Prediction as the durable edge
The tests did not justify treating point return forecasts as a dependable source of portfolio advantage.
01 / THESIS
The question before the result
Can risk structure survive where prediction does not?
BASE began with a practical problem: a portfolio still has to allocate capital when return forecasts are fragile. The test combined a diversified risk structure with time-series trend, sleeve-level volatility targeting, a deliberate blend, and a governed drawdown brake.
The aim was not to make uncertainty disappear. It was to determine which mechanisms remained useful after the predictive claims were removed.
Research record / locked
The current repository reports an approximately 0.97 Sharpe ratio, 1.30 Sortino ratio, and −11% maximum drawdown for the production profile. These figures are part of the research record, not values generated by this tour.
0.97reported Sharpe
1.30reported Sortino
−11%reported maximum drawdown
02 / REJECTION
The result was narrower than the thesis
The forecast failed. The structure did not.
Direct predictive-return tests did not persist out of sample. That matters because a polished model can still be an unstable explanation of the past.
Complexity as proof
More components did not earn authority by themselves. Each layer had to justify a distinct risk-management role.
Why the rejection stays visible
A keeper is credible only when the claim it replaced remains on the record.
03 / SURVIVOR
What the evidence kept
Five mechanisms earned a place.
The survivor was not a single signal. It was a sequence of bounded decisions that could still function without pretending to know the next return.
01Inverse-volatility and equal-risk-contribution base
02Twelve-month time-series trend
03Volatility targeting at the sleeve level
04A deliberate base and trend blend
05A governed drawdown regime brake
Diversification premium
Risk allocation remained useful when precise return forecasts did not.
Trend and governed scaling
Exposure management survived as a more durable function than prediction.
04 / GOVERNANCE
From research artifact to operating rail
A result becomes real only when the full path closes.
BASE completed the governed route from market data to reconciled positions. Research logic alone is not enough. The system must know what it intended, what it ordered, what actually filled, and whether the resulting position agrees with the record.
DATA
Acquire and validate the observable inputs.
SIGNALS
Translate the surviving mechanisms into targets.
TARGET RISK
Scale sleeves and apply the regime brake.
ORDERS
Convert intended exposures into instructions.
FILLS
Record what the market actually executed.
RECONCILE
Compare positions, orders, and system state.
05 / CURRENT ROLE
Where BASE sits now
The survivor became the spine.
BASE is the governed live rail around which the wider Corpus is organized. Other sleeves can remain in validation, research, concept, or null status without being dressed up as production systems.
The research record remains fixed. The Labs scenario explorer lets you supply hypothetical base and trend outcomes, change their blend, and apply a brake. Its arithmetic explains the mechanism; it does not rewrite the repository result.
The interaction contract
This is the research record. That is a user-controlled scenario. Neither becomes a new claim without returning through the full research process.
