And the primer on branch prediction! youtu.be/nczJ58WvtYo

I did have an idea that did not work. Take a continuous indicator of regime change a use it a $vol or gmv modulator over time. Say, something that says “if the value is above one, run at full volatility. Below zero go to cash.” A large class is made by indicators of the form a*drawdown(t) b*t. This includes stop-loss, cusum, and approximate others. It took me two full days (and a lot of claude code) to determine that the only case that “works” is b=0. I.e., stop-loss. “Works” means that a solution exists and is nontrivial. So there is something special about stop-loss. My tentative recommendation is *not to use statistics used for tests of hypothesis outside of the intended application.* Obvious, maybe. And another surprise. Stop-loss reduces sharpe, but only up to a theoretical max reduction of 50%. I feel that extensions to quadratic costs are within reach. I am writing this down and sending to my colleagues (esp. the sys PMs). Notes: 1. The model is in continuous time. The math generated by claude seems correct but is a bit above me. 2. I thought of this problem last Sunday. This took two days of math and code-assisted simulations. I could ask questions between meetings, and at night. On my own, I don’t know if I could have gotten an answer ever. 3. There are still five days this week. Gotta hurry up. I wonder what we’ll find out. There is no time like the present, people.

The smartest man ever John von Neumann made groundbreaking contributions across fields. He was one of the key mind behind the atomic bomb, early computers, game theory, and quantum mechanics. He even theorized self-replication before DNA was discovered.

And then go here: ema.cri-info.cm/wp-content/u…

Many #LeanLang users were first introduced to Lean via the Natural Number Game, a gamified approach to learning mathematical proofs developed by Kevin Buzzard. The Lean Game Server now hosts 8 games, including real analysis, linear algebra, and introduction to proofs. Open source, so educators can build their own too. Explore: adam.math.hhu.de/ #FormalMathematics #MathEducation #OpenSource