the book · open strategies and closed positions
Everything the desk is running, has run, or refuses to run — listed in order of embarrassment, ascending. Source stays private while the opinions stay free; when a strategy survives a year out of sample, it graduates to a memo.
An expert advisor that traded lunar phases. Full moon: risk-on. New moon: flat. The moon turned out not to be a leading indicator of anything except tides. Blown up on a waxing crescent — which was, in hindsight, the system's single most accurate signal.
A sentiment model that reads earnings-call transcripts and scores how much of each sentence is load-bearing. Works beautifully, except on Tuesdays, quarter-ends, and any call where the executive has been media-trained. Currently being interrogated on data it has never met, per filing № 002. It is allowed two parameters and it knows why.
Five positions, rebalanced when the vibes stabilize, which is rarely. The strategy is the refusal to touch it — every urge to trade is logged, stamped, and filed under cost of doing business. Expected to be boring for thirty years, at which point it becomes the most interesting thing on the desk.
Static HTML, one stylesheet, two typefaces, no build step, no dependencies, no analytics. The only system on this desk that has never blown up, and the only one whose complexity compounds downward. You are looking at the complete documentation; the source is the product.