Thread by @tarunchitra

The VC vs. trader “war” of crypto is reminiscent of the previous talent “war” between HFT and online ads: All of these boil down to latency vs. bandwidth trade-offs where "event-driven" investing depends on the condition number of a participants' value function

Trader: need max and min eigenval. of value fn. to be "close" (low condition number) because of regret minimization between your worst and best case outcomes

If your value function is smooth, this gives uniform bounds on the max/min eigenval. of hessian of your val. function

VC: need max eigenval. of value function to optimized

Things like the Tracy-Widom law force you to chase fat tails, terrible Sharpe, and anomalous portfolio construction

But in crypto, interpolating between “condition number maximizer” and “highest eigenval.” maximizer is possible and market condition dependent

DeFi, for better or worse, is better for condition number optimizers

But: All of this depends on knowledge of your value function.

Much of online and reinforcement learning assumes you don’t know it or learn it along the way

Philosophical question: can you ever know if a market favors condition number or max curvature optimization from data?

Local measurements likely can’t tell you the answer — one need only look to Morse theory for simple yet impractical counter examples

So who can interpolate the extremes and be the Hunter S. Thompson of this market?

🧐🧐🧐

For what it’s worth: I’m not saying that value functions are smooth, but like a lot of machine learning, I’m trying to gain intuition by thinking of what the world looks like *if* adversaries value functions *happened to be* smooth

tl;dr: traders can't hold too, long, VCs need to hold for a while, but holding for a while only works in illiquid market (low bandwidth) whereas holding for short times works in liquid markets (low latency)

Value function properties encode this difference!