A Framework for Classifying Future Strategies: Engines, Orthogonal Alpha, Hedges, and Opportunistic Trades
Part 8 below talks about my framework for evaluating new strategies and models
This is part 8 of my series — Building & Scaling Algorithmic Trading Strategies
As I’ve been building out the long–short allocator and experimenting with potential add-on sleeves, one thing has become clear:
Not all strategies belong in the same category, and they absolutely cannot share the same sizing logic.
Some strategies compound.
Some hedge.
Some spike.
Some do nothing for months and then deliver a 10× hit.
Some only matter during macro events or flow seizures.
Some are core — others are satellites.
I think it is easy to get carried away as you design new strategies and treat everything like an “alpha strategy” or everything like a “hedge” when many are just ideas at best.
That leads to overfitting, unnecessary complexity, and misaligned sizing. Not to mention a lot of wasted time and energy.
So I’ve tried to formalize a framework to classify every future strategy I test. The volatility sleeve ended up being the perfect test case to force this on paper.
1. The Catalyst: What the Vol Sleeve Taught Me
I initially assumed the volatility sleeve would behave like a hedge.
But after aligning 1,281 days of data, the reality was:
ROI 1,016.6%
CAGR 60.8%
Sharpe 0.70
Max DD –28.0%
Correlation to allocator: ~0.02
Beta: ~0.10
And under stress?
Dual negative days → vol sleeve mildly positive
Dual worst 5% days → vol sleeve negative
Dual worst DD window → vol sleeve barely positive
That means:
The vol sleeve is not a hedge.
It’s an orthogonal alpha engine with its own volatility.
And that is a different category — with different sizing rules.
This insight led me to formalize a complete classification system for future sleeves.
2. The Four Buckets All Strategies Must Fit Into
So I have decided that every future strategy — macro, vol, equities, RV, commodities, flow-based, machine learning, whatever — has to be bucketed before I size it or blend it.
The buckets are:
Core Combinatorial Engines
Orthogonal Alpha Engines
True Hedges
Opportunistic / Regime-Specific Plays
Here is how I decide where something belongs.
3. Bucket 1: Core Combinatorial Engines (The “True Alphas”)
These reinforce or extend the allocator.
They share the same directional DNA: trend, flow, velocity, acceleration, Sharpe gating, etc.
Characteristics
High correlation to the allocator
Improve compounding directly
Smooth, stable return profiles
Respond well to dynamic sizing
Add consistency, not convexity
Examples
Additional long/short factors
Trend engines on related assets
Momentum overlays
Short-term mean reversion tied to allocator signals
ML-based signal refinements
Sizing
Dynamic
Daily
Signal-driven
Volatility-aware
These are the “engines” that actually grow capital.
4. Bucket 2: Orthogonal Alpha Engines (Independent Convexity)
This is where the volatility sleeve belongs.
These strategies have their own cycles and own signatures. They don’t hedge your core strategy; they simply make money (or lose money) independently.
Characteristics
Correlation ~0
Sometimes positive, sometimes negative beta
High convex upside
Sometimes negative in left tails
Path-dependent
Do NOT hedge core drawdowns
Dynamic sizing usually makes things worse
Examples
VIX term-structure spreads
Commodity carry/momentum hybrids
ETF decay arbitrage
Rates/credit dislocations
Cross-asset convergence trades
Sizing
Small, stable
Fixed allocation
Updated monthly or per regime
Hard caps only
These are alpha satellites, not hedges. They help the portfolio by being different — not protective.
5. Bucket 3: True Hedges (Rare and Expensive)
Very few strategies qualify as true hedges.
For something to count, it must:
Pay off sharply in allocator drawdowns
Maintain strongly negative correlation in stress
Reduce portfolio max drawdown
Have convex crisis upside
Examples
Deep OTM equity crash puts
Direct long-VIX positions (not spreads)
Flight-to-safety duration hedges
Tail hedge funds (if you can afford bleed)
Characteristics
Lose small amounts consistently
Win big rarely
Reliable left-tail behavior
High carrying cost
Sizing
Tiny
Static
Never dynamically upsized
Treated like insurance
A hedge is not meant to “make money”—it’s meant to save the system from itself during multi-sigma events, absorbing the kind of tail risk that would otherwise overwhelm the portfolio based on its chosen risk profile.
6. Bucket 4: Opportunistic / Regime-Specific Trades
These strategies activate only under conditions:
scheduled (CPI, FOMC, earnings)
structural (vol crush windows, end-of-quarter flows)
seasonal (turn-of-month, holidays)
macro (liquidity cycles, rate pivots)
Characteristics
No cost when inactive
High edge when active
Time-bound or event-bound
Can be large when “on”
Zero when “off”
Examples
Delta-hedged CPI straddles
Earnings vol crush trades
Quarterly rebalance flows
Macro-data volatility spikes
Sizing
Binary (on/off)
Opportunistic
Heavily rule-based
No daily drift sizing
These trades don’t go in the core book; they are calendar-based scalps.
7. The Framework: How I Decide Where a New Strategy Goes
Whenever I test a new idea, I run it through these questions:
Correlation
Is it correlated with the allocator?
→ If yes → Core Engine
→ If ~0 → Orthogonal Alpha
Tail Behavior
Does it help in allocator drawdowns?
→ If strong → Hedge
→ If weak → Orthogonal Alpha
Convexity
Does it generate bursts or smooth returns?
→ Bursts → Orthogonal
→ Smooth → Core Engine
Activation Pattern
Always on?
→ Core or OrthogonalOnly active in scheduled windows?
→ Opportunistic
Sizing Sensitivity
Does dynamic sizing improve it?
→ Core engineDoes dynamic sizing break it?
→ Orthogonal alphaDoes scaling destroy optionality?
→ Hedge
Cost Profile
High bleed → Hedge
Low bleed → Alpha (core or ortho)
Zero bleed → Opportunistic
This prevents me from forcing a strategy into the wrong mental model.
┌──────────────────────────┐
│ NEW STRATEGY IDEA │
└────────────┬─────────────┘
▼
Is it correlated with the allocator?
(ρ or β ≳ 0.3)
│
┌──────────────┴──────────────┐
│ │
▼ ▼
YES → CORE COMBINATORIAL ENGINE NO → Does it help in
• trend / momentum allocator left tails?
• flow / velocity logic • crisis convexity?
• smooth, additive PnL • DD protection?
• responds to dynamic sizing
│
▼
┌────────────┴────────────┐
│ │
▼ ▼
WEAK/NO → ORTHOGONAL ALPHA ENGINE STRONG → TRUE HEDGE
• correlation ~0 • reliable negative
• independent convexity correlation in stress
• path-dependent bursts • convex payout in tails
• dynamic sizing harmful • high carry cost
• examples: vol TS, RV, decay • static, tiny sizing
│
▼
Only active in specific
events or regimes?
(CPI, FOMC, rebalance flows,
seasonality, liquidity shifts)
│
┌─────────────┴─────────────┐
│ │
▼ ▼
YES → OPPORTUNISTIC / NO → Remains in
REGIME-SPECIFIC PLAY assigned bucket
• event-driven scalps
• scheduled vol plays
• macro/liquidity windows
• zero bleed when inactive8. Applying the Framework: Vol Sleeve = Orthogonal Alpha
Based on the full 1,281-day analysis:
Correlation: ~0
Mixed behavior in allocator drawdowns
Strong convex bursts
Highly path-dependent
Dynamic sizing degraded performance
Therefore:
The volatility sleeve is Orthogonal Alpha, not a Hedge.
So I size it:
small
stable
capped
reviewed periodically
independent of allocator signals
This keeps the system clean and stops me from overfitting something that shouldn’t be micromanaged.
9. Closing: The Portfolio Is Not One Strategy — It’s an Ecosystem
Going forward, every new sleeve will be classified before I even test performance:
If it reinforces the allocator → Engine
If it diversifies it → Orthogonal Alpha
If it protects left tails → Hedge
If it activates under specific conditions → Opportunistic
That’s the operating system for the future of my algorithmic trading platform.
Next, I’ll write about how I detect volatility regimes (floor → compression → breakout → panic → decay) and how those regime tags influence which sleeves stay on, which stay small, and which activate.
The information presented in Math & Markets is not investment or financial advice and should not be construed as such.