saves.
I accepted my fate. Though the manner of our execution was sickening, in truth we’d not been good all season.
The improbable strike force of Delfouneso and Anichebe had performed way above expectation in my second season, but not in my third. Anichebe’s injuries hadn’t helped
But before terminating the save, I decided to do an experiment. Granted, the main reason for our relegation was performance over the season, but were we actually destined to lose that final match?
So I exited FM without saving, restarted it, went on holiday, and looked at the score when I returned. I did this numerous times.
The scores varied. In the genuine match, we’d lost 2-1. In the ‘holiday’ iterations we usually lost, by narrow margins — though there was a 3-0 defeat. And there were also a couple of draws.
Subsequently I’ve occasionally experienced the game-crashing during a match. In those circumstances, of course, you have to replay. As matter of ethics, I don’t change the starting line-up or tactics in the light of what I’ve learnt from the aborted match (on the basis that the AI won’t have that advantage). I don’t, though, feel duty bound to make the same in-game decisions, since every match evolves in its own way.
But, as with the Stoke ‘holiday’ matches, I notice that, from the start, the second iteration of a match can develop differently.
What to take from these experiences? The most evident learning point is that there are stochastic factors built into the match. Whether a lino (as I still call them) spots an offside, for example.
I guess FM-ers all know that.
Yet a review of online content about FM reveals that we are very loath to accept the import of this point. We make a change; something happens; ‘Ah!’ we think, ‘That happened because of the change I made!’
Well, maybe it did. But we shouldn’t assume it did — because there are things in the game beyond our control and some of those are governed by stochastic factors.
The Romans had a phrase for the assumption of causation: ‘post hoc ergo proper hoc’ — in other words, ‘after that and therefore because of that’.
In thinking about FM, this assumption is virulent. You see it in FM blogs, for example, all the time. It’s everywhere.
And though the rational part of my brain knows it’s an assumption that cannot be relied upon, I still fall prey to it. As I say, it’s virulent.
Recently our manager, Grigor Pasha, had a wretched run with his Pomerie side. They weren’t getting hammered, but they just couldn’t score and so they sank into the relegation zone.
Then he made two changes. He flipped his asymmetric formation to its mirror image (playing a wide target man on the right instead of the left, for example). And he decided to play a higher defensive line.
Pomorie won 4-1.
‘Well done, Grigor,’ I thought. ‘Those changes worked!’
But then Pomorie lost a match 0-1. What do we say then?
The big picture: the kind of game FM is
If we compare games, we can arrange them on a spectrum according to the contribution that luck makes. At the one end, chess; at the other, snakes and ladders.
My argument here is that a look at FM content and, I suggest, reflection on our own thought processes, reveals that we constantly act as though FM is closer to chess, and further from snakes and ladders, than is actually the case.
The corollary, I suggest, is that the biggest single thing we can do improve our performance as managers is simply to catch ourselves falling prey to this assumption. If we can do that, we give ourselves a chance of not over-estimating the significance of short-run data.