Research
Research
Two engines on one data stack. Hypothesis tests ask whether a trading idea had a historical edge — validated against random entry, three sub-periods and ±15% parameter shifts. Scenario stresses ask what a future shock would do to a portfolio — shocks and betas both estimated from real data. Diary-framed throughout: historical observations and sensitivity estimates, never predictions or advice.
SaaS valuation, decomposed
Split a software valuation into its three components — cash-flow base, discount rate and duration — then dial each, alone or together, and watch the price move.
Chip re-rating, decomposed
The cyclical sibling: split a semiconductor valuation into normalized earnings, the discount rate and the boom-bust re-rating — and see why spot P/E lies. Memory archetype.
HYPOTHESIS
Semiconductors after a sharp fall in 10Y yields
9.0% avg · 2.5% excess @63d
56 events · FAIL
Open →
HYPOTHESIS Tech recovery after a QQQ drawdown
15.6% avg · 8.2% excess @63d
24 events · FAIL
Open →
SCENARIO AI / Tech de-rating
-17.1% est. portfolio impact
4 holdings · closelook_calibrated
Open →
SCENARIO Credit crisis
-18.2% est. portfolio impact
4 holdings · closelook_calibrated
Open →
SCENARIO Inflation re-accelerates
-1.0% est. portfolio impact
4 holdings · closelook_calibrated
Open →
SCENARIO Rates rise sharply
-1.3% est. portfolio impact
4 holdings · closelook_calibrated
Open →
SCENARIO Recession shock
-24.2% est. portfolio impact
4 holdings · closelook_calibrated
Open →
SCENARIO USD spike
-8.5% est. portfolio impact
4 holdings · closelook_calibrated
Open →
Methodology. Hypotheses detect rising-edge triggers point-in-time (no look-ahead, entry t+1) and measure forward returns vs a benchmark; the verdict requires beating a random-entry null, holding across a three-sample split, and surviving ±15% parameter shifts. Scenarios apply a factor shock via each holding's downside beta (estimated on factor down-days). Data: Closelook lake + EODHD + FRED. Read more →