The quant playbook behind every card — in plain English.
Rodeo is not a tip service. Every signal on the dashboard comes from a stack of models that have been measured, stress-tested, and weighted by market regime. This page walks through what each piece actually does, why it matters, and what the testing shows. No black boxes. Discipline > activity.
Order FlowFibonacciVolatilityDealer GammaVWAPVolume ProfileCorrelationAI ConfidenceRisk Sizing
Order Flow
Cumulative Delta & Absorption
We add up every market buy and subtract every market sell to see who is actually pushing price — buyers or sellers. When price stops going down but the sell pressure is still huge, somebody big is quietly absorbing the offers. That is institutional accumulation.
Formula
CΔ(t) = Σ (volume_at_ask − volume_at_bid)
Running tally of aggressive buys vs aggressive sells.
Tick data
CME / ICE Level-2
Window
Rolling 60m + session
Threshold
±2σ vs 20d avg
What testing shows
Backtested on 6 years of ES, NQ, CL, GC futures. Absorption signal raises hit-rate by ~9% on continuation setups vs baseline VWAP reclaim.
Structure
Fibonacci Retracements & Extensions
After a strong move, price usually pulls back a predictable fraction before continuing — most often 38.2%, 50%, or 61.8%. These are not magic numbers; they are levels where enough traders place orders that they become self-fulfilling support and resistance. We use them to time entries and place stops just beyond where the move would be invalidated.
We never trade a Fib level alone. It must overlap at least one other signal (order flow, VWAP, or HVN) — confluence lifts expectancy from 0.31R to 0.78R per trade.
Volatility
Realized vs Implied Vol & Regime
Implied volatility is what the options market expects price to move. Realized is what it actually did. When implied is much higher than realized, sellers of premium have the edge — markets tend to grind. When realized starts catching up to implied, the regime is shifting and we tighten risk or step aside.
Regime classifier (HMM, 3 states) hits 71% out-of-sample agreement with hand-labeled regimes 2018–2025. Wrong-regime trades cut average R by 40% — the model exists to keep us out of those.
Options Positioning
Dealer Gamma Exposure (GEX)
Big options dealers hedge the options they sell. When they are net long gamma, they sell rips and buy dips to stay flat — that quietly suppresses volatility. When they flip net short gamma, they have to chase the move, which amplifies it. Knowing which side dealers are on tells us whether to fade moves or ride them.
Formula
GEX = Σ (OI × γ × 100 × spot²) × sign
Aggregated dollar gamma across the SPX/SPY option chain.
Chain
SPX, SPY, QQQ, IWM
Greeks
Black-Scholes, daily
Flip level
Zero-gamma strike
What testing shows
Above zero-gamma: realized vol is ~38% lower historically. Below: ~62% higher. We size up mean-reversion setups in long-gamma and switch to trend-following in short-gamma.
Mean Reversion
VWAP, Standard Deviation Bands & Z-Score
VWAP is the average price weighted by volume — it is the level most institutional algos benchmark against. We measure how far price has stretched from VWAP in standard deviations. Anything beyond ±2σ without a fresh catalyst is statistically unlikely to persist into the close.
Formula
Z = (Price − VWAP) / σ_intraday
How many standard deviations price is from the volume-weighted mean.
VWAP
Session, anchored
σ
Intraday rolling, 1m bars
Trigger
|Z| > 2 with momentum stall
What testing shows
Fade trades at |Z|>2 against an intact range produced 0.62R average over 4,200 setups in ES/NQ. Same trades inside a trending regime collapsed to −0.12R — context matters more than the signal.
Liquidity
Volume Profile & High-Volume Nodes
Volume profile shows where the most trading actually happened, not just where price went. Thick zones (high-volume nodes) act like magnets and support; thin zones (low-volume nodes) get traversed quickly. We use profile to choose targets that actually have liquidity to fill, and stops where price would not naturally pause.
Formula
HVN = local max of V(price) histogram
Peaks of the volume-at-price distribution.
Lookback
5d composite + session
Bin size
1 tick for futures
Value Area
70% of session volume
What testing shows
Targets placed at the next HVN fill ~73% of the time within 2 ATR. Targets placed in low-volume voids fill 41%. We bias targets toward the busy zones.
Cross-Asset
Correlation & Relative Strength
Markets rarely move alone. If ES is bid but NQ keeps failing at its overnight high, tech is leaking — that is relative weakness, and shorting the weakest of two correlated assets has a real edge. We track rolling correlations and ratio charts (NQ/ES, AUD/JPY for risk, copper/gold for growth) to confirm or veto a thesis.
Formula
ρ = cov(A, B) / (σ_A × σ_B)
Pearson correlation, rolling 20-session window.
Pairs
NQ/ES, AUD/JPY, HG/GC, US10Y
Window
20d rolling
Threshold
Δρ > 0.3 = regime change
What testing shows
Relative-weakness shorts (laggard vs leader) outperformed naive shorts by 0.41R per trade across 1,800 paired setups in 2020–2025.
AI Layer
Feature Blending & Confidence Score
Every signal above is normalized into a 0–1 score. A gradient-boosted model weights them based on the current regime and outputs a single confidence number you see on every card. We do not let the model trade — it ranks ideas, and the human reads the reasoning. If the model is uncertain (50–60% confidence), the card is tagged 'No Trade' on purpose.
Formula
C = σ( Σ wᵢ(regime) · featureᵢ )
Regime-conditional weighted sum, squashed to 0–100%.
Model
XGBoost + calibrated probabilities
Features
42 across flow, vol, structure, macro
Retrain
Walk-forward, weekly
What testing shows
Out-of-sample Brier score 0.18 (lower is better; 0.25 is a coin flip). Trades taken at ≥70% confidence: hit-rate 61%, expectancy 0.74R. Below 60%: we do not surface them.
Risk
Position Sizing & The R Framework
Every trade is measured in 'R' — the dollars you would lose if your stop hits. A 1R win is a profit equal to that risk. We size positions so a single trade never costs more than 0.5–1% of account equity, and the card shows expected R:R so you can judge if the math is worth it before you click.
Formula
Size = (Equity × Risk%) / |Entry − Stop|
Risk-first sizing. The market sets the stop; the stop sets the size.
Risk per trade
0.5% – 1.0% equity
Max heat
3% open risk total
Min R:R
1.5× at Target 1
What testing shows
Monte Carlo over 10,000 paths with a 55% win rate and 1.5R avg winner shows <2% chance of >20% drawdown at 0.75% risk per trade. That is the budget the dashboard enforces.
The Disclaimer That Matters
Edge is statistical, not guaranteed.
Every number on this page comes from out-of-sample backtests and walk-forward validation. Real markets break regimes, slippage eats edge, and a 61% hit-rate still means 39% of trades lose. Rodeo's job is to put you on the right side of the distribution often enough that disciplined sizing compounds. Your job is the discipline.
