Tool Tutorials

Relative Volume (RVOL) Tutorial

Chapter 1

Mathematical Foundations of Relative Volume

Raw volume metrics (e.g., trading 40 million shares) provide zero contextual insight without referencing an asset's baseline liquidity. Relative Volume (RVOL) normalizes raw share count by expressing a session's volume as a multiple of its historical rolling average.

Prior 30 sessions build SMA_30 → Session t is measured against it 30-session average 2.3× Session t RVOL
RVOLt = Vt ÷ SMA of the prior N sessions. Session t is never included in its own baseline.

The Unbiased RVOL Formula

The Relative Volume multiple for session t is defined as the ratio of current session volume Vt to the N-period simple moving average of prior volume:

RVOLt = Vt ÷ V̅t−1, N

Where the historical baseline V̅t−1, N is calculated strictly across prior sessions:

t−1, N = (1 ÷ N) × Σ Vt−i for i = 1…N
  • RVOL = 1.0: Baseline session (volume equals the N-period norm).
  • RVOL = 2.0: Volume expansion at 200% of the historical average.

Excluded Denominator Principle

To prevent mathematical distortion, session t must never be included in its own baseline calculation.

Incorrect (Biased): SMAN includes Vt ⇒ inflates baseline, dampens true RVOL
Correct (Unbiased): SMAN = average of prior N sessions only ⇒ measures expansion against an uncorrupted baseline
Denominator Isolation
Including an extreme spike (e.g., 4x volume) in its own baseline raises the denominator by ~10%, artificially dampening the calculated RVOL multiple to ~3.6x. Strict exclusion preserves the true size of the expansion.
Open SPY
Chapter 2

Thresholding Modes & Configuration Mechanics

Quantitative tools implement flexible thresholding models to segment volume regimes effectively.

ModeFormula DefinitionAnalytical Use Case
Greater ThanRVOLt ≥ ThresholdOpen-ended tail analysis
Range ModeMin_Threshold ≤ RVOLt ≤ Max_LimitBanded, non-extreme regimes
Threshold Selection Modes

Grid Rounding & Boundary Alignment

To prevent auto-detection boundary issues where rounding might exclude the seeding session, thresholding algorithms apply consistent grid precision across filtering parameters:

RVOLdisplayed = Round(Vt ÷ V̅t−1, 30, 1)
StepValue
Raw RVOL2.66x
Rounded Grid2.7x
Threshold Filter2.7x (session included)
Threshold Boundary Resolution
Configuration Tip
When auto-detecting parameters on low-activity sessions (e.g., RVOL = 0.6), the filter threshold expands to include thousands of historical sessions. To isolate true institutional liquidity events, set fixed thresholds at RVOL ≥ 1.5x or higher.
Chapter 3

Directional Asymmetry in Volume Spikes

Segmenting high-RVOL sessions by price direction reveals a fundamental market asymmetry: institutional volume expansion is heavily skewed toward down days.

RVOL Threshold Green Sessions Red Sessions Green Session Share (%)
0.5x (Baseline)2,6992,23254.7%
1.5x16938030.8%
2.3x215826.6%
SPY Directional Distribution Across RVOL Thresholds (Historical Benchmark)

Institutional Accumulation vs. Liquidation Dynamics

Baseline Equity Drift ⇒ Green Share ≈ 55%   vs.   High Volume Expansion (RVOL ≥ 2.3) ⇒ Green Share ≈ 27%
Accumulation (Buying)Liquidation (Selling)
Passive limit ordersAggressive market orders
Patient, multi-day executionSudden, concentrated execution
Absorbed across timeCreates instant volume spikes
What Drives the Volume
  1. Buying Structure: Institutional accumulation is typically executed via algorithms designed to minimize market impact, spreading buying across multiple sessions.
  2. Selling Structure: Portfolio de-risking and forced liquidations prioritize execution speed over price impact, compressing large order volume into brief, highly visible spikes.
Sample Size Caution
At extreme thresholds (RVOL ≥ 2.3x), sample size drops rapidly (e.g., 21 green sessions). Treat high-threshold percentages with appropriate statistical caution due to wider confidence intervals.
Chapter 4

Forward Expectancy & Post-Spike Dynamics

To evaluate whether high RVOL acts as trend continuation or exhaustion, we measure forward performance across 1-day, 5-day, and 10-day time frames (t+n).

RVOL Threshold Direction Next-Day Green % Avg Win (%) Avg Loss (%) Expectancy
0.5x (Baseline)Green Day54.0%+0.62%−0.65%+0.035%
0.5x (Baseline)Red Day56.0%+0.68%−0.61%+0.112%
1.5x ExpansionGreen Day43.0%+0.71%−0.82%−0.162%
1.5x ExpansionRed Day57.0%+0.89%−0.75%+0.185%
2.3x ExtremeGreen Day43.0%+1.12%−1.05%−0.117%
2.3x ExtremeRed Day52.0%+2.23%−1.42%+0.478%
Forward Performance Following Volume Expansion (SPY Benchmark)

Up-Move Exhaustion vs. Down-Move Persistence

Post-Green Volume Spike Win rate 54% → 43% (exhaustion) RVOL expansion Post-Red Volume Spike Avg win +2.23% vs avg loss −1.42% RVOL expansion
High-volume green sessions often mark short-term buyers' exhaustion. High-volume red sessions keep a near coin-flip win rate but with asymmetric payoff on the rebound.
  1. High-Volume Green Days (Exhaustion): When RVOL reaches 1.5x on a green session, t+1 win rates drop from 54% to 43%. Buying volume expansion often marks temporary buyers' exhaustion rather than trend continuation.
  2. High-Volume Red Days (Capitulation & Asymmetry): On red sessions with 2.3x RVOL, t+1 win rates remain near 52%, but the average winning move (+2.23%) significantly exceeds the average loss (−1.42%).

Mathematical Expectancy Equation

Evaluating win rates without payoff size yields incomplete analysis. Expected return E(R) incorporates both parameters:

E(R) = (Pwin × Avg Gain) − ((1 − Pwin) × Avg Loss)
For Red Session RVOL ≥ 2.3x: E(R) = (0.52 × 2.23%) − (0.48 × 1.42%) = +0.478%

Despite a near-coin-flip win rate (52%), positive payoff skew creates a favorable positive expectancy environment.

Chapter 5

Establishing Control Baselines

To determine whether an anomaly exists, observed data must be evaluated against a control baseline.

Experimental Comparison Protocol Threshold 0.5x Baseline win rate ~54% Target 1.5x Experimental win rate ~43% Delta = 43% − 54% = −11% shift
Set 0.5x to capture nearly all sessions as a control, then raise the threshold and measure the divergence from that baseline.

Generating Ticker-Specific Baselines

  1. Set the RVOL parameter to 0.5x to capture nearly all trading sessions (~99%).
  2. Record baseline directional distribution, t+1 win odds, and average move sizes.
  3. Increase the parameter to the target threshold (e.g., 1.5x or 2.0x) to measure the true divergence (Δ).
ΔExpectancy = E(R)Target − E(R)Baseline
Ticker-Specific Baselines
Baselines vary by asset class and liquidity profile. SPY's 54% baseline win rate cannot be applied to small caps, leveraged ETFs, or commodities. Always generate a fresh 0.5x control baseline for the specific asset under evaluation.
Build a baseline on SPY