Pratik Bang

SYSTEMATIC TRADING RESEARCH · CHICAGO, ILLINOIS

Systematic strategies, from hypothesis to implementation.

I study trend following, time-series momentum, portfolio construction, information diffusion, strategy decay, and real-world trading-system design using Python, statistical analysis, and reproducible backtests.

Explore Research View GitHubpending

Research interests

What I work on

Nine themes that recur across my projects. Each one links into the research archive.

Trend Following

How persistent directional price behavior can be measured, sized, and traded across many markets.

Time-Series Momentum

Whether an asset's own past return predicts its future return, and over which horizons.

Strategy Decay

How and why measured edges weaken as markets, participants, and costs change.

Portfolio Construction

Turning many weak signals into one risk-controlled portfolio through sizing and correlation.

Financial Econometrics

Inference that survives overlapping data, cross-sectional dependence, and multiple testing.

Transaction-Cost Modeling

The distance between a gross backtest and an implementable set of returns.

Indian Equity Markets

Cross-sectional return behavior, liquidity, and information diffusion in NSE-listed stocks.

Trading Analytics

Reconstructing performance, exposure, and risk from broker records and market data.

Machine Learning for Markets

Where statistical learning adds information beyond price, and where it mostly adds overfitting.

Featured research

Selected projects

Three current lines of work. Status is stated on every card, and no numerical results appear until final outputs exist.

All research
COMPLETEDJAN 2026 — SEP 2026

Modern-Era Trend Following in Global Futures

QuestionDoes systematic trend following continue to generate economically meaningful returns in modern futures markets, and how has its performance changed across time, market structure, volatility regimes, and implementation assumptions?

Research area
Futures · Trend Following · Portfolio Construction
Frequency
Daily data, monthly signal evaluation and rebalancing

Methods

Non-overlapping monthly returnsSign and magnitude regressionsNewey–West inferenceTwo-way clustered standard errors+6 more

Current statusCompleted. Net Sharpe of the monthly 12/1 benchmark falls from 1.51 before 2009 to 0.42 after 2008. Combining 12 pre-specified lookback and holding-period rules into one equal-weight, internally netted book raises post-2008 net Sharpe to 0.62, lowers maximum drawdown from 18.8% to 12.8%, and reduces annual cost drag from 1.20% to 0.75%. In the final 74-market funded simulation, the multi-horizon book has the highest Sharpe above cash at 0.64 with a 9.1% maximum drawdown.

Full project page Report Codepending
IN PROGRESS2026 - ONGOING

Slow Diffusion of Information in NSE Stocks

QuestionDo returns in NSE-listed stocks display evidence of slow information diffusion, and can delayed price response help explain subsequent return continuation?

Research area
Equities · Momentum · Statistical Research
Frequency
Daily data, monthly portfolio formation and rebalancing

Methods

Cross-sectional portfolio sortsPanel regressions with controlsTwo-way clustered standard errorsNewey–West inference+3 more

Current statusWork in progress. The design tests continuation while controlling for liquidity and size, so that a diffusion interpretation is not confused with a liquidity premium.

Full project page Reportpending Codepending
IN DEVELOPMENTNOV 2025 - ONGOING

Multi-Broker Trading Analytics and Risk Dashboard

QuestionCan positions, trades, cash flows, and account values from several brokerage accounts be consolidated into one auditable daily NAV series and a consistent set of portfolio performance and risk metrics?

Research area
Trading Infrastructure · Portfolio Construction
Frequency
Daily, with intraday position refresh

Methods

Daily NAV reconstructionTime-weighted return constructionCash-flow separationBroker reconciliation and break reporting+1 more

Current statusA private analytics system in development. The core problem is accounting discipline: separating trading performance from capital flows, and reconciling every derived number back to a broker record.

Full project page Reportpending Codepending

Research process

How a question becomes a result

The same eight steps apply whether the outcome supports the hypothesis or rejects it.

  1. 01

    Define an economic question

    Start from a mechanism worth believing in, not from a pattern found in a data sweep.

  2. 02

    Form falsifiable hypotheses

    State in advance what evidence would count against the idea.

  3. 03

    Acquire and validate data

    Document sources, adjustments, gaps, and everything that could quietly bias a result.

  4. 04

    Build a reproducible baseline

    Implement the simplest canonical version first so later choices can be attributed.

  5. 05

    Test alternative explanations

    Ask what else could produce the same numbers: risk, structure, liquidity, or luck.

  6. 06

    Add costs and constraints

    Introduce commissions, slippage, rolls, capacity, and rebalancing discipline.

  7. 07

    Measure uncertainty and robustness

    Bootstrap, subperiod, and parameter-sensitivity work before any conclusion is written.

  8. 08

    Interpret and document limitations

    Separate what the test shows from what I infer, and say plainly what remains unknown.

Skills and tools

Methods, code, and data

Grouped by how they are used rather than listed as one undifferentiated set.

Research and statistics

  • Hypothesis development
  • Panel regressions
  • Newey–West inference
  • Clustered standard errors
  • Bootstrap confidence intervals
  • Structural-break analysis
  • Walk-forward and subperiod analysis
  • Drawdown and tail-risk analysis

Programming and data

  • Python
  • pandas
  • NumPy
  • SciPy
  • statsmodels
  • scikit-learn
  • SQL
  • Jupyter
  • Google Colab
  • Git and GitHub

Trading research

  • Time-series momentum
  • Trend following
  • Volatility scaling
  • Portfolio construction
  • Futures data
  • Equity cross-sectional research
  • Transaction costs
  • Slippage
  • Contract rolls
  • Performance attribution

Platforms and data workflows

  • RealTest
  • Norgate/MetaStock data
  • Broker APIs
  • Market-data pipelines
  • Dashboard development

Contact

I'm interested in systematic trading research, quantitative roles, research collaboration, and thoughtful conversations about market behavior.

Email MeConnect on LinkedInView GitHubpending