The Strategy Health Card surfaces the four highest-severity auto-warnings on its header. There are more warnings under the hood — every sub-panel can raise its own — but only the most critical four are pinned. This article documents the full warning catalog, the threshold rules that trigger each one, and what each warning is really telling you.
Severity scale
- Critical: Strategy has a structural flaw that prevents trustworthy deployment. Action: redesign.
- Warning: Strategy has a measurable weakness in a specific dimension. Action: targeted fix or risk overlay.
- Info: Strategy has a feature worth noting but not actionable on its own. Action: context for downstream decisions.
Factor warnings
Critical: "Factor half-life shorter than rebalance cadence"
Raised when the median per-factor half-life (see Half-Life Derivation) is shorter than the configured rebalance frequency. Translation: the factor edge dies between rebalances, so the captured return is a fraction of the available alpha. Fix: shorten rebalance frequency or change to a slower-decaying factor.
Warning: "IC near zero (|IC| < 0.02)"
Raised when the cross-sectional Spearman correlation between factor scores and realised forward returns is statistically indistinguishable from zero. The factor has no signal. Common cause: poorly-defined factor metric, or a factor that worked in academia on US large-caps but doesn't survive in your universe.
Warning: "Negative average IC"
Worse than zero: the factor is inverted in this sample. Either the factor is genuinely reversed (rare but happens — see Frazzini & Pedersen 2014 on the low-vol anomaly relative to CAPM beta), or there is a sign error in the score definition.
Info: "Specific return > 50%"
More than half the strategy's return is unexplained by the named factors. This is not a problem per se — stock picking can be skill — but it means the strategy is not truly a "factor" strategy. Worth disclosing in any narrative.
Risk warnings
Critical: "Concentration HHI > 0.20 (effective N < 5)"
The effective number of holdings — the reciprocal of HHI (Hirschman 1964) — is below 5. The portfolio's returns are dominated by a handful of names; what looks like a diversified strategy is actually a small-portfolio bet. See Concentration HHI.
Critical: "Positive Sharpe in only 1 of N regimes"
The macro regime classifier (VIX + term spread + trend, see Macro Regime Classification) divides the backtest period into 2–4 regimes. If positive Sharpe shows up in just one, the strategy is regime-dependent. Acceptable if the regime is common and you have a regime detector live; otherwise a critical structural issue.
Warning: "Max drawdown / annual volatility > 3"
The ratio is the "Pain Index" proxy: a strategy with high MDD relative to its vol has experienced a tail event the volatility number alone hides.
Warning: "Sortino < 0.5"
Downside-only volatility (Sortino & van der Meer 1991) is large relative to mean return. Even if Sharpe looks acceptable, the downside path is rough.
Cost warnings
Critical: "Turnover > 200% annually with bps drag > 200"
Excessive turnover combined with realistic spread + impact assumptions (see Transaction Cost) wipes out the gross edge. Common cause: rebalance frequency higher than the factor's half-life justifies.
Warning: "Tax drag > 30% of gross return"
Most realised gains are short-term (under 1 year). For a taxable account this is a structural issue: the after-tax CAGR drops materially. See Tax Drag for the math.
Warning: "Capacity ceiling < $1M at 10% participation"
Average daily volume on the universe is too low to absorb meaningful capital without moving prices. See Capacity & Liquidity.
Sub-panel-only warnings
- Style Box drift > 30% across cells — strategy is style-rotating, which may be intentional or accidental.
- Counterfactual Sharpe range > 0.7 — strategy is parameter-fragile.
- Single drawdown episode contains > 70% of total drawdown — the published max drawdown is dominated by one event.
- Risk Decomp R² < 0.30 — factor model explains little variance; specific return dominates.
How thresholds are chosen
Each threshold is calibrated against two anchors: (1) statistical significance where applicable (e.g., IC thresholds use a 2-sigma rule given typical sample sizes), and (2) practitioner norms documented in Grinold & Kahn (1999) and Pardo (2008). The thresholds are not magic — they are conservative defaults. A serious deployment process customises them to the firm's risk appetite, capital base, and tax status.
A common anti-pattern is to dismiss warnings as too conservative and proceed. The warnings encode lessons from documented strategy failures; ignoring them is taking on a risk the platform has explicitly priced for you. The right response is either to understand why the threshold doesn't apply in your case (e.g., tax warnings don't apply in a tax-deferred account) or to act on the warning.
Further Reading
Foundational papers
- Frazzini, A. & Pedersen, L. H. (2014). Betting Against Beta. Journal of Financial Economics, 111(1), 1–25.
- Bailey, D. H., Borwein, J. M., López de Prado, M. & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the AMS, 61(5), 458–471.
Textbook references
- Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). Wiley.
- Grinold, R. C. & Kahn, R. N. (1999). Active Portfolio Management: A Quantitative Approach for Producing Superior Returns and Controlling Risk (2nd ed.). McGraw-Hill.
Related QuanterLab articles
Try it in QuanterLab
When a critical warning fires, do not silence it — address the underlying structural issue. Critical warnings exist because the corresponding pattern has caused documented strategy failures.