> research · 3 studies
Market ideas, tested on real data.
Out of sample, costs included, sources linked. We publish the answer either way — most ideas that sound clever don’t survive the test, and knowing which ones is worth money.
latest · Sep 23, 2026 · backtest
−10.6%No edge
weather model, 20 months out of sample
Arabica moves on weather forecasts, days before the thermometer confirms anything — and the stocks that buy the beans feel it a year later, through margins.
stocks in itSBUXSJMKDPMCDKOXLP

data · ERA5 weather · ICE arabica · 2016–2026
Earlier studies
question · answer · evidenceSep 20, 2026
field study
Portfolio, Positions, Holdings and three more. Seven brokers don’t open on your holdings at all, and one buries them three menus down.
data · 17 brokers, menu paths recorded · Sept. 2026 · import from your broker →
6 namesfor one screen, across 17 brokersNo convention
Sep 19, 2026
event replay
Gold rose 7% in the two weeks after the 2022 invasion, then rate hikes took it below where it started. A long-run diversifier, not a two-week hedge.
stocks in itGLDXLESPY
data · Brent, COMEX gold, STOXX 600 · 2019–2026 · replay shocks on your portfolio →
−14.6%gold vs its pre-war price, Nov. 3, 2022Not for long
House rules
- 01
Out of sample
A rule is judged on data it never saw. In-sample wins are where most backtests go to lie.
- 02
Costs counted
Every trade pays for itself. An edge that vanishes after fees was never an edge.
- 03
Failures published
A clean “no” is a result. It saves you the money of finding out yourself.
- 04
Sources linked
Every figure traces to its source, and the code behind each backtest is kept with it.
The same tests, on what you own
Replay past crashes on your holdings, see how concentrated you really are, and track every position in one place.
Research, not recommendations. Not investment advice.