Heisengames and the importance of patience in business

Marton Trencseni - Mon 08 February 2016 - Business


Most bets businesses take, be it hiring, features, products or strategy don't work out. Only a few of them are ever successful. Analytical employees often notice this, and become negative: "We just hired executive E, she did X, Y and Z, and it's not working!", and similar sentiments.

Still, many businesses are successful despite setbacks. A negative attitude---even when the analysis of the situation is in fact correct and X, Y and Z were failures---may be missing the bigger picture. Putting aside the demotivating psychological aspect of negativity, how can you be right and still miss the bigger picture?

Toy model

Let's model the business as a dice game you're playing at a casino. The game is simple: if you roll a six, you get $X. The game costs $1 to play at the casino. Clearly, the breakeven point is X=6.

Suppose the casino offers the game at X=7. In this case, you still lose 5 times out of 6. However, when you win, you offset your losses, and in the long run you will make a lot of money. (Unless you hit a losing streak and lose all your money, in which case you don't have any more money to keep playing the game; but let's ignore this.) This toy model illustrates the point about business and negativity: most bets will be "failures", but in the long run, it can still be a (very) profitable game!

Of course, it's also possible that X=5. In this case the negativity is justified, because even when we win, it doesn't offset the cost of playing the game. Unless...


We can make the game even more interesting. Suppose X=5, but every time you win, the casino increases the payout by a cent. In this case, most of the time you lose, and initially you also lose out on average, but if you keep it up long enough (eg. using venture capital), eventually the game will become (very) profitable! I call this Heisengame, because playing the game changes the game itself.

We can also turn it around. Suppose X=7, but every time you win, the casino decreases the payout by a cent. In this case, analysis will show that the game is currently profitable on average, but it won't be in the long run.


In real life, the payout itself is also not fixed. You can model it like this: the casino may pay you a random amount when you roll a six, say according to a normal distribution centered on X. In this case, you need to play even longer to learn what the game actually is, because you need to collect many data points to understand the payout function. You need to have patience to understand the game.


These are just toy models, but they illustrate important points:

  1. Just because most bets are losers isn't a problem in itself. As long as the winners generate a big enough win, the game is worth playing. To know which game the business is playing, long term data and analysis is needed.
  2. Business is a Heisengame: playing the game changes the rules of the game. Long term data and analysis is needed to tell the direction of the change.
  3. In the real world, you don't know the parameters of the game. You need to be patient while you collect data and measure the parameters.

So next time you see a bet not working out, consider that maybe the business just rolled a five. As long as you win in the long run, it's okay.