Tools
Each of these is code you can read, with every number computed and none typed. The point is rarely the answer. It is seeing which assumption the answer depends on, and then changing that assumption yourself.
Each of these is code you can read, with every number computed and none typed. The point is rarely the answer. It is seeing which assumption the answer depends on, and then changing that assumption yourself.
A discounted cash flow, a sum of the parts, and a Monte Carlo that collapse the whole bet into one question about the containerboard cycle. Set what you believe and the page prices it, then sizes the position with Kelly. Built from the public filings and twenty-one tie-out checks against them.
The Treasury par curve, the ten-year over time, and the household's tickers, written to data/market.json by a scheduled job. The page you open first when the portfolio has had a big week in either direction.
Two brokerage exports go through tools/import_positions.py and become holdings-data.js; every number on the ledger page is derived from that file at load. Contributions are the one hand-kept list, because no export carries them.
A Black-Scholes engine with Greeks, implied volatility and a binomial tree, specified by tests transcribed from Hull's worked examples; retrieval over the books I own, with page citations; and an agent that explains but never does the arithmetic. The engine is my homework, written one function at a time. The workbench comes after.
The operations-research work, the corrugator scheduler, the truck-cubing model, and the writing about them live on my own site, justfetz.github.io. This site is only the money and the letters.