Speaker Q&A with Mayank Saxena, Hybrid Derivatives Trader Vice President at Société Générale.

As someone operating at the intersection of derivatives and systematic investing, how do you approach structuring multi-asset quant portfolios that achieve true diversification across equities, rates, FX, credit, and commodities, rather than simply remixing correlated risk premia?

As someone bridging derivatives and systematic investing, true multi-asset diversification in quant portfolios goes far beyond naive asset-class allocation (e.g., equities + rates + FX + credit + commodities). Most "multi-asset" approaches merely remix correlated macro bets - growth-sensitive equities/credit, inflation-hedge commodities, risk-on sentiment across most classes—leading to conditional correlation spikes when it hurts most. The 2022 regime of persistent inflation and rising rates exemplified this: equities, credit spreads, and duration all sold off together, erasing traditional 60/40 benefits as stock-bond correlations flipped positive.

The goal is orthogonal exposures: Return streams driven by distinct, weakly correlated risk premia that behave differently across regimes (growth/inflation, risk-on/off, liquidity/vol spikes). From a systematic perspective, this means harvesting persistent, well-researched premia cross-asset, rather than passive beta.

Key principles and approaches I favor:

a) Identify Orthogonal Premia Sources: Focus on factors with low/negative conditional correlations:

• Value: Cheap vs. Expensive assets, cross-asset via fundamentals like book-to-market or yield spreads.
• Carry: Higher-yielding assets, e.g., Short Volatility, bond roll-down, commodity futures basis.
• Momentum/Trend: Time-series or cross-sectional, often convex in crises via trend-following CTAs.
• Quality/Defensive: Strong balance sheets, low leverage and stable earnings.
• Low Volatility/Min-Vol: These extend robustly across equities, fixed income, FX, credit, and commodities with pairwise correlations often near zero or negative.

b) Cross-Asset Implementation via Derivatives: Use liquid derivatives to isolate premia efficiently:

• Futures/swaps for trend/carry in commodities/FX/rates (low cost, high capacity).
• Credit default swaps/indexes for credit value or carry.
• Equity index options for vol premia.

c) Portfolio Construction for Orthogonality:

• Risk Parity / Hierarchical Risk Parity: Allocate such that each premia contributes equal risk not capital.
• Multi-Style Composites: Combine premia in ensembles (e.g., 50% fundamental/value-carry, 30% technical/trend-momentum, 20% defensive/low-vol), targeting low portfolio correlations to benchmarks (~0.1-0.3).
• Regime-Adaptive Overlays: Monitor regime signals (inflation surprises, vol-of-vol, correlation matrices) to tilt weights—e.g., boost trend in high-vol/inflation regimes, scale down carry when term structures invert and market is fragile.
• Capacity & Liquidity Controls: Scale exposures inversely to liquidity (e.g., tighter ADV caps in commodities), with turnover penalties to preserve implementability.

In practice, such frameworks deliver a higher Sharpe vs. passive multi-asset with better drawdown control in 2022-like stressed scenarios. The key is to backtest with realistic costs, stress non-stationary correlations, and avoid overcrowding popular premia. True diversification isn't more assets—it's engineered orthogonality through systematic premia harvesting.

From a QIS trading perspective, what frameworks do you find most effective for combining heterogeneous alpha streams—such as trend, carry, value, volatility strategies, and machine-learning signals—into a single, capacity-aware portfolio without diluting your edge?

From a QIS trading perspective, combining heterogeneous alpha streams—trend-following (convex, low-turnover), carry (regime-persistent), value (mean-reverting), volatility premia (short/long), and machine-learning signals (adaptive but noisy/high-turnover)—into a single, capacity-aware portfolio requires frameworks that maximize risk-adjusted returns while preserving each stream's edge through low/negative correlations and explicit frictions.

I find these approaches most effective, drawing from MPT extensions, empirical quant research, and live multi-strat books:

a) Risk-Based & Hierarchical Risk Parity (HRP): Treat alphas as assets and allocate via risk parity (equal risk contribution) rather than equal notional weights. HRP enhances this by using hierarchical clustering to group similar streams, then applying weights to the components within each cluster. This yields superior diversification, especially with estimation error or regime shifts—often improving Sharpe by 15-30% over naive mean-variance in heterogeneous setups. Embed ADV caps and turnover penalties in the objective to scale positions realistically.

b) Integrated Optimization with Constraints: Mix optimized sub-portfolios, then apply mean-variance/CVaR optimization with hard/soft constraints: turnover limits, position caps (% ADV), and leverage bounds. Use HRP for initial diversification, then refine with performance weighting. These preserve edge better than unconstrained blending.

c) Stress-Testing & Robustness Layer: No framework is complete without rigorous stress-testing: Replay crises (GFC, 2020 COVID, 2022 inflation), amplify shocks (correlations +20-50%, vol 2x), and use Monte Carlo simulations with fat tails. Compute stressed VaR/CVaR, drawdowns, and capacity impacts when liquidity evaporates. Trigger regime-based reweights (e.g., boost trend/convexity in breakdowns). This ensures the portfolio remains scalable at higher AUMs without forced liquidation.

In practice, structural diversification, dynamic weighting, constraint-heavy optimizer and forward stress are key to building a robust and scalable strategy. Backtests with realistic costs often show net Sharpe gains of 0.3-0.6 over naive combos, with drawdowns reduced by 10-25% stress. Success depends on avoiding data-mined overfitting, and regime-adaptive tuning to sustain the long-term edge.

Given your expertise in derivatives, how do you determine when convexity, optionality, or tail-risk hedging should be embedded directly into the portfolio construction engine versus applied as an overlay? What specific signals or risk conditions typically trigger that decision?

From a derivatives and QIS perspective, the decision to embed convexity or optionality directly into the portfolio construction engine versus applying it as an overlay depends on four main considerations: (1) whether the convexity is structural to the strategy, (2) the persistence of the risk regime, (3) cost and capital efficiency, and (4) whether the risk being hedged is endogenous or exogenous to the portfolio.

You embed convexity when it is part of the strategy’s core return generation mechanics. Strategies that are structurally short gamma—carry, mean reversion, volatility selling, or anything pro cyclical—benefit from embedding convexity into the construction engine. In these cases, the convexity is not insurance but rather part of the design (e.g. using delta hedging/gamma scalping as a return engine). You also embed convexity when the portfolio has persistent negative crash beta or when alpha implicitly loads on vol, correlation, or liquidity factors.

Convexity is applied as an overlay when the objective is tactical protection against temporary or regime shift risks. These are environments where hedging cost is high, the timing is uncertain, and you want flexibility—e.g., vol of vol accelerating, cross asset correlation converging, skew steepening, or bid/ask widening. In these cases, the convexity is meant to be episodic rather than structural. VIX calls, downside equity puts, long gamma macro-overlays, etc. protects the portfolio during this temporary phase.

Typical triggers for overlay convexity include:

a) Cross asset correlation rising and dispersion collapsing

b) Vol of vol breaking out relative to spot vol

c) Stress in liquidity indicators or widening funding spreads

d) Crash beta becoming unstable

e) Macro uncertainty dominating micro dispersion

f) Risk off moves where hedging is cheap vs. realized volatility

The key distinction is simple: embed convexity when it is part of the portfolio’s structural alpha generation; use overlays when the purpose is tactical protection against tail risk or regime shift dynamics.

Considering that liquidity, turnover, and leverage constraints vary widely across asset classes, what practical controls or modeling techniques do you use to ensure real-world implementability, especially during volatility spikes or cross-asset correlation breakdowns?

In multi asset QIS portfolios, the main challenge is that liquidity, turnover, and leverage behave very differently across asset classes, and these differences become amplified during volatility spikes or correlation breakdowns. To maintain real world implementability, I rely on a combination of ex ante liquidity modelling, execution cost estimation, and dynamic constraint systems that adapt to market stress. There must be robust controls to withstand the stressed market when turnover spikes due to rebalancing, bid/ask spread widens, and leverage jumps coupled with correlation spikes. Covid, 2022 Inflation spike & Liberation Day to name a few examples of such scenarios. Outlining a few practical techniques adopted in real world below

a) Size positions using asset class specific liquidity metrics such as ADV/Open Interest and bid/ask volatility to ensure that a reasonable size can be executed without incurring slippage and without non-trivial market impact. For example, equities can handle 10% of ADV without significant slippage whereas for commodities this figure is close to 3%-5%.

b) Transaction costs are state-dependent. Hence, the backtest must include volatility scaled slippage models. During stressed regimes—Covid, 2022 inflation shock—bid/ask spreads and market impact expand non linearly. Modelling these costs prevents the backtest from overstating Sharpe.

c) Regime based rebalancing would help in allocating more weights to more liquid asset classes during that regime which would preserve cost. This reduces forced turnover exactly when liquidity is at its worst.

d) High turnover dilutes the performance. Applying a constraint of daily maximum turnover could smooth out the execution and would prevent excessive trading in a volatile market.

e) Dynamically volatility targeting would scale the leverage inversely to the realized volatility thereby deleveraging when the volatility spikes and re-leveraging in a low volatility environment benefiting from higher participation in the upside and lower participating in the downside.

f) Apply caps to the maximum leverage to protect the portfolio from extreme tail events

Overall, the goal is to design a portfolio that is self stabilizing under stress. Robust liquidity modelling, state dependent execution cost assumptions, adaptive leverage, and turnover controls ensure the QIS strategy remains implementable during real world market dislocation.

Finally, as market regimes shift rapidly, how do you monitor, stress-test, and adapt multi-asset quantitative portfolios in real time? Specifically, how do you respond when historical relationships, term structures, or volatility surfaces stop behaving as expected?

In multi-asset quantitative investment strategies (QIS), rapid regime shifts—from low-volatility/growth environments to high-volatility/inflationary regimes, market crashes (e.g., GFC), or liquidity crunches (e.g., Covid)—require a proactive, systematic framework for real-time monitoring, forward-looking stress-testing, and adaptive portfolio adjustments. This ensures heterogeneous alpha streams (trend, carry, value, volatility, ML signals) remain resilient, preserving edge without excessive whipsaw or turnover.

Stress-testing is non-negotiable for robust, scalable QIS. It quantifies tail risks and reveals vulnerabilities when historical relationships break down (e.g., equity-bond correlations flipping positive), term structures behave unexpectedly, or volatility surfaces deviate (e.g., skew steepening in calm markets or term vol inversions).

a) Stress-Testing and Scenario Analysis

Combine historical replays with hypothetical and live-data shocks to evaluate portfolio response under non-stationary conditions.

i. Replay crises: 2008 GFC, 2011 Euro debt, 2020 COVID, Liberation Day with amplified shocks—e.g., equity-bond correlations from -0.4 to +0.5-0.8, VIX spikes from 20 to 80, or liquidity evaporation (bid-ask widening 2-4x).

ii. Hypothetical scenarios: Monte Carlo simulations incorporating fat tails; stress term structures, vol surfaces (e.g., skew steepening/backwardation), spots, rates, and correlations.

iii. Dynamic integration: Bootstrap from recent data when surfaces/relationships deviate; decompose risks by factor/alpha stream and shock independently

iv. Key outputs: Conditional VaR (CVaR at 99%), expected shortfall, liquidity-adjusted drawdowns, and implementation shortfall estimates. Run full tests daily, trigger on breaches.

These reveal hidden vulnerabilities—e.g., short-vol strategies losing 15-25% more than baseline in inverted surfaces.

b) Systematic Adaptation Responses

Adaptations must be rule-based and automated to avoid discretionary panic. Trigger on monitoring/stress signals, with human oversight for extreme cases.

i. Volatility targeting: Dynamically scale gross/net exposure inversely to portfolio vol to preserve convexity while auto-deleveraging in spikes.

ii. Risk reallocation on correlation breakdowns: When correlations compress/spike (e.g., equities-bonds positive in inflationary regimes); deleverage correlated clusters (e.g., cut equity-credit 20-30%), boost decorrelated streams (trend, commodities).

iii. Term structure inversions/unexpected behavior: Shorten durations in fixed income, pivot carry/value signals to relative-value trades (e.g., curve steepeners via futures), overlay macro-ML predictors; use adaptive filters (e.g., Kalman) for parameter updates.

iv. Volatility surface deviations: (e.g., skew steepening, term vol inversion): Cap/reduce short-vol carry positions, add cheap convexity (OTM options or long-vol sleeves), recalibrate vol strategies with live implied data; retrain ML signals on recent windows (3-6 months).

v. Liquidity/turnover management: Introduce dynamic exposure caps (e.g., % of ADV, tightened 30-50% in stress), threshold-based rebalancing (only if deviations >5-10%), and higher turnover penalties.

In backtests/live performance (e.g., 2022 correlation flips), such frameworks often mitigate 10-25% of potential drawdowns by reallocating toward trend/long-vol while containing whipsaw.

This integrated, data-driven cycle—real-time monitoring, forward stress-testing, systematic adaptation—turns regime uncertainty into an opportunity, delivering more consistent risk-adjusted returns across heterogeneous QIS alphas.

Join Mayank on Day 1 of Future Alpha 2026 at 11:40 AM for an insightful Panel Discussion: Navigating the Quant Landscape: Alpha, Risk, and Convexity in a Volatile World. 

Secure your pass today!