Benchmarks
Compare trading research and backtesting tools on correctness-matched workloads, with exact runtimes, interactive charts, methods, and downloadable results.
Choose by what you want to measure
Start with the question closest to your own work. You can browse every benchmark below.
Run the same strategy across engines
See how general-purpose backtesters scale when they produce the same trades and final account value.
View benchmark Explore many settingsScale a parameter search
Compare what happens when one strategy expands from a few settings to a large research grid.
View benchmark Work with many assetsRebalance a portfolio
Measure shared-cash portfolio allocation as the number of assets grows.
View benchmark Model detailed eventsProcess orders and state changes
Compare ordered market and account events while preserving the same order behavior.
View benchmarkCompare backtests as the workload grows
Compare common backtesting jobs as the amount of data, number of settings, or portfolio size grows.
Single-asset bar simulation
The same long-only strategy, fill timing, costs, and final equity across general-purpose backtesters.
View benchmarkParameter-grid scaling
Runtime and memory as one deterministic strategy expands from a few settings to a large research grid.
View benchmarkPortfolio rebalancing
Shared-cash, multi-asset allocation with identical rebalance dates, sizing, and transaction costs.
View benchmarkMeasure complete trading tasks
Explore workloads built around ordered events, complete crypto-bot flows, and reinforcement-learning environments.
Stateful event and order processing
Ordered market and account events with an explicit record of every state transition.
View benchmarkCrypto candle bots
Credential-free candle backtests with pinned local data, costs, and cache behavior.
View benchmarkReinforcement-learning environments
Deterministic environment steps across increasing asset counts, reported as scaling rather than a leaderboard.
View benchmarkInspect the numerical core
Inspect focused numerical work where implementation details and compilation strategy matter.
Numba and Rust kernels
Correctness-checked VectorBT PRO kernels grouped by input size, with compilation excluded from warm runtime.
View benchmarkSee which tools are included
Each tool appears only in benchmarks that fit the way it works.
Fair comparisons first
Results are published only after the tools produce the same answer under the same conditions.
- 1Same job
Use the same data, strategy rules, costs, and sizing.
- 2Same answer
Check fills, positions, cash, and equity against independent reference code. Reference timings are not included in comparison charts.
- 3Clear evidence
Make the underlying values and test details available to download.