Under fat tails, look for quick reasons to say yes
This note summarizes my thoughts and learnings from a number of interesting articles:
Fat Tails are weird in the sense that the flip the standard logic of optimal evaluation on its head, often necessitating the exact opposite behavior from what one might typically default to. Evaluation routines that make complete sense in a thin-tailed world are completely wrong in the realm of fat tails.
For example, in the realm of thin tails, it’s possible to, with a moderate amount of data, reach a fairly precise estimate of the importance of various characteristics to some outcome. With fat tails, there’s no amount of data within reason that will get you to precise parameter values.
Further, we must be vary careful about over-relying on these estimates, as the cost of saying no to a potentially very good thing could be quite high. One should Bet on positively fat-tailed opportunities, which necessitates suspension of disbelief for a moment. Humility is key – we don’t really know what we think we know, so don’t bother trying to use what we don’t actually know that well to eliminate potentially incredible options.
We should look for reasons to say yes, rather than reasons to say no. Avoid checklists. Almost any amazing opportunity will fail some number of requirements on a naive checklist. Overuse of checklist leads to only picking “well-rounded” options, which might have a decent chance of being better than average but have a poor chance of being exceptional. I see this in so many of my peers, as one example, but it applies in many places.
It’s fine to have dealbreakers, but they should be few in numbers. Use them sparingly, and only use the ones you have incredible confidence in. Be quick in your decisions, but don’t be fickle.
Focus more of your efforts on finding reasons to say yes. What are the characteristics that, if present, could lead to an exceptional outcome? Look for the presence of these characteristics. The best option will spike in some particular way (Looking for the Spike in a Haystack), but may be mediocre in many other ways. Pass on opportunities because they lack these spiky characteristics, rather than because they have (or don’t have) some other, moderately bad (good) characteristics. Rule things in, rather rule things out.
Don’t evaluate each option in isolation, from a blank page. This is likely to be error prone and exhausting at a minimum. Place each decision in the context of past decisions made by yourself and the collective experiences of others. Does this item look like the others that have done well? What are the shared, potentially exceptional characteristics?
Once you know what you want, be ruthless and quick in evaluating opportunities. Don’t dwell. You need to take many samples to arrive at a potential “high draw” from a positively fat-tailed distribution. Overindexing or optimizing over a single or small numbers of draws is a waste of time, as measured in terms of the opportunity cost of what you could have if you were only to draw more samples.
Remember, You likely haven’t seen the best or worst, so if something is clearly far from the optimal outcome, you should cut bait quickly and feel no guilt in doing so. Similar to product-market fit, “fit” that isn’t found quickly is likely to never materialize (Product-Market Fit is Lindy), so don’t beat a dead horse. Get a move on!
References
You likely haven’t seen the best or worst
Craftspeople How to Create Good Work
Transclude of Craftspeople-How-to-Create-Good-Work#23bc81
The Case for American Seriousness - by Katherine Boyle
Transclude of The-Case-for-American-Seriousness---by-Katherine-Boyle#abc2eb