Paper: arXiv 2507.16701
Authors: Akash Deep, Chris Monico, W. Brent Lindquist, Svetlozar T. Rachev, Frank J. Fabozzi
Abstract
We propose a machine learning-based extension of the classical binomial option pricing model that incorporates key market microstructure effects. Traditional models assume frictionless markets, overlooking empirical features such as bid-ask spreads, discrete price movements, and serial return correlations. Our framework augments the binomial tree with path-dependent transition probabilities estimated via Random Forest classifiers trained on high-frequency market data. This approach preserves no-arbitrage conditions while embedding real-world trading dynamics into the pricing model. Using 46,655 minute-level observations of SPY from January to June 2025, we achieve an AUC of 88.25% in forecasting one-step price movements. Order flow imbalance is identified as the most influential predictor, contributing 43.2% to feature importance. After resolving time-scaling inconsistencies in tree construction, our model yields option prices that deviate by 13.79% from Black-Scholes benchmarks, highlighting the impact of microstructure on fair value estimation. While computational limitations restrict the model to short-term derivatives, our results offer a robust, data-driven alternative to classical pricing methods grounded in empirical market behavior.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 6.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper integrates advanced theoretical concepts like the Minimal Martingale Measure and binary tree extensions with machine learning, indicating significant mathematical density. It also presents concrete empirical results from backtesting with 46,655 high-frequency SPY observations, including AUC metrics and price deviation percentages.
Research Flowchart
flowchart TD Start["Research Goal<br>Develop ML extension of binomial option pricing<br>incorporating market microstructure effects"] --> Input["Input Data<br>46,655 minute-level SPY observations<br>Jan - June 2025"] Input --> Method["Methodology<br>Binomial Tree augmented with<br>Random Forest Classifiers"] Method --> Process["Computational Process<br>Estimate path-dependent transition<br>probabilities using high-frequency data"] Process --> Outcome1["Outcome 1: Prediction Accuracy<br>AUC 88.25% in one-step price movements"] Process --> Outcome2["Outcome 2: Feature Importance<br>Order flow imbalance: 43.2% contribution"] Process --> Outcome3["Outcome 3: Option Pricing Impact<br>13.79% deviation from Black-Scholes<br>after resolving time-scaling issues"] Outcome1 --> Final["Key Finding<br>Robust, data-driven alternative to<br>classical pricing grounded in empirical behavior"] Outcome2 --> Final Outcome3 --> Final
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a machine‑learning‑based extension of the classical binomial option pricing model that incorporates market microstructure effects via Random Forest estimated transition probabilities.
Source sentence
We propose a machine learning-based extension of the classical binomial option pricing model that incorporates key market microstructure effects.
Assumes 4
- Traditional option pricing models assume frictionless markets, ignoring bid‑ask spreads, discrete price changes, and serial return correlations. model
Source sentence
Traditional option pricing models assume frictionless markets, ignoring empirical phenomena such as bid‑ask spreads, discrete price changes, and serial correlations in returns.
- The binary tree must satisfy the no‑arbitrage condition p∗(s)·u(s)+(1−p∗(s))·d(s)=erΔt for all states. model
Source sentence
For a binary tree to be arbitrage‑free, the adjusted probabilities p∗(s) must satisfy: p∗(s)·u(s) + (1 −p∗(s))·d(s) = erΔt
- State‑dependent transition probabilities can be learned from high‑frequency features including recent returns, spread metrics, volume, and order flow imbalance. data
Source sentence
Recent returns (rt−1, rt−2, . . . , rt−k) • Bid‑ask spread metrics (spreadt, spreadt/St, ∆spreadt) • Volume and trade metrics (Vt, sign(Vt), OFIt)
- The Minimal Martingale Measure adjustment yields risk‑neutral probabilities that preserve empirical information while satisfying no‑arbitrage. model
Source sentence
The MMM minimizes the Kullback‑Leibler divergence from the physical measure while ensuring no‑arbitrage: pMMM(s) = arg min p∗DKL(p∗||pRF(s)) subject to p∗· u(s) + (1 −p∗)· d(s) = erΔt
Data
- SPY minute‑level trading data market-prices unknown
Source sentence
Our empirical analysis employs minute‑level trading data for the SPDR S&P 500 ETF (SPY) from January 2, 2025, to June 25, 2025. The dataset contains 46,655 observations with complete open, high, low, close, volume, and tick count information.
Costs
Unclear - sources disagree
Source sentence (the authors' words)
Bid‑ask spread metrics (spreadt, spreadt/St, ∆spreadt) • Volume and trade metrics (Vt, sign(Vt), OFIt)
Finds 5
- The Random Forest classifier achieves an AUC of 88.25% in forecasting one‑step price movements. predictive-accuracy
Source sentence
... we achieve an AUC of 88.25% in forecasting one‑step price movements.
- Order flow imbalance is the most important predictor, contributing 43.2% to feature importance. performance
Source sentence
Order flow imbalance is identified as the most influential predictor, contributing 43.2% to feature importance.
- Option prices produced by the model deviate by 13.79% from Black‑Scholes benchmarks. pricing-calibration
Source sentence
our model yields option prices that deviate by 13.79% from Black‑Scholes benchmarks
- The model’s implied volatility is 21.0% versus a historical volatility of 24.3%, indicating improved efficiency. statistical
Source sentence
The model’s implied volatility of 21.0% compared to historical volatility of 24.3% suggests potential efficiency gains
- Computational performance: pricing completes in 0.004 seconds per option, but tree construction with 10 time steps takes 87.31 seconds, limiting scalability. other
Source sentence
actual option pricing completes in 0.004 seconds per option after initial tree construction, the 87.31 seconds required for tree construction with only 10 time steps severely constrains practical application.
Checked by others
Supported by
- Price predictability in limit order book with deep learning model
Compare the findings
This paper: Order flow imbalance is the most important predictor, contributing 43.2% to feature importance.
That paper: Incorporating volume imbalance markedly improves directional prediction performance.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (36)
- Binomial models for option valuation-examining and improving convergence (1996)
- A lattice framework for option pricing with two state variables (1988)
- Binomial Tree Method for American Option Pricing: Discrete Cosine Transform Approach (2025)
- Efficient procedures for valuing European and American path-dependent options (1993)
- American option valuation: new bounds, approximations, and a comparison of existing methods (1996)
- Option replication in discrete time with transaction costs (1992)
- The factor structure in equity options (2018)
- High-frequency trading in the stock market and the costs of options market making (2024)
and 28 more
- High-frequency trading and price discovery (2014)
- Order flow and expected option returns (2016)
- Option volume and stock prices: Evidence on where informed traders trade (1998)
- Option pricing using machine learning (2021)
- Machine learning methods for pricing financial derivatives — → 2406.00459
- Deep learning volatility: a deep neural network perspective on pricing and calibration in (rough) volatility models (2021)
- Volatility forecasting with machine learning and intraday commonality (2024)
- Machine learning-based analysis of volatility quantitative investment strategies for American financial stocks (2024)
- Empirical asset pricing via machine learning (2020)
- Random dynamics and finance: Constructing implied binomial trees from a predetermined stationary density (2007)
- Pricing European and American derivatives under a jump-diffusion process: A bivariate tree approach (2005)
- Regime-switching recombining tree for option pricing (2010)
- A general version of the fundamental theorem of asset pricing (1994) (2006)
- Pricing European options with stochastic volatility under the minimal entropy martingale measure (2016)
- The minimal entropy martingale measure and the valuation problem in incomplete markets (2000)
- On the minimal martingale measure and the m "o llmer-schweizer decomposition (1995)
- Financial modelling with jump processes (2003)
- High frequency market microstructure (2015)
- ETF arbitrage: Intraday evidence (2013)
- Using high-frequency data in dynamic portfolio choice (2008)
- Option pricing: A simplified approach (1979)
- The pricing of options and corporate liabilities (1973)
- Advanced financial market forecasting: integrating Monte Carlo simulations with ensemble Machine Learning models (2024)
- Risk-Adjusted Performance of Random Forest Models in High-Frequency Trading — → 2412.15448
- A multifactor analysis model for stock market prediction (2023)
- Unifying Market Microstructure and Dynamic Asset Pricing — → 2304.02356
- Martingales and arbitrage in multiperiod securities markets (1979)
- Martingales and stochastic integrals in the theory of continuous trading (1981)
3 of 36 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.