Research reports

REGIME REVERSAL STUDY

Where The Signal Layer Looks Most Structured

The current evidence is strongest around reversal states: extreme zero-gamma distance, high wall-bias strength, and dealer amplification. This is the most useful product story so far because it frames Contessa as a market-structure intelligence layer.

AssetSPY
Interval5 minutes
Complete days168
Rules tested272

FAMILY 01

Zero-Gamma Distance Reversion

The strongest family trades mean reversion when spot is far from estimated zero gamma, especially inside negative-gamma or amplifying regimes. The top-quartile negative-gamma variant produced +1,004 bps net at 1 bp slippage with a 51.8% directional hit rate.

FAMILY 02

Wall-Bias Reversal

Extreme wall-bias readings also behaved more like reversal pressure than direct trend prediction. Top-strength wall-bias fade rules were among the most stable in the holdout sample, but many still underperformed active-window buy-and-hold.

ROBUST RANKING

Rules That Stayed Positive In The Chronological Holdout

The ranking below uses the lower of full-sample net return and holdout net return as a simple robustness screen. This is intentionally conservative compared with sorting only by headline PnL.

RuleNetHoldout netHoldout excessSharpeMax DDTrades
Zero-gamma reversion, top quartile gap, negative gamma+1,004 bps+467 bps-41 bps5.16-394 bps2,555
Zero-gamma reversion, top 5 percent gap+561 bps+438 bps+11 bps10.06-153 bps647
Zero-gamma reversion, top 5 percent gap, negative gamma+549 bps+426 bps-19 bps10.06-153 bps624
Zero-gamma reversion, top half gap, negative gamma+485 bps+415 bps-61 bps1.70-471 bps4,395
Wall-bias reversal, top decile strength, high volatility+651 bps+399 bps-44 bps7.85-229 bps1,039
Wall-bias reversal, top quartile strength, amplifying regime+912 bps+383 bps-42 bps4.78-350 bps2,554

PRODUCT INTERPRETATION

The sellable claim is structure, not a black-box strategy.

The current evidence supports saying that Contessa transforms raw options data into regime variables that can be researched, filtered, and integrated into trading or risk systems. The next proof step is expanding the full-year window and validating these families out of sample.