Quantitative Credit Strategies for Hedge Funds: A Practitioner's Guide to CDS, Capital Structure Arb, and Credit Factor Models in 2026

Credit is not rates with a spread. The return distribution is fundamentally different — spread × duration captures carry, but the jump-to-default component generates a discrete, unhedgeable loss that has no analogue in rates or equity. Running a systematic credit book requires a quant framework built from scratch for non-linear, fat-left-tailed return distributions, not an equity or rates framework with credit spreads substituted in. This guide is for credit PMs at multi-strat funds and dedicated credit funds who already trade CDS and want opinionated mechanics on what works systematically in 2026. Quantitative trading software for hedge funds running credit books must handle CDS curve analytics, default probability models, and instrument-specific liquidity tiers natively — equity frameworks applied to credit generate plausible backtests and then fail when spreads gap on a credit event.

Why Credit Is a Distinct Quant Problem

Credit total return = spread × duration + default jump + price appreciation. The default jump is the problem: a 500 bps CDS position on a name that files Chapter 11 generates a loss of (1 − recovery rate) × notional, typically 60–70% of face value, regardless of the spread carry earned up to default. This non-linearity means the Sharpe ratio is a misleading summary statistic for credit — a strategy with Sharpe 1.2 that has a fat left tail and maximum drawdown of 40% in a default cycle is not comparable to an equity strategy with the same Sharpe. Sortino, conditional VaR, and maximum drawdown under a default stress scenario are the correct credit risk metrics.

The illiquidity premium in credit is real but path-dependent. HY CDS single-name bid/offer averages 30–60 bps vs. CDX.HY at 5–10 bps — that 25–50 bps spread reflects genuine compensation for illiquidity risk, but it evaporates in distress exactly when you need to exit. Post-2022 credit market structure has made this more acute: higher base rates (SOFR 4–5%) have widened IG/HY spread dispersion as leveraged issuers face meaningful refinancing risk, creating more single-name dispersion and more fundamental differentiation within rating buckets. Credit cycle timing differs from rates timing: the credit cycle lags the rate cycle by 12–24 months as refinancing risk accumulates, making 2024–2026 a structurally different environment for credit than 2019–2021. Fixed income quant strategies for institutional investors focused on rates duration are operating in a structurally different regime from credit — the instruments overlap (corporate bonds) but the systematic signals and risk frameworks are categorically distinct.

The Credit Instrument Universe

CDX.IG (investment grade, 125-name index, on-the-run series) and CDX.HY (high yield, 100-name index) are the flow credit instruments with institutional depth. CDX.IG Series 42+ notional outstanding $500B+; CDX.HY at $100–200B outstanding in the on-the-run series. Bid/offer on CDX.IG: 0.5–1.0 bps; CDX.HY: 5–10 bps. European equivalents: iTraxx Main (IG, 125 names) and iTraxx Crossover (sub-IG, 75 names) — Main at 0.5–1.5 bps bid/offer, Xover at 5–10 bps. These are the instruments for macro credit beta and systematic hedging.

Single-name CDS are the fundamental credit instruments. IG single-name contracts: $50–200M notional typical per trade; bid/offer 3–10 bps on liquid names (IBM, JPM, GS) widening to 20–50 bps on illiquid IG. HY single-name: $10–50M notional; bid/offer 20–80 bps on liquid HY, 50–150 bps on distressed names. CLO tranches (AAA down to equity) add structured credit exposure with non-linear waterfall mechanics; AAA CLOs trade at 150–175 bps over SOFR and provide diversified senior credit exposure with $25–100M ticket sizes. TRS on corporate bonds allows funded or unfunded credit exposure without the documentation overhead of CDS. The critical distinction: CDX is flow credit — you trade the index for beta, hedges, and macro views; single-name CDS is fundamental credit — you trade the individual name for fundamental views, basis trades, and capital structure arb. Algorithmic trading strategies for institutional investors in credit must model instrument-specific liquidity independently: CDX can absorb $500M+ blocks; single-name HY capacity caps at $10–20M before market impact is significant.

Five Core Systematic Strategies

1. CDS Basis Trades

The bond/CDS basis = CDS spread − (asset swap spread or Z-spread on the bond). Theoretical parity: the bond and CDS should price the same default risk. In practice, deviations arise from cheapest-to-deliver (CTD) optionality in the CDS (the protection buyer can deliver any eligible obligation, biasing CDS wider than bond), repo funding costs, and bond/CDS documentation differences. Negative basis (CDS spread < bond spread) creates a carry trade: buy the bond, buy CDS protection, earn the basis differential. Typical carry: 50–150 bps annualized when the negative basis is persistent. The trade is not riskless — 2008 and March 2020 saw negative basis blow out to 200–500 bps as funding dried up and mark-to-market losses on the bond leg overwhelmed carry. Basis trades require careful repo funding analysis (GC vs. specific repo, haircut changes in stress) and mark-to-market margin monitoring. For credit PMs running cleared CDS (CDX, single-name via clearing mandates), the repo and collateral mechanics connect directly to CCP initial margin optimization — including LCH PRISMA IM on cleared CDS, CDX.IG/HY SPAN inter-commodity credits, and collateral transformation cost modeling for CCP-eligible collateral. Risk management software for hedge funds running basis books must model repo funding risk as a first-class exposure — the funding leg failure in 2008 was the primary mechanism destroying basis trades, not the credit leg.

2. Capital Structure Arbitrage

Merton (1974) prices equity as a call option on firm value and debt as the residual. The empirical signal: implied vol from equity options + equity price → Merton-implied CDS spread. When actual CDS trades 80+ bps wide of Merton-implied spread (or vice versa), the capital structure is mispriced. The CDS-spread/equity-vol ratio — analogous to the VIX/credit spread relationship — flags when equity and credit markets are pricing materially different default probabilities. Stub equity trades (long equity + long CDS protection) on distressed names with low stub equity values capture recoveries where equity has optionality the CDS market has not priced. IC 0.10–0.20 in dislocated credit markets (2009, 2016, 2020); near-zero IC in tight spread regimes (<100 bps IG index). Regime filter is mandatory: when IG CDX < 70 bps, the capital structure arb signal degrades to noise. Machine learning in quantitative finance has improved Merton model calibration: gradient-boosted trees on equity vol surface features (term structure slope, skew, surface curvature) predict CDS spread deviations better than the classical at-the-money vol input by 15–25 bps RMSE.

3. CDX Index vs. Single-Name Dispersion

The CDX index spread should equal the weighted average of its constituent single-name spreads, adjusted for correlation and index convexity. When the index trades wide of intrinsic value (sum of weighted constituent spreads), sell index protection and buy single-name protection on the widest names (index > intrinsics trade). When index is tight to intrinsics, buy index protection and sell single-name protection on the tightest names (compression trade). CDX roll mechanics add a systematic signal: the on/off-the-run spread differential (typically 2–5 bps for CDX.IG) creates roll trades around the quarterly roll date when liquidity concentrates in the new on-the-run series. Capacity: $200M–1B on CDX.IG dispersion trades; $50–200M on CDX.HY. Portfolio optimization for institutional investors running credit factor models alongside index/single-name dispersion must control for the correlation between the two signals — dispersion trades have implicit factor exposure (overweight wide-spread names = value tilt) that can create unintended double-counting with explicit credit factor positions.

4. Credit Factor Models

The four credit factors with persistent out-of-sample evidence: value (OAS vs. model fair value from DCF + default probability model; IC 0.06–0.10 in IG, 0.10–0.15 in HY), momentum (3–6 month spread momentum; IC 0.06–0.12, with the caveat that credit momentum is driven by fundamental deterioration, not price anchoring as in equity momentum), quality (interest coverage ratio, net leverage, free cash flow margin; IC 0.08–0.14 in HY where quality differentiation is wider), and size (smaller issuers trade at 20–40 bps spread premium to comparably rated large issuers in HY). Factor orthogonalization to equity factors is non-trivial: credit value correlates 0.4–0.6 with equity HML; credit quality correlates 0.5–0.7 with equity profitability. Residualizing credit factors against equity factor exposures reduces correlation and improves the standalone Sharpe. Diversified credit factor portfolio (equal-weighted value/momentum/quality/size): Sharpe 0.7–1.1 net of execution costs. Factor investing for hedge funds in credit uses the same signal construction principles as equity factor models but with different economic intuitions — credit value is about default-adjusted carry, not book/price ratios; credit momentum reflects fundamental trajectory, not price momentum.

5. Volatility and Skew in Credit Options

CDX option vol surface provides tradeable credit vol signals. The swaption on CDX.IG 5Y: at-the-money implied vol typically 30–50% annualized; term structure is upward sloping (near-term event risk flattens it). Credit vol term structure mean-reverts faster than equity vol — event-driven spikes (earnings, ratings actions, macro data) create vol overshoots that revert in 5–15 trading days. In tight spread regimes (CDX.IG < 60 bps), CDX options implied vol trades 5–12 vol points above realized: selling tail risk (delta-hedged CDX payer spreads) generates positive carry with Sharpe 0.8–1.2 in range-bound regimes. When VIX > 25, credit vol typically trades cheap to fair value as flow demand for equity hedges overwhelms credit options demand — buy CDX payers as equity portfolio hedges when the credit/equity vol ratio is below the 20th percentile. Options volatility strategies for hedge funds that run vol books in equity and credit simultaneously must model the correlation structure of vol surfaces across asset classes — credit vol spikes in default cycles (2009, 2020) are not well predicted by equity vol alone and require credit-specific regime features as inputs.

Credit-Specific Risk Management

Jump-to-default is the dominant risk in single-name credit. A book of 5 single-name CDS positions with 2% notional each is highly concentrated — one default wipes (1 − 40% recovery) × 2% = 1.2% of AUM in a single event. At 50+ names with 0.5–1% notional per name, the default risk is diversified and the expected loss from one default is 0.3–0.6% of AUM — a manageable event rather than a book-threatening one. CS01 (spread DV01, sensitivity to a 1 bp parallel spread move) and DV01 (interest rate sensitivity) are the primary risk metrics for ongoing position monitoring. Curve risk — the 5yr/10yr CDS spread differential — requires explicit hedging for long-duration fundamental credit positions. Ratings migration risk (BBB→BB crossing the IG/HY divide drives forced selling from IG mandates) creates asymmetric spread widening that must be stress-tested separately from continuous spread movement scenarios.

CDX roll P&L is systematic and must be explicitly modeled: when the on-the-run series rolls quarterly, existing off-the-run positions must be rolled (transacting bid/offer twice) or maintained in the off-the-run series (accepting reduced liquidity). The roll cost for CDX.IG is 0.5–1.5 bps per quarter; for CDX.HY, 5–15 bps — meaningful for high-turnover systematic books. Mark-to- market vs. accrual accounting: CDS protection sold accrues premium daily; mark-to-market reflects spread movements. A book that manages to accrual and ignores mark-to-market is hiding spread widening risk that crystallizes at exit. Real-time market data infrastructure for credit desks requires CDS composite price feeds (Markit daily composites are the industry standard) plus bond TRACE data for basis monitoring — the two data feeds must be timestamped and synchronized for intraday basis monitoring to be meaningful.

Data and Signal Construction

Markit CDS composite data: daily composite quotes are the standard credit reference for most systematic books — the composite averages dealer quotes across 5+ dealers, reducing single-dealer noise. Real-time CDS quotes (via Bloomberg CDSW or direct dealer APIs) are required for basis monitoring and intraday strategies; expect 30–50 bps bid/offer on HY single-name real-time vs. 15–25 bps on the daily composite. TRACE bond reporting provides disseminated corporate bond transactions with a 15-minute reporting delay for most bonds (immediately for certain investment-grade bonds). That reporting delay is exploitable: TRACE data shows large block trades before composite quotes fully update, creating a 5–20 minute window where bond/CDS basis signals contain information. Alternative data strategies for institutional investors in credit include earnings call NLP: sentiment and language complexity scores from earnings transcripts predict CDS spread moves with IC 0.10–0.16 in HY — management tone degradation precedes spread widening by 2–4 weeks on average. Earnings surprise vs. CDS spread reaction: a 10% negative earnings surprise generates 5–20 bps CDS spread widening on HY names within 48 hours; the signal persists 3–6 weeks before fully incorporating. Point-in-time accounting data is required for fundamental factor construction — as-reported financial statements (not restated) with quarterly vintage timestamps prevent look-ahead in quality factor backtests.

Backtesting Pitfalls Specific to Credit

Survivorship bias in the CDS universe is severe: names that defaulted (Lehman, Washington Mutual, Caesars, Hertz, Revlon) are removed from today's CDS universe — any backtest using current CDS universe composition projected back 10+ years understates default frequency by 15–25% and overstates credit factor returns materially. Use as-of-date CDS universe membership data (Markit maintains historical index composition) and explicitly include defaulted names in factor model backtests. TRACE reporting delays must be modeled in backtests: incorporating a TRACE trade timestamp before the public dissemination time is look-ahead. Stale composite quotes on illiquid single-name CDS (names with fewer than 3 contributing dealers) can show zero spread moves for weeks — these staleness artifacts create spurious mean-reversion signals in factor models. How to backtest a quantitative trading strategy for credit requires all three: as-of-date index membership, TRACE publication-timestamp enforcement, and staleness filters on illiquid names — none of which are handled by generic backtesting platforms built for equities.

2008 and 2020 liquidity events are mandatory stress scenarios, not optional. The March 2020 CDX.IG spread move from 60 bps to 165 bps in 15 trading days — with bid/offer on single-name HY widening to 100–200 bps — is the liquidation scenario that dominates long-term credit strategy Sharpe ratios. Any credit strategy backtest that doesn't include execution costs calibrated to March 2020 conditions is producing academic results, not institutional results. CDX series roll artifacts (spread differences between on/off-the-run series in the backtest) must be explicitly accounted for; a backtest that treats CDX as a continuous series without roll adjustments overstates momentum signal returns by 10–20 bps per year. Statistical arbitrage strategies for hedge funds face analogous survivorship and liquidity backtesting pitfalls — the mechanics differ (equity delisting vs. CDS default removal) but the effect is the same: overstated historical returns that don't survive first contact with live markets.

AlphaEdge AI for Systematic Credit Books

The infrastructure gap for systematic credit is not signal ideas — it is the production data and analytics plumbing. CDS basis monitoring requires synchronized Markit composite data and TRACE bond transactions with publication-timestamp enforcement; CDX intrinsic value tracking requires as-of-date index composition and constituent spread aggregation; credit factor models require point-in-time accounting data and default-inclusive CDS universe construction. Each of these components is 3–6 months of engineering work before any signal logic runs.

AlphaEdge AI provides the full systematic credit stack out of the box: CDS basis monitor with Markit/TRACE integration and CTD-adjusted theoretical parity, CDX vs. intrinsic value tracker with daily index composition and roll schedule, credit factor model covering value (OAS vs. model fair), spread momentum, and quality (coverage, leverage, FCF margin), jump-to-default hedging module with CS01 position limits enforced by issuer and sector, and real-time spread monitoring with CDX on/off-the-run roll P&L attribution. The backtesting engine enforces as-of-date index membership, TRACE publication timestamps, and staleness filters on illiquid single-name composites — the three credit-specific pitfalls that corrupt generic backtesting platforms.

The surrounding infrastructure is covered across the platform: systematic global macro strategies for hedge funds that use credit spreads as macro signals (IG/HY spread levels as regime indicators for risk-on/risk-off), execution algorithms for institutional traders adapted for CDS liquidity (roll-window-aware scheduling, bid/offer impact modeling for single-name HY), and high-frequency trading infrastructure for desks that need real-time credit event monitoring and automated hedging triggers. Crypto quant strategies for institutional desks offer context on managing non-linear return distributions in a different asset class — the jump risk mechanics (sharp drawdowns on regime change) are structurally analogous to credit default risk even though the underlying instruments are entirely different. Commodity quant strategies for institutional investors running at multi-strat funds alongside credit books benefit from the same risk aggregation framework — the cross-asset correlation structure between credit spreads and commodity prices (energy credit in particular) requires integrated portfolio-level risk monitoring rather than siloed per-book VaR.