Portfolio Value-at-Risk, three ways, with risk you can decompose
See your one-day VaR three ways — parametric, historical and Monte-Carlo — at 95% and 99%, plus CVaR, tail stats, benchmark capture, per-holding risk contribution and concentration flags, all computed from your actual positions.
US coverage included in your free 14-day trial · no credit card required
StoqPulse turns your live holdings into hard, deterministic risk numbers: 1-day VaR computed three ways — parametric, historical and Monte-Carlo — at 95% and 99%, CVaR (expected shortfall), non-normal tail stats, beta and capture versus SPY, annualised volatility, max drawdown, per-holding risk contribution, a correlation heatmap and concentration flags. US coverage is free for 14 days (no card), so you can study your tail risk in detail. Every figure is computed by pure, testable math — never invented by an AI.
VaR three ways: parametric, historical and Monte-Carlo
StoqPulse estimates one-day Value-at-Risk three complementary ways, so you are never anchored to a single model. Parametric (variance-covariance) VaR scales your portfolio's daily return volatility by the standard-normal critical value (z = 1.645 at 95%, z = 2.326 at 99%). Historical VaR reads the loss straight off your actual empirical return distribution at the same confidence levels — no normality assumption. Monte-Carlo VaR runs a seeded bootstrap resample of your real returns, so it is deterministic: the same portfolio renders identical numbers every render, never a random draw. Each method is shown at 95% and 99%, in dollars and as a percent of book, side by side — and where the three disagree is itself a signal of how fat-tailed your book is.
CVaR / expected shortfall, and the tail's true shape
VaR tells you the threshold; CVaR (expected shortfall) tells you how bad it gets past it — the average loss once you are in the worst 5% of days. StoqPulse reports 95% CVaR alongside each VaR method, so the gap between the cutoff and the tail average is obvious. To capture what a symmetric model misses, it also computes the return distribution's true shape: the annualised Sortino ratio (return per unit of downside deviation), skewness (is the tail to the left?) and excess kurtosis (how fat are the tails versus a normal curve), with plain-English fat-tailed or left-skewed flags. These non-normal stats need roughly three months of overlapping daily history to be meaningful.
Beta, drawdown, Sharpe — and benchmark capture vs SPY
Beyond tail risk, StoqPulse measures how your book moves with the market. Beta is computed against SPY with both return series aligned to the exact same trading dates, alongside annualised volatility (252 trading days), the Sharpe ratio against a 4% risk-free rate, and trailing max drawdown from your actual portfolio value path. A benchmark-relative block layers on top: tracking error and active return versus SPY, the information ratio, and up- and down-market capture — how much of the market's gains you captured on up days and how much of its losses you took on down days. Together they frame whether your returns are paying you for the risk you carry.
Which holdings actually drive your risk
Portfolio weight and risk share are not the same thing, so StoqPulse decomposes total risk holding by holding with a Euler decomposition of the covariance matrix. Each position shows its marginal VaR (how total VaR moves if you add to it), its component VaR (its share of the total — and the component VaRs reconcile back to the displayed figure), and its percentage of overall portfolio volatility, so a 6% position that is 20% of your risk is impossible to miss. A correlation heatmap over your largest holdings then shows where names move together, exposing the hidden overlap that concentration-by-weight alone can hide. All of it is factual research, never a recommendation to trade.
Concentration flags from your real weights
Tail math misses single-name and sector blow-ups, so StoqPulse adds deterministic concentration flags. Any holding whose weight exceeds your position limit, or any sector above its limit, is flagged with the exact percentage and the limit it breached. Two market-risk flags layer on top: beta above 1.30 (amplifies market moves) and annualised volatility above 30%. Limits are configurable, not hardcoded, so the flags reflect your own risk policy. Sector and top-position weight bars sit next to the flags for context.
Computed from your live holdings — numbers never from AI
Every figure comes from your actual positions priced on trailing daily history, aligned to the set of dates common to all holdings so the value series is internally consistent. The math lives in pure, unit-tested functions — returns, variance, covariance, beta, drawdown, all three VaR/CVaR methods, the tail stats, the capture ratios and the risk decomposition — so results are reproducible and auditable; even the Monte-Carlo path is seeded, so it is repeatable rather than random. A plain-English AI risk narrative (Pro tier) reads the output for you, but the AI never produces the numbers themselves. The quantitative analytics stay available on every plan.
Honest about the method
No single VaR model is the whole truth, so StoqPulse shows three and names their limits. Parametric VaR is fast and transparent but assumes roughly normal returns, so it can understate fat-tailed or skewed risk; historical and Monte-Carlo VaR read the real return distribution but are only as rich as the sample behind them. Parametric VaR appears with about a month of overlapping daily history (roughly 20 return points); the historical VaR, Monte-Carlo VaR and tail stats need roughly three months before they show. CVaR is reported at the 95% level. We name these limits up front so you read the numbers in the right context — a one-day estimate is a gauge of current exposure, not a crisis forecast.
Frequently asked questions
All three. StoqPulse computes parametric (variance-covariance) VaR, historical VaR from your actual empirical return distribution, and Monte-Carlo VaR from a seeded bootstrap resample of your real returns — each at 95% and 99%, side by side. Parametric appears with about a month of history; the historical and Monte-Carlo figures need roughly three months. Comparing the three is itself informative: a big gap points to fat tails a normal model would miss.
VaR is reported on a one-day horizon at 95% (z = 1.645) and 99% (z = 2.326) confidence, in dollars and as a percent of book, for all three methods. CVaR (expected shortfall) — the average loss across the worst 5% of days — is shown at the 95% level alongside each method, so you see both the cutoff and the tail beyond it.
Beta is measured against SPY as the market benchmark. StoqPulse aligns your portfolio returns and SPY returns to the exact same trading dates before computing covariance over variance, so the comparison is apples to apples. The same alignment powers a benchmark-relative block — tracking error, active return, information ratio and up/down-market capture versus SPY. A beta above 1.30 also triggers a market-risk flag, since it means your book amplifies market moves.
Yes. StoqPulse decomposes total portfolio risk holding by holding, reporting each position's marginal VaR, its component VaR (the component VaRs reconcile back to the displayed total) and its share of overall portfolio volatility — so a small weight that carries outsized risk stands out. A correlation heatmap over your largest holdings shows where names move together. It is factual research, not a recommendation to trade.
Any single holding whose weight exceeds your position limit, or any sector above its sector limit, is flagged with the exact percent and the breached limit. Two more flags cover market risk: beta above 1.30 and annualised volatility above 30%. The limits are configurable to match your own risk policy rather than fixed in code.
Yes — free for 14 days, no card. During your free trial the VaR methods, CVaR, tail stats, beta, capture ratios, volatility, drawdown, Sharpe, risk contribution and concentration flags are available for US portfolios at no cost. The optional AI risk narrative — a plain-English read of your exposures — is a Pro feature, but every quantitative number stays available on your plan. After the trial you choose a paid plan; see live pricing for current tiers.
Parametric VaR, volatility, Sharpe and beta need at least about a month of overlapping daily history (roughly 20 return points) across your holdings. Historical VaR, Monte-Carlo VaR and the tail stats (Sortino, skew, kurtosis) need a fuller sample — roughly three months. Max drawdown needs only a short value path. New positions with little shared history may show a dash until enough common trading days accumulate.
Explore more StoqPulse tools
Start free for 14 days — US coverage included
Build the discipline into your process. No credit card required.