Skip to content

Summary

Diversification is the precondition that makes TOMIC's 2% sizing rule workable: no single trade, sector, or event can reach the account. But diversification measured with average correlations is fragile — correlations rise in adverse markets, so the diversification benefit shrinks precisely in the scenarios risk models are meant to protect against.3 This page distills TOMIC's diversification guidance and the correlation failure modes.

TOMIC Diversification Rules

TOMIC diversifies along three independent axes:1 2

Axis Rule Rationale
Underlying / sector ≥ 5 sectors; no sector > 20–25% of capital Blocks sector-specific events (oil spill, single-stock blowup)
Strategy Mix of theta-positive structures (condors, flies, calendars, ratios) Different structures fail differently in the same move
Time / expiration Staggered expirations, trades < 90 days Avoids all risk rolling off (or expiring) on the same date

TOMIC's illustrative book (SPX, RUT, NDX, ETFs, and five equities across sectors) shows the intent: the index short-vol sleeve is deliberately complemented by underlyings with different drivers, and explicit tail insurance ("units" — cheap OTM puts / VIX calls, 5–10% of capital) covers the systemic layer that diversification cannot.1

Addressable Risk Layers

TOMIC separates risk into layers with different mitigations:1

Risk Example Primary mitigation
Systemic Financial-system collapse (Lehman 2008) Hard assets / tail hedges — not diversification
Market Macro shocks hitting everything (2008, 2020) OTM puts, VIX calls ("units"); cash as a position
Sector Industry events (BP spill) Sector caps, per-sector options
Company Firm-specific (Enron) Underlying diversification

Correlation Fails in Crashes

The VaR mathematics makes the failure precise. The two-asset formula σ(X+Y) = √(σX² + σY² + 2ρσXσY) shows the diversification benefit shrinking monotonically as ρ rises; on Hull's four-index portfolio, switching from average to EWMA-updated inputs on September 25, 2008 more than doubled 1-day 99% VaR ($217,757 → $471,025) because volatilities and correlations were far above their 500-day averages, and "correlations rise in adverse markets."3

Practical consequences for an options book:

  • Stressed correlations must be scenario-tested: Hull's 2008 four-index example shows stressed-window inputs more than doubling VaR versus averages — re-estimate correlations under stressed windows before trusting the diversification math in a drawdown3.
  • Short premium behaves as one factor in stress: the short-vol book's losses concentrate when vol spikes and correlations rise together, so treat the strategy axis — not underlying count — as the diversification frontier1.
  • Drawdown depends on structure, not count: defined- versus undefined-risk structures and position sizing determine how a stress scenario hits the book1.
  • Dynamic hedges assume continuity: portfolio insurance via delta replication performed poorly in 1987 because gaps and illiquidity broke the replication assumptions4 — correlation-based protection fails for the same reason at the same moments.

Operating Discipline

  • Treat 20–25% per sector as a hard cap enforced at entry, not a target average.1
  • Stress the book with correlations forced to 1 (or to 2008/2020 historical worsts) before adding correlated size — a home-grown stress test complementing the VaR layer (see VaR and Margin).
  • Hold the "units" tail sleeve regardless of diversification breadth; it hedges the market-wide tail risk that diversification cannot cover — though systemic counterparty/system risk additionally calls for the hard-asset and cash mitigants in the risk-layer table above.1
  • Cash is a position: skipping marginal trades in elevated-correlation regimes is itself a diversification decision.1

Links

Source Notes

  1. TOMIC bundle, topic Risk Management. ↩↩↩↩↩↩↩↩

  2. TOMIC bundle, topic Trading Plan. ↩

  3. Hull bundle, topic Value at Risk (EWMA example, correlation-in-adverse-markets result). ↩↩↩

  4. Natenberg bundle, topic Hedging with Options (portfolio insurance and the 1987 crash). ↩