Realized vs Implied Volatility: How to Spot the Volatility Risk Premium

What Is Realized Volatility?

Realized volatility (RV) measures how much an asset's price actually moved over a specific lookback window. It is backward-looking and computed from observed returns. The simplest estimator uses daily closing prices, but more advanced estimators extract additional information from intraday highs and lows to produce tighter estimates.

Close-to-Close (Standard) Estimator

The most common RV calculation. Take the standard deviation of log returns over N trading days, then annualize:

Where \(C_i\) is the closing price on day \(i\) and \(\bar{r}\) is the mean log return. This estimator is simple and widely used, but it ignores all intraday price action. A stock that gaps up 5% at the open, trades in a 3% range, and closes unchanged registers zero return for that day — even though it was highly volatile.

Parkinson (High-Low) Estimator

Parkinson (1980) uses daily high and low prices instead of just closes, capturing intraday range information:

Parkinson is approximately 5x more efficient than close-to-close — it achieves the same statistical precision with one-fifth the data. The tradeoff: it assumes continuous trading (no overnight gaps), so it underestimates volatility in assets with large opening gaps.

Yang-Zhang Estimator

Yang-Zhang (2000) combines overnight returns (close-to-open), opening returns (open-to-close), and the Rogers-Satchell intraday estimator into a single minimum-variance formula:

This is the most efficient unbiased estimator for assets with meaningful overnight gaps — which includes nearly every equity. It handles gap risk, intraday range, and drift simultaneously. When you see RV quoted by professional platforms, it is often a Yang-Zhang or Rogers-Satchell variant.

What Is Implied Volatility?

Implied volatility (IV) is the market's forward-looking estimate of how much the underlying will move over the life of an option. It is not directly observed — it is implied by the option's price via the Black-Scholes model or its extensions. Given the current option price, underlying price, strike, time to expiry, and risk-free rate, IV is the volatility input that makes the model price match the market price.

The key distinction: RV tells you what happened. IV tells you what the market expects (and fears) will happen. The gap between them is where the edge lives.

The Volatility Risk Premium: IV Almost Always Exceeds RV

The Volatility Risk Premium (VRP) is the spread between implied volatility and subsequent realized volatility:

Empirically, IV exceeds RV approximately 85-90% of the time on major indices. This is not a market inefficiency — it is compensation for risk. Option sellers bear the risk of extreme moves (crashes, gap events, black swans). Option buyers — primarily institutional hedgers — are willing to overpay for protection because the cost of being unhedged during a tail event far exceeds the cost of persistent premium leakage.

The VRP exists because of a structural supply-demand imbalance. Fund managers must hedge. Pension funds must protect portfolios. Retail investors buy protective puts. This persistent demand for insurance inflates IV above fair value. Option sellers — the insurance providers — collect this premium as compensation for bearing the tail risk.

VRP Regimes: The Five States of the Spread

Not all VRP environments are the same. FlashAlpha's /v1/volatility/{symbol} endpoint classifies the current IV-RV spread into one of five regimes, each with distinct trading implications:

Pulling VRP Data from the FlashAlpha API

The GET /v1/volatility/{symbol} endpoint returns ATM implied volatility, realized volatility across five lookback windows (5d, 10d, 20d, 30d, 60d), the IV-RV spread, and the regime assessment. Here is how to pull and interpret it:

import requests

API_KEY = "your_api_key"
BASE_URL = "https://lab.flashalpha.com"

def get_vrp(symbol: str) -> dict:
    """Fetch volatility profile and VRP for a symbol."""
    resp = requests.get(
        f"{BASE_URL}/v1/volatility/{symbol}",
        headers={"X-Api-Key": API_KEY}
    )
    resp.raise_for_status()
    return resp.json()

# Pull AAPL volatility data
vol = get_vrp("AAPL")

print(f"AAPL Volatility Risk Premium Analysis")
print(f"{'='*42}")
print(f"  ATM IV:          {vol['atm_iv']:.2f}%")
print(f"  RV (5d):         {vol['realized_vol']['rv_5d']:.2f}%")
print(f"  RV (10d):        {vol['realized_vol']['rv_10d']:.2f}%")
print(f"  RV (20d):        {vol['realized_vol']['rv_20d']:.2f}%")
print(f"  RV (30d):        {vol['realized_vol']['rv_30d']:.2f}%")
print(f"  RV (60d):        {vol['realized_vol']['rv_60d']:.2f}%")
print(f"  VRP (20d):       {vol['iv_rv_spreads']['vrp_20d']:.2f}%")
print(f"  Assessment:      {vol['iv_rv_spreads']['assessment']}")
print(f"  Term Structure:  {vol['term_structure']['state']}")

Example output:

AAPL Volatility Risk Premium Analysis
==========================================
  ATM IV:          27.80%
  RV (5d):         19.40%
  RV (10d):        21.30%
  RV (20d):        23.10%
  RV (30d):        24.50%
  RV (60d):        22.80%
  VRP (20d):       4.70%
  Assessment:      healthy_premium
  Term Structure:  contango

The 20-day VRP of 4.70% tells you that implied volatility is pricing in 4.7 percentage points more annualized movement than the stock has actually delivered over the past 20 trading days. The healthy_premium assessment confirms this is a favorable environment for selling premium.

Comparing RV Windows: Short-Term vs Long-Term Realized Vol

Comparing different RV windows reveals whether volatility is accelerating or decelerating — a critical signal for timing premium sales:

import requests

API_KEY = "your_api_key"
BASE_URL = "https://lab.flashalpha.com"

def analyze_rv_term_structure(symbol: str):
    """Compare short-term vs long-term RV to detect vol acceleration."""
    resp = requests.get(
        f"{BASE_URL}/v1/volatility/{symbol}",
        headers={"X-Api-Key": API_KEY}
    )
    resp.raise_for_status()
    vol = resp.json()

    rv = vol["realized_vol"]
    atm = vol["atm_iv"]

    windows = [
        ("5d",  rv["rv_5d"]),
        ("10d", rv["rv_10d"]),
        ("20d", rv["rv_20d"]),
        ("30d", rv["rv_30d"]),
        ("60d", rv["rv_60d"]),
    ]

    print(f"\n{symbol} — RV Term Structure vs ATM IV ({atm:.1f}%)")
    print(f"{'Window':<8} {'RV':>8} {'VRP':>8} {'IV/RV':>8}")
    print("-" * 36)
    for label, rv_val in windows:
        vrp = atm - rv_val
        ratio = atm / rv_val if rv_val > 0 else float('inf')
        print(f"{label:<8} {rv_val:>7.1f}% {vrp:>7.1f}% {ratio:>7.2f}x")

    # Detect acceleration/deceleration
    rv_5 = rv["rv_5d"]
    rv_60 = rv["rv_60d"]
    if rv_5 > rv_60 * 1.3:
        print(f"\n⚠ Short-term RV accelerating: 5d ({rv_5:.1f}%) >> 60d ({rv_60:.1f}%)")
        print("  Caution: recent movement is elevated — VRP may compress.")
    elif rv_5 < rv_60 * 0.7:
        print(f"\n✓ Short-term RV decelerating: 5d ({rv_5:.1f}%) << 60d ({rv_60:.1f}%)")
        print("  Favorable: recent calm amplifies the premium available.")
    else:
        print(f"\n— RV stable across windows. No acceleration signal.")

    return vol

# Scan a watchlist
for ticker in ["SPY", "TSLA", "NVDA", "AAPL"]:
    analyze_rv_term_structure(ticker)

When short-term RV (5d) is dropping while long-term RV (60d) remains elevated, IV tends to be sticky — creating a widening VRP. This is the ideal setup for selling premium. When short-term RV is spiking above long-term RV, the market is in an acceleration phase and the VRP may be about to compress or invert.

Multi-Ticker VRP Scanner

The most practical use of VRP data is scanning a universe for the best premium-selling opportunities. This scanner ranks tickers by VRP richness and filters by regime:

import requests
import pandas as pd

API_KEY = "your_api_key"
BASE_URL = "https://lab.flashalpha.com"

watchlist = [
    "SPY", "QQQ", "IWM", "AAPL", "MSFT", "NVDA", "TSLA", "AMZN",
    "META", "GOOGL", "AMD", "NFLX", "COIN", "PLTR", "BA", "XOM"
]

results = []
for ticker in watchlist:
    try:
        resp = requests.get(
            f"{BASE_URL}/v1/volatility/{ticker}",
            headers={"X-Api-Key": API_KEY}
        )
        resp.raise_for_status()
        vol = resp.json()

        atm = vol["atm_iv"]
        rv_20 = vol["realized_vol"]["rv_20d"]
        vrp = vol["iv_rv_spreads"]["vrp_20d"]
        assessment = vol["iv_rv_spreads"]["assessment"]
        iv_rv_ratio = atm / rv_20 if rv_20 > 0 else 0

        results.append({
            "ticker": ticker,
            "atm_iv": round(atm, 1),
            "rv_20d": round(rv_20, 1),
            "vrp_20d": round(vrp, 1),
            "iv/rv": round(iv_rv_ratio, 2),
            "regime": assessment
        })
    except Exception:
        continue

df = pd.DataFrame(results)

# Best opportunities: healthy or very_high premium, sorted by VRP
sellable = df[df["regime"].isin(["healthy_premium", "very_high_premium"])]
sellable = sellable.sort_values("vrp_20d", ascending=False)

print("=== Premium Selling Opportunities (Healthy/Rich VRP) ===\n")
print(sellable.to_string(index=False) if len(sellable) > 0 else "No opportunities today")

# Danger: negative spread or danger zone
avoid = df[df["regime"].isin(["negative_spread", "danger_zone"])]
if len(avoid) > 0:
    print("\n=== AVOID — Negative/Danger VRP ===\n")
    print(avoid.to_string(index=False))

Trading Implications: When to Sell and When to Stay Away

The VRP regime directly maps to trading decisions. Here is the framework:

The most common mistake is selling premium based solely on IV level. An IV of 40% might look "high" — but if the stock is realizing 45% over the past week, you are selling options below fair value. Always compare IV to RV. The spread is what matters, not the level.

Strategy Selection by VRP Regime

Related Reading

• Quantitative Architectures for VRP Harvesting — deep dive into systematic VRP strategies with advanced Greeks
• IV Rank vs IV Percentile — complementary metrics for assessing whether IV is elevated
• IV Crush Explained — how event-driven VRP collapses after earnings
• Volatility Surface API — build full IV surfaces for deeper analysis

Frequently Asked Questions