Volatility Analysis
230.36
1.91
(0.84%)
09/04/2026
NVIDIA Corporation (NVDA)Showing 1 year of volatilty data. To view all data, Upgrade to PRO plan!
Summary
- NVDA is realising 44.4% annualised volatility over the last 21 sessions, the high end of its range — higher than 85% of readings over the past 1 year.
- Volatility has risen from 43.3% to 44.4% over the past month (+3% relative), so the near-term trend in risk is against short-volatility positioning.
- Close-to-close vol (44.4%) runs well above the Parkinson range estimate (26.6%). Because Parkinson only sees the intraday high/low, that gap is overnight jump risk — the move happens between the closing bell and the next open.
- 33% of total variance is delivered overnight (22.4% annualised) against 31.7% during the session — risk is mostly intraday and can be managed inside the session.
- A GARCH(1,1) fit puts next-session vol at 37.6% against a long-run anchor of 38.0%, with persistence of 0.505 — shocks decay with a half-life of about 1 trading days.
- On the empirical distribution, a 1-day 95% VaR is -3.96% with an expected shortfall beyond it of -4.76%.
- Absolute returns show no significant lag-1 autocorrelation (0.057), so volatility in this name is closer to unpredictable noise than to a persistent regime — treat the GARCH forecast with caution.
- Worst peak-to-trough drawdown over the window was -20.2% (trough 2026-03-30), with an Ulcer Index of 10.06 capturing how long it stayed underwater.
Realized Vol · 21d
44.4%
High
calm85th pctilestressed
Context · 1 year lookback
85%Percentile
1.12Z-score
38.3%Median vol
+3%1-mo change
79%Vol of vol
37.6%GARCH next
38.0%GARCH long-run
1dShock half-life
45.8%EWMA λ=.94
Historical Volatility Close-to-Close · 21-day rolling · annualised ×√252
Volatility Cone Close-to-Close · 1 year of overlapping windows
Every realized-vol reading the stock has produced at each horizon, as a percentile envelope. The marker is today. A dot riding the top of the cone says vol is stretched for that horizon — which is the comparison that matters, since short windows are naturally noisier than long ones.
| Horizon | Min | 10th | 25th | Median | 75th | 90th | Max | Current | Percentile | Obs |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 Week 5d | 12.3 | 20.9 | 27.8 | 39.6 | 49.6 | 59.9 | 82.5 | 34.0 | 40 | 247 |
| 2 Weeks 10d | 15.3 | 26.8 | 32.1 | 39.0 | 44.6 | 51.2 | 59.6 | 59.5 | 99 | 242 |
| 1 Month 21d | 25.0 | 30.3 | 34.7 | 38.3 | 42.9 | 45.1 | 47.5 | 44.4 | 85 | 231 |
| 2 Months 42d | 27.8 | 33.7 | 35.9 | 38.4 | 40.3 | 42.9 | 45.1 | 43.3 | 95 | 210 |
| 3 Months 63d | 32.0 | 34.4 | 35.6 | 37.3 | 40.1 | 41.5 | 44.3 | 40.6 | 79 | 189 |
| 6 Months 126d | 35.0 | 35.6 | 36.0 | 37.3 | 38.8 | 39.9 | 40.2 | 40.0 | 92 | 126 |
Realized Term Structure
Current realized vol at each horizon against the median for that same horizon. When the short end sits above the long end the curve is inverted — the market is pricing a near-term event, and the spread has historically closed by the short end falling rather than the long end rising.
Estimator Comparison
Five estimators of the same quantity. They disagree in informative ways: Parkinson and Garman-Klass read the intraday range and are blind to overnight gaps; Rogers-Satchell is drift-independent, so it does not inflate on a strong trend; Yang-Zhang combines overnight, open-to-close and Rogers-Satchell and is the one to quote against implied vol.
| Estimator | 1 Month | 3 Months | 6 Months | 1 Year | Reads |
|---|---|---|---|---|---|
| Close-to-Close | 44.4 | 40.6 | 40.0 | 38.0 | total risk, incl. gaps |
| Parkinson | 26.6 | 31.3 | 30.8 | 30.8 | intraday range only |
| Garman-Klass | 25.9 | 31.1 | 30.3 | 30.4 | range + open/close |
| Rogers-Satchell | 24.9 | 31.0 | 29.8 | 30.1 | drift-independent |
| Yang-Zhang | 35.4 | 38.9 | 37.0 | 37.5 | gaps + range + drift |
Conditional Volatility Models
GARCH(1,1) fitted by maximum likelihood with variance targeting, so the model reproduces the observed long-run vol exactly. α is how hard vol reacts to a shock, β how long it remembers one; their sum is persistence. The forecast curve is the average vol expected over each horizon — the quantity an option of that tenor is exposed to, not the single-day path.
0.0285α (reaction)
0.4764β (memory)
2.84e-4ω
0.5049Persistence α+β
1.0 daysShock half-life
37.58%Next-day vol
38.03%Long-run vol
45.84%EWMA λ=0.94
251 daysSample
Tail & Drawdown Risk
VaR and expected shortfall are historical — read straight off the empirical return distribution, with no normality assumption. The parametric figures next to them assume a Gaussian; the difference between the two is the size of the tail your model would have missed.
38.03%Annualised vol
37.04%Downside vol
39.07%Upside vol
-3.96%VaR 95% (hist)
-5.07%VaR 99% (hist)
-4.76%CVaR 95% (ES)
-5.71%CVaR 99% (ES)
-3.94%VaR 95% (normal)
-5.56%VaR 99% (normal)
0.092Skew
0.41Excess kurtosis
4.38%Days beyond 2σ
0.80%Days beyond 3σ
-20.22%Max drawdown
10.06Ulcer Index
38.28%CAGR
1.01Sharpe (rf=0)
1.03Sortino (rf=0)
3σ days occurred 3.0× more often than a normal distribution predicts (0.80% vs 0.27%). Any position sized off a Gaussian assumption is under-reserved for this name.
Overnight vs Intraday
Variance decomposed into the gap between one close and the next open, versus the regular session.
Overnight gap · 22.4% annualisedIntraday session · 31.7% annualised
Volatility Clustering ACF, 251 returns
Autocorrelation of raw returns, absolute returns and squared returns. Raw returns should be near zero (an efficient market); absolute and squared returns should not be — that persistence is the ARCH effect every conditional-vol model exists to capture. Bars outside the ±0.124 band are significant at 95%.
| Lag | Return | |Return| | Return² | Significance (|r|) |
|---|---|---|---|---|
| 1 | -0.081 | 0.057 | 0.033 | |
| 2 | 0.060 | 0.038 | 0.024 | |
| 3 | -0.081 | 0.039 | -0.005 | |
| 4 | -0.148 | 0.034 | 0.053 | |
| 5 | 0.045 | -0.079 | -0.094 | |
| 6 | -0.071 | -0.162 | -0.120 | |
| 7 | -0.031 | 0.015 | 0.002 | |
| 8 | 0.026 | -0.115 | -0.094 | |
| 9 | -0.060 | -0.061 | -0.065 | |
| 10 | 0.080 | 0.034 | -0.004 |
Volatility Change by Month month-over-month % change in 21-day realized vol
Not price returns — this is how much the volatility itself moved each month. Persistent positive months flag a season when risk tends to build in this name.
| Year | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec | Min | Max | Avg |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026 | -15.5 | 77.5 | -19.4 | -3.0 | 12.1 | 19.9 | -12.1 | 9.9 | -3.1 | -19.4 | 77.5 | 7.4 | |||
| 2025 | 12.4 | -29.9 | -29.9 | 12.4 | -8.8 | ||||||||||
| Avg change | -15.5 | +77.5 | -19.4 | -3.0 | +12.1 | +19.9 | -12.1 | +9.9 | -3.1 | — | +12.4 | -29.9 | -29.9 | 77.5 | 4.4 |
| % months vol rose | 0 | 100 | 0 | 0 | 100 | 100 | 0 | 100 | 0 | — | 100 | 0 | |||
KlickAnalytics AI
Ask the market a question. Get a calculated answer.
The AI is not a chatbot bolted onto a document store. It calls the same analytics engine that powers every screen on this platform — so what comes back is a number it computed from raw history, with the command that produced it.
86,000+ instruments
Global equities, ETFs, funds, options, FX, commodities, crypto, economics, filings, transcripts and news — one normalised symbol universe with adjusted history.
A real analytics engine
Screening, backtesting, technicals, options analytics, correlations, seasonality and factor models — computed on demand from raw prices, never a stale cache.
It shows its working
Answers arrive with the charts, tables and tool calls behind them, so you can check the number instead of trusting a paraphrase.
Your own documents
Upload filings, decks and research. Ask across them and the answer cites the page it came from.
Agents and workflows
Multi-step research that runs the platform's tools for you — screen, pull the history, compute, compare, then write it up.
MCP, CLI and API
The same command catalogue from Claude, your own agent, a shell or your pipeline. The answer on screen is the answer your job gets at 4am.
You ask
“How does NVDA usually trade through earnings?”
It calls
→ ka.earnings_moves(NVDA)
→ ka.options_expected_move(NVDA)
→ ka.options_expected_move(NVDA)
It answers
NVDA has averaged a 9.2% absolute move on the day after earnings and closed higher 67% of the time. Two in three reactions land between −4.2% and +16.3% — the distribution is skewed right, not symmetric.
Every figure computed live from our own history — not scraped, not summarised.
Or start with
“Screen the S&P 500 for oversold quality names.”
“Compare AAPL and MSFT over the last 5 years.”
“What moved the market today, and why?”
“Backtest a 50/200 crossover on TSLA.”
Create a free account to start asking
No credit card required. Confirm your email and the platform is open.
Create a free account
Sign in