Insights
Research notes, not market noise.
A working knowledge base for algorithmic trading, AI/ML, and the structural frameworks — Camarilla, Fibonacci, CPR, Volume Profile, options market microstructure — behind our research.
AI/ML Stack
NumPy: The Numerical Backbone of Algorithmic Trading
How vectorized array computation makes it possible to process years of 5-minute OHLCV data across an entire index in milliseconds, not minutes.
5 min read
Pandas for Market Data Analysis in Algo Trading
DataFrames, time-series indexing, and rolling windows: the tools that turn raw Zerodha Kite-format OHLCV bars into research-ready features.
6 min read
Scikit-learn: Classical Machine Learning for Trading Signals
Why the boring, interpretable, rigorously cross-validated tools — TimeSeriesSplit, CalibratedClassifierCV, preprocessing pipelines — do most of the real work.
6 min read
XGBoost: Gradient Boosting for High-Signal Trading Models
Why gradient-boosted trees dominate tabular prediction problems — and how that maps directly onto touch-prediction at a Camarilla or Fibonacci level.
6 min read
LightGBM: The Meta-Classifier Behind Our Daily Regime Calls
Leaf-wise tree growth makes LightGBM fast enough to retrain daily — which is exactly the job we give it: classifying the day's market regime and dominant strategy.
6 min read
Neural Networks: The Building Block of Deep Learning in Finance
Layers, weights, and activation functions — the universal function approximator underneath every deep learning model we build.
6 min read
Convolutional Neural Networks: Pattern Recognition Applied to Price Charts
Chart patterns are local, repeating shapes — exactly the kind of structure CNNs were built to detect, whether in images or price sequences.
6 min read
LSTM Networks: Modeling the Last Hour Before a Level Touch
Why our touch-prediction engine reads a 12-bar, one-hour lookback window through an LSTM instead of judging each 5-minute bar in isolation.
7 min read
TensorFlow: The Deep Learning Framework Behind Our AI/ML Research
The computational engine that trains and serves our CNN, LSTM, and neural network models at scale.
5 min read
Model Architecture
Ensemble Stacking: Blending Gradient Boosting and LSTM Predictions
No single model wins alone. How a calibrated logistic-regression meta-learner blends tabular and sequential predictions into one probability.
6 min read
Triple-Barrier Labeling: How We Turn Price Paths Into Training Targets
A touch event isn't a label by itself. The triple-barrier method — reversal, breakout, and time-based exit — turns what happens *next* into a clean target.
7 min read
Walk-Forward Validation: Why Random Cross-Validation Lies to You in Finance
Shuffled K-fold validation leaks the future into the past. Date-grouped TimeSeriesSplit is the only honest way to validate a trading model.
6 min read
Probability Calibration: Making a Model's "70%" Actually Mean 70%
A classifier can rank outcomes well and still be badly calibrated. Isotonic regression and the Brier score fix that — and why it matters for position sizing.
6 min read
Structural Levels
Camarilla Pivot Points: Intraday Support & Resistance Levels
The R3/S3/R4/S4 matrix, computed fresh every morning from yesterday's High, Low, and Close — one of five structural frameworks in our daily engine.
5 min read
Fibonacci Retracement & Extension: Mapping Reversal Zones in Price
Where the 61.8% level actually comes from, why it isn't magic, and how a full retracement-and-extension matrix feeds a touch-prediction model.
6 min read
Central Pivot Range (CPR): Reading Market Balance Before the Bell
Pivot, Top Central, Bottom Central — and CPR width, a single number that tells you whether today is likely to trend or chop, before the first candle prints.
6 min read
Initial Balance Extensions: Projecting the Day From Its First Hour
The high and low of 9:15–10:15 AM set the day's Initial Balance. What happens at 1.0x, 1.5x, and 2.0x of that range separates trend days from range days.
6 min read
Volume Profile & Point of Control: Reading Where the Market Actually Traded
Price tells you where the market is. Volume profile tells you where it agreed to be — and the POC is the single price it agreed on most.
7 min read
Volume Profile Proxy: Approximating the POC Without Tick Data
True volume-at-price needs tick-level data most retail and even institutional feeds don't cleanly provide. Here's the rolling 5-day proxy we use instead.
6 min read
Footprint Charts: Order Flow, and Why We Use Wick Ratios Instead
True footprint charts need bid/ask volume at every price tick. When that data isn't clean, candle wick geometry is a surprisingly good order-flow proxy.
6 min read
VWAP: The Reference Line Every Intraday Model Checks Against
The volume-weighted average price isn't just an execution benchmark — it's the default reversal target in our triple-barrier labeling logic.
5 min read
Options & Derivatives
Max Pain: Why Option Writers Quietly Pull Price Toward One Strike
The strike where aggregate option-writer losses are minimized isn't folklore — it's a computable, trackable dollar-value pain curve across the chain.
6 min read
Gamma Exposure (GEX) and the Zero-Gamma Strike
Where dealer hedging flips from dampening moves to amplifying them — and why that single strike matters more than most indicators on the chart.
7 min read
Open Interest Walls & Put-Call Ratio: Where Institutions Are Positioned
The highest Call OI strike acts as a ceiling; the highest Put OI strike acts as a floor. PCR tells you which side is currently winning the argument.
6 min read
Implied Volatility Skew & Smile: Reading Institutional Tail-Risk Positioning
Why out-of-the-money puts almost always carry richer implied volatility than equidistant calls — and what a shifting skew is telling you.
6 min read
Advanced Derivatives
What Is Advanced Derivatives Trading?
Beyond reading OI walls and gamma exposure: the pricing theory — Black-Scholes, Monte Carlo, the math underneath both — that lets you value a derivative from first principles.
6 min read
Linear Algebra in Quantitative Finance
Covariance matrices, portfolio variance, PCA on risk factors, and the matrix multiplications underneath every neural network in the stack.
6 min read
Calculus in Quantitative Finance: Where the Greeks Come From
Delta, Gamma, and Theta are literally derivatives — first and second order. And gradient descent, training every model in this stack, is calculus too.
6 min read
The Black-Scholes Model: Options Pricing From First Principles
Where the formula actually comes from, the assumptions it leans on, and why the volatility smile is really the market's way of correcting it.
7 min read
Monte Carlo Simulation for Derivatives Pricing
When there's no closed-form solution, simulate thousands of random price paths instead — and let the average payoff across all of them be the price.
6 min read
Quant Statistics
Probability Theory for Trading Signals: Thinking in Odds, Not Certainties
Every signal in a systematic strategy is a conditional probability statement. Treating it as a certainty is where most retail strategies fail.
6 min read
Regression Analysis in Quantitative Trading
From simple linear regression to ridge and Lasso — how fitting a line through noisy data underpins factor models and risk estimation.
6 min read
Time-Series Forecasting Models for Market Data
ARIMA, exponential smoothing, and ML-based forecasters — what each assumes, where each breaks, and why every forecast needs an error band.
7 min read
Research Workflow
Obsidian: Building a Connected Knowledge Base for Quant Research
Why a local-first, linked note graph beats a folder of disconnected documents — and how it doubles as the landing page for our model's own output.
5 min read
Why Our Models Write Their Own Markdown (.md) Files
Plain text, version-controlled alongside code, and machine-writable — the unglamorous reason every inference run ends in a YAML frontmatter block.
5 min read