Stop losses protect single trades, not portfolios. Robust algo risk management layers position sizing, exposure caps, volatility throttles, drawdown rules, liquidity-aware execution, options hedges, and regime adaptation—monitored through VaR/ES, max drawdown, and risk dashboards. The goal is consistent risk per unit of return, not merely fewer losing trades.
Why this matters
Indian markets can flip from calm to chaotic around RBI policy, earnings clusters, or index rebalances. For algos trading NIFTY, Bank Nifty, or liquid stock futures, the difference between surviving and scaling is a portfolio risk framework—not just a stop-loss on each leg.
Where stop losses fall short
Stop losses help cap price risk on a position. They don’t address:
- Correlation spikes (everything sells off together).
- Gap risk (opens beyond your stop).
- Liquidity & slippage (especially during F&O ban lists or circuit moves).
- Model risk (your edge decays, regime shifts).
- Concentration & leverage (too much in one sector/idea).
Takeaway: Treat stop losses as a last line of defence, not the first.
1) Position sizing: your first risk control
Volatility targeting (per instrument):
Position size = Target Daily Vol (₹)Forecast Daily Vol (₹) per contract\dfrac{\text{Target Daily Vol (₹)}}{\text{Forecast Daily Vol (₹) per contract}}
- Forecast vol from EWMA/ATR.
- Example (NIFTY futures): If you target ₹50,000 daily vol and one lot’s expected daily move is ₹10,000, hold 5 lots. When vol doubles, size halves—risk stays stable.
Fractional Kelly for algos:
Full Kelly f∗=μ/σ2f^* = \mu / \sigma^2 (µ: expected excess return per period; σ²: variance).
Use ½-Kelly or less to reduce drawdown sensitivity and estimation error.
Risk parity (portfolio level):
Allocate so each sleeve contributes similar ex-ante vol or marginal VaR. This naturally reduces concentration in high-beta or high-vol names.
2) Exposure, leverage & concentration limits
Set hard caps that the engine enforces pre-trade:
- Gross/Net Exposure: e.g., Gross ≤ 250% of equity; Net between −50% and +150%.
- By Instrument/Sector: e.g., single stock ≤ 10% of gross; sector ≤ 30% of gross.
- Leverage & Margin: Respect broker/RMS and SEBI peak-margin constraints; keep a free-cash buffer for spikes in SPAN/Exposure margins.
- Event caps: On RBI policy, major index rebalances, or budget day, cut gross exposure (e.g., 50%) or switch to time-based exits.
3) Dynamic overlays: volatility & drawdown controls
Volatility throttle (intra-day): If realized or implied vol exceeds a regime threshold, scale down new orders or reduce participation rate.
Equity-curve stop (account level):
- Max DD stop: If peak-to-trough drawdown hits, say, −12%, flatten risk by 50–100%.
- Cooling-off: Trade with half risk for N sessions; re-enable only after recovery to a moving-average of equity curve.
Time stops: Close positions that fail to realize expected edge by a maximum bar count (e.g., 20–40 bars for intraday).
4) Liquidity & execution risk controls
- Participation caps (POV): e.g., never exceed 5–10% of market volume in mid-caps; 2–4% around opens/closes.
- Slippage budget: If realized slippage exceeds X bps vs model, de-risk or halt the strategy.
- Order shelf-life: Cancel unfilled limits after N seconds; avoid chasing during sweeps.
- Venue & order type sanity: Prefer passive/pegged near fair value; avoid aggressive sweeps during auction or low-depth windows.
- Pre-trade checks: Reject orders if symbol in NSE F&O ban list, if upper/lower circuits are near, or if margin headroom < buffer.
5) Hedging smartly (when it pays)
- Protective puts: Buy OTM puts on NIFTY/Bank Nifty against long-bias books.
- Collars: Finance puts with OTM call sales (mind assignment).
- Calendar hedges: If intraday algos suffer only on “gap” days, hold overnight index puts or a small long-VIX proxy when available.
- Cost control: Define a hedge budget (e.g., 1–2% annualized drag). If realized tail risk > budget, retain hedges; else, taper.
Measuring risk: simple, consistent, decision-useful
Track these at strategy and portfolio levels:
Core ratios
- Sharpe =E[Rp−Rf]σp= \frac{E[R_p – R_f]}{\sigma_p}
- Sortino =E[Rp−Rf]σdown= \frac{E[R_p – R_f]}{\sigma_{\text{down}}}
- MAR =CAGRMax DD= \frac{\text{CAGR}}{\text{Max DD}}
Drawdown math
- Max Drawdown (MDD): Largest peak-to-trough fall.
- Ulcer Index: RMS of drawdowns; penalises depth & duration.
Tail risk
- Parametric VaR (95%) ≈1.65×σdaily×t\approx 1.65 \times \sigma_{\text{daily}} \times \sqrt{t} (₹ terms with current equity).
- Expected Shortfall (ES, 95%) ≈ ϕ(z)1−0.95×σ\frac{\phi(z)}{1-0.95} \times \sigma under normality; prefer historical ES in practice.
Risk of Ruin (heuristic)
If win-prob pp, win/loss ratio bb, and fraction risked ff:
Large ff + low edge → ruin rises sharply. Keep ff modest (fractional Kelly), target stable vol, and cap drawdowns.
Monitoring
- Heatmaps: Monthly P&L and risk by symbol/strategy.
- Attribution: P&L bridges (alpha vs carry vs beta).
- Alerts: Slippage > budget, DD > threshold, turnover spikes, correlation breaks.
Regimes & model risk: adapt, don’t overfit
- Detect regimes: Cluster on realized vol, trend strength, liquidity, and breadth.
- Switch playbooks: Mean-reversion size down in high-vol trend regimes; trend models cut risk in choppy, low-breadth phases.
- Walk-forward & OOS: Retrain only on scheduled cadence; use ensemble models to avoid single-point failure.
- Kill-switches: Disable a sleeve if live hit ratio or edge drifts beyond statistically expected bands.
Implementation blueprint (checklists you can lift)
Pre-trade
- Margin headroom ≥ buffer (e.g., 20%).
- Symbol not in F&O ban; no near-circuit prints.
- Sizing uses current forecast vol; limits respected (gross, net, sector, instrument).
- Events filter (RBI, big earnings) applied.
In-trade
- Participation ≤ cap; slippage within budget.
- Volatility throttle active; time stop armed.
- Equity-curve guard (intra-day DD) monitored.
Post-trade
- Attribution & drift diagnostics.
- Risk dashboard update: VaR/ES, DD, exposures.
- Parameter decay review; regime labels refreshed.
Mini case study (India-centric)
An intraday trend-follower on NIFTY & Bank Nifty futures suffered −18% MDD during volatile policy days. We added: (i) vol targeting (ATR-based), (ii) event caps (50% gross on RBI days), (iii) equity-curve stop at −10% with a 5-day cool-off, and (iv) slippage budget halt.
Result (next 9 months): MDD reduced to −10.5%, volatility fell ~30%, CAGR dipped slightly but MAR improved meaningfully. The system scaled safely with higher notional.
Quick reference table
| Control | What it mitigates | How to set it (starter rule) |
| Volatility Targeting | Inconsistent risk, regime shifts | Fix daily ₹ risk; size = target / forecast vol |
| Gross/Net Caps | Leverage shocks | Gross ≤ 2.5×, Net in −0.5× to +1.5× |
| Concentration Limits | Single-name/sector shocks | Name ≤10% gross; sector ≤30% |
| Equity-Curve DD Stop | Strategy breakdown | Cut risk 50–100% at −10–15% DD |
| Participation Limits | Slippage/liquidity | ≤2–5% volume in index; ≤1–3% in mid-caps |
| Time Stops | Drift/model error | Exit after N bars if no progress |
| Options Hedges | Gap/tail risk | Protective puts; collar within 1–2% annual cost |
| Kill-Switch | Model failure | Halt if live metrics breach control bands |
FAQs
1) Are time stops better than price stops?
They address opportunity cost and model drift; use both. If price never reaches the stop but the thesis fails to play out in time, exit.
2) What’s a sensible volatility target to start with?
Back-solve from tolerance: if daily P&L swings of ±0.7% of equity are acceptable, set target daily vol ≈ 0.7% and size positions to that.
3) Should I always hedge with options?
Hedge when gap/tail risk dominates your loss distribution. Treat hedges as insurance with a budget; don’t expect them to be profit centers.
4) How do I react when vol doubles overnight?
Your size should halve automatically (vol targeting). If liquidity thins or spreads widen, tighten participation caps and widen price bands.
Conclusion
Great algos don’t chase every tick; they stabilize risk through sizing, limits, overlays, and monitoring. In Indian markets—where events, liquidity, and execution quality vary—robust risk engineering is your durable edge. Build the stack, automate the checks, and promote capital longevity over trade-by-trade heroics.