How We Research
Our approach blends mathematics, code, and observation. We use a combination of historical data, real-time feeds, and deep learning models to test ideas before they ever reach a trading screen. We focus on three main areas of research:
Implied Volatility and Term Structure
We study how implied volatility behaves over time and across expirations. We map the term structure to see when short-term and long-term volatility move in sync or drift apart. This helps us understand when the market is calm, balanced, or stressed.
Dealer Positioning and Exposure
We track how dealers hedge positions through advanced Greeks such as Gamma Exposure (GEX), Delta Exposure (DEX), Charm, Vanna, and Vomma. By studying these metrics, we can see where market makers might need to buy or sell as prices move. This gives us insight into pressure points that often lead to reversals or accelerations.
Machine Learning and Pattern Recognition
We build neural networks that analyze millions of data points across multiple assets. These models do not predict. They detect. They recognize repeating behavior in volatility, flow, and price structure that traditional models miss. When the same conditions appear again, we are prepared.
Testing and Validation
Every idea is tested in three stages: Historical BacktestingWe replay years of market data to see if the idea holds through different cycles, regimes, and shocks. Forward TestingWe run the idea in real-time simulations to watch how it reacts to live volatility, liquidity, and order flow. IntegrationIf the idea proves stable, it becomes part of our live trading system.We continue to monitor it daily and remove it the moment data stops supporting it. Nothing goes live without passing all three stages.
Tools and Data Sources
Our work depends on a wide range of inputs, including:
Real-time option chain data across multiple exchanges Volatility surface modeling and skew tracking Liquidity and open interest analytics Cross-asset correlations and macro volatility indexes Proprietary deep learning models for signal clustering Each dataset is stored, structured, and versioned for accuracy.We believe transparency with our own data builds confidence in our decisions.
The Role of Advanced Greeks
Traditional Greeks explain how an option reacts to simple changes in price, time, or volatility. Our research expands that view. We study higher-order Greeks that reveal how one force affects another.
Gamma Exposure (GEX): shows where hedging activity can pin or push price Delta Exposure (DEX): measures total directional bias in dealer positioning Charm: tracks how delta changes over time, showing when slow drifts can speed up Vanna: connects volatility changes with directional movement Vomma and VEX: measure how volatility itself becomes volatile Term Structure and Skew: show where fear or complacency builds in the market By combining these readings, we can map where stress builds, where momentum fades, and where probability tilts in our favor.
From Research to Trading
Once a concept passes testing, it becomes a rule inside our trading framework. Each rule defines when to enter, when to exit, and how to size a position based on volatility and exposure. We do not rely on opinions or predictions.We rely on probability, structure, and discipline. Research and trading are not separate at Bergenstone.They are part of one continuous loop that refines itself with every trade.
Our Commitment to Independence
All of our research supports one purpose: to improve how we trade our own capital. We do not sell signals, manage client funds, or share live positions. Our work stays internal and self-funded. This allows us to stay focused on truth in data, not performance pressure or marketing. We research to understand, not to impress. They are part of one continuous loop that refines itself with every trade.
Research is what keeps Bergenstone evolving. It gives us clarity in uncertain markets and confidence in every decision we make.
By combining deep learning, advanced Greeks, and disciplined testing, we continue to refine the small edges that compound into long-term results.
Bergenstone Research
Independent. Data-driven. Always learning.