Welcome to the ta4j Wiki

ta4j gives you the building blocks for technical-analysis-driven systems in Java: bar series, indicators, rules, strategies, reports, and a unified trading-record model that now spans backtests, paper trading, and live execution.

The current wiki reflects ta4j’s newer unified trading stack:

• BaseTradingRecord is the default trading-record implementation for both backtests and live or paper-trading flows.
• BarSeriesManager is the default single-strategy backtest driver and now accepts either its own default record factory or a record you provide.
• BacktestExecutor builds on BarSeriesManager when you want to rank or tune many strategies at once, including weighted normalized leaderboards.
• Manual loops are still the right tool when orders and fills are decoupled, partial fills matter, or your broker confirms executions asynchronously.
• LiveTradingRecord and ExecutionFill remain available only as 0.22.x compatibility facades. New code should use BaseTradingRecord and TradeFill.

Choose your audience path

• Production users and integrators should start in Start Here, then follow Canonical User Journey.
• Maintainers and design contributors should use architecture and delivery artifacts under architecture/ and completed-features/README.md, which are not part of the primary user onboarding path.

What’s Newer On Current Master

• Configurable backtest execution models: BarSeriesManager and BacktestExecutor can now stay on the default next-open model or switch to current-close, slippage, or stop-limit execution.
• Weighted strategy ranking: BacktestExecutionResult#getTopStrategiesWeighted(...) and WeightedCriterion let you rank strategies by a normalized composite score instead of a single raw metric.
• One trade-record story for partial fills: New code can stream TradeFill values directly with TradingRecord.operate(fill) or group an order with Trade.fromFills(...), then inspect getCurrentPosition() and getOpenPositions() on the same record.
• Broader analysis surface: Recent current-master additions include SharpeRatioCriterion, SortinoRatioCriterion, CalmarRatioCriterion, OmegaRatioCriterion, and volume pressure indicators such as ForceIndexIndicator, EaseOfMovementIndicator, and KlingerVolumeOscillatorIndicator.
• Forecast foundation targeting 0.23.1: Num-only summaries now declare empirical or analytic support, return estimators compose canonical ReturnMoments, and feature vectors publish representation-bound schemas.
• Exact and explicit price models: MonteCarloPriceForecastIndicator summarizes transformed terminal-price paths exactly; LognormalApproximationPriceForecastIndicator is the clearly named analytic alternative when paths are unavailable.
• State-conditioned projections and calibrated tails: AnalogReturnProjectionIndicator builds deterministic empirical neighbor forecasts, and RollingConformalForecastProjectionIndicator widens tails from matured residuals without changing base provenance.
• Composable rough-volatility state: RoughVolatilityForecastStateIndicator keeps canonical return moments while adding bounded roughness, vol-of-vol, cumulative horizon variance, and an explicit analog feature schema.
• Bayesian regime state: OnlineChangePointForecastStateIndicator adds canonical constant-hazard run-length inference, window-qualified recent-change posterior mass, typed component summaries, and direct analog composition.
• Serialization named shorthand preview: The companion ta4j PR #1507 targets 0.23.1 with compact expressions such as SMA(7,21), SmaCrossUp(7,21), RSI(14), and SharpeRatio while keeping canonical descriptor JSON as the durable storage format.
• Forecast migration required: The 0.23.1 correction intentionally replaces the forecast API introduced in 0.23.0. See Migration and Version Compatibility.

Start Here

• Primary onboarding lane: Getting Started -> Usage Examples -> Backtesting -> Live Trading -> Live Trading Runbook -> Troubleshooting Hub
• Getting Started - Install ta4j, build a strategy, and pick the right driver
• Canonical User Journey - Follow the end-to-end production path from data ingestion to live operations
• Backtesting - BarSeriesManager, BacktestExecutor, supplied records, and manual simulation loops
• Live Trading - Event-driven live or paper flows with BaseTradingRecord
• Usage Examples - Runnable examples, including parity and bot loops
• Serialization and Named Shorthand - Preview guide for ta4j PR #1507 strategy, rule, indicator, and criterion serialization
• Execution Decision Matrix - Choose execution and simulation path by workload
• Migration and Version Compatibility - Preferred APIs and incremental migration guidance
• Release Notes - Version-by-version changelog and migration notes

Core Concepts

• Bar Series & Bars - OHLCV data, aggregation, moving windows, and streaming updates
• Data Sources - Loading bars or trades from files and HTTP providers
• Num - Precision-aware numeric types such as DoubleNum and DecimalNum
• Technical Indicators - Indicator composition and caching
• Forecast Indicators - Forward-looking return and price distributions, point projections, and strategy filters
• Forecast State Estimation - Minimal lifecycle state, canonical, rough-volatility, and Bayesian regime moments, representation-bound schemas, and provenance-aware summaries
• Forecast Projection Models - Analog neighbors, rolling conformal calibration, maturity guards, tuning, and failure behavior
• Trading Strategies - Rules, strategies, unstable bars, and serialization
• Serialization and Named Shorthand - Preview guide for canonical JSON, compact strategy JSON v2, and named asset expressions
• Charting - Visual overlays, trading-record rendering, and analysis charts

Pick The Right Execution Path

Use the canonical matrix for execution choices:

• Execution Decision Matrix for driver and simulation-path selection
• Backtesting for implementation patterns once you choose a backtest path
• Live Trading for fill-driven and event-driven live loops

Where To Go Next

• Analysis Criteria and Risk Metrics - Rolling criteria, risk-adjusted return, open exposure, and risk-unit scoring
• Walk-Forward Research - Strategy walk-forward runs and generic prediction research workflows

Community

• Chat on the ta4j Discord
• Explore the ta4j repository and its ta4j-examples module
• How to Contribute - Development setup and contribution workflow
• Roadmap & Tasks - Planned work and known gaps
• Related Projects - Ecosystem projects built around ta4j

Maintainer design docs

• Architecture proposals - Active design drafts and in-progress PRDs
• Architecture archive - Historical decisions and implementation records
• Completed feature dossiers - Delivered PRD/checklist records