Making a game more fair reduces its variance

# Making a game more fair reduces its variance [[Fair games can be gamed]]. As a consequence, there is usually a single clear path to win at the game, which means that the sort of people who succeed will tend to be similar. The fact that [[Fair games are self-limiting]] compounds the similarity of winners because dissimilar people won’t even play. It feels a little tautological that if you reduce the variance of the people who are playing the game, it will create lower variance output. For games like marathons or chess, that’s fine - their outputs naturally have capped variance. By contrast, capping the variance of games that depend on outlier results - science, startups, hedge funds, art - seems deadly to its output. [[You can’t cut off just one tail of a distribution]]. This may be one reason for why science is stagnating ([[Is Science Slowing Down? - SSC]]) while the startup world feels relatively healthy. The road to academia has been made incredibly legible - and thus, low variance. University admissions, grad school admissions, faculty hiring, and tenure all have clear paths to success. While there are some high level similarities, the ingredients to startup success are still incredibly opaque. (And hence the proliferation of gurus, thought leaders and bullshit vendors - [[Full-time thought leaders try to become gatekeepers to knowledge]].) In addition to destroying the output of institutions that depend on high-variance outcomes, reducing variance can reduce the average quality of people in the institution by making it clear to high-variance people who don’t fit the mold that they won’t be able to have a good outcome. This feedback loop with the fact that [[Fair games are self-limiting]] pushes high-variance people towards the remaining high variance games - in the early 21st century, this is startups. Note how everybody in startups are obsessed with “talent” but nobody in science is. This seem like an odd statement consider how much we attribute scientific success to “genius”[^1] and the fact that [[The things that make “great talent” in high variance industries boils down to the ability to successfully make things happen under a lot of uncertainty]]. But on the ground, it feels like people don’t act as though this is true: I don’t see professors or departments obsessing over finding and recruiting the best grad students or postdocs. There’s no feeling that “getting the best of the best is absolutely critical to our success.” By contrast, the startup world has, if anything overindexed on that with the idea of the 10x engineer etc. This of course, is a generalization from my experience. It might also be an artifact of the different organizational structures: in science the people in the department get less benefits from the success of others and there may be an implicit assumption that nobody is going to do truly great original work as a grad student? ### Related * [[Power laws in people quality matter more in high-variance activities]] * [[The ARPA Model is high-variance]] [^1]: Thanks to [[Michael Nielsen]] for calling this out.  [Web URL for this note](http://notes.benjaminreinhardt.com/Making+a+game+more+fair+reduces+its+variance) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Making+a+game+more+fair+reduces+its+variance)