Perception of chance in past and future: the narrative fallacy

Let me first state and discuss briefly two truisms
  1. Learning from experience is A Good Thing;
  2. We perceive the future differently from how we perceive the past.
For (1), remember as a child learning how to crack an egg; common sense says most of our (non-instinctive) physical abilities, and many mental abilities, were learned by the "trial and error" method, adjusting behavior in the light of personal experience of what worked last time. In (2) I am thinking of one's personal past or future. Of course the future is uncertain, like a very fuzzy photo, while the past is definite, like a sharp photo, but this has perceptual consequences we're less aware of. In research described in Dance with Chance some students were given a brief description of a fictional person about to leave on a trip and asked to imagine and write what might happen on the trip; other students were asked instead to imagine the person having just finished a trip. The latter tended to give more detailed descriptions than the former, even though the assignments are logically the same.

How is this relevant to our perception of chance? Consider a hypothetical (American) football game, with zero point spread -- the consensus is that the teams are equally good. At the start of the game the outcome is "just chance", in the sense of unpredictable. What do we say, after the game, about why the winning team won? We might ascribe the result to
particular instances of skill or lack thereof -- particular good or bad plays;
more general "playing well" or badly;
particular instances of good or bad luck.

Now what I suspect (an interesting undergrad project would be to get data by looking at actual news reports) is news reports and analysis refer much more to skill than to chance. Moreover, I suspect that when the result is attributed in part to chance -- e.g. success or failure of a last-second long field goal attempt -- a report implicitly identifies the chance result of the match with the chance result of this specific play, even though this particular play actually makes no more contribution to the final result than many other potentially or actually scoring plays.

So here's the point. Details of what happened within one match may be relevant to players or coaches -- who have the scope for learning from experience -- but aren't really relevant to predictions for the next game (aside from cases like an injury or a substantial change in your assessment of a player's ability).

The phrase the narrative fallacy was popularized in the The Black Swan, and here's how I interpret it. When we write the details of some interesting past Episode -- a 3 hour football match or a 3 day battle or a 3 year war or a 30 year career or a 300 year rise and fall of an empire -- we select and recount events sequentially, and within this "narrative" a reader is inclined to assume that one event was caused by the previously described events. Sometimes the writer is explicit about causation (I bet no historian ever wrote "war broke out on the date ...." without attempting to say why!) and sometimes the writer isn't; but our mental processes tend to lead us unconsciously toward the post hoc ergo propter hoc assumption. The main point is that even if one can identify proximate causes of particular events within a larger Episode correctly, after the fact, this doesn't contradict the hypothesis that the way the Episode as a whole developed was unpredictable. Seeing the historical details makes us forget it might have just been chance; the Frog might not have been a Prince after all.