Research note
Score Backtest: four replays, unknown frozen results
Directional, contrarian, plumbing-defensive and SPY buy-and-hold retain distinct rules. No frozen four-strategy result or overall verdict is recorded.
Backtest verdict: Unknown — no frozen result recorded in the inventory. Directional, contrarian, plumbing-defensive and SPY buy-and-hold are separate replay strategies; none receives a page-wide PASS here.
Open this backtest in Backtesting
The rule
The market-wide engine trades SPY. Directional hysteresis positioning enters long at or above the long threshold, exits to cash at or below the cash threshold and holds its previous position in between. Contrarian positioning reverses those thresholds. Plumbing-defensive holds SPY unless the plumbing kill switch is tripped; buy-and-hold stays long. This is not a shorting model.
The Score page opens at 10 years, a 14-day rebalance cadence, long threshold +20, cash threshold −20 and 5 bps per position change. Its 5-year range uses the same cadence; 15 years uses 21 days. These are defaults, not measured results. Prior-score positions apply to the next return; stops, targets and trails are not specified by these four rules. The retained Lab defaults tool separately opens at 5 years.
Methodology
Evidence as of: the INVENTORY review dated November 5, 2026; this is the consolidation planning baseline, not a new observation window. Publication does not extend any study's sample.
The replay combines reconstructed composite observations with the engine's SPY price history, using prior-score positioning rather than same-period returns. Native statistics, sampled equity, rules and notes remain attached to each result. Position-transition count is not necessarily a count of closed round trips.
The inventory records no frozen start/end, four-strategy output, comparable drawdowns, significance or study verdict. Check the captured output and Yahoo price basis rather than assuming total return. Macro revisions, availability/lag, the reconstructed sample and threshold selection limit inference; changing thresholds is another experiment.
Results
| Study / variant | Inventory evidence | Window / basis | Strategy return / CAGR | Benchmark | Max drawdown | Closed trades / observation N | Significance / IC | Native verdict |
|---|---|---|---|---|---|---|---|---|
| Directional | D07 / F15 / H16 | not recorded | not recorded | not recorded | not recorded | not recorded | not recorded | Unknown — no frozen result |
| Contrarian | D07 / F15 / H16 | not recorded | not recorded | not recorded | not recorded | not recorded | not recorded | Unknown — no frozen result |
| Plumbing-defensive | D07 / F15 / H16 | not recorded | not recorded | not recorded | not recorded | not recorded | not recorded | Unknown — no frozen result |
| SPY buy-and-hold | D07 / F15 / H16 | not recorded | not recorded | not recorded | not recorded | not recorded | not recorded | Unknown — no frozen result |
The linked workspace retains the complete native output and notes when available. Missing evidence is not a zero return, a successful empty result or a failed strategy.
Verdict and limitations
Unknown — no result/verdict recorded in this inventory. These replays are tools for inspecting rules and returns, not evidence that all composite strategies pass. A favorable curve in one sample does not validate different thresholds, costs or windows. The in-sample Lab mode and separate annual/OOS validation should not be conflated.
Sources and live backtest
INVENTORY D07, F15/H16; F10 identifies the retained Lab configuration. Sources to check: frontend/components/strategy/ScoreBacktestBoards.tsx, frontend/lib/score-backtest-page.ts and backend/app/scoring/strategy_backtest.py (_apply_strategy, rules and run output). Any numerical claim requires the exact historical result/window/config; none was acquired here.