ArcticDB is a dataframe database for Python from Man Group. It runs inside your Python process and keeps data in S3, Azure or local files. Most alternatives replace the database. If your real need is a backtest, the alternative may be a different kind of tool.
Why teams look for ArcticDB alternatives
- Its licence is free only for non-commercial, personal or academic use. Business use needs a paid licence, including research and development, and prices are by quote.
- They want SQL, or a database other languages can query, not one tied to Python.
- They use it mainly to backtest, and have to write the fills and costs themselves.
ArcticDB alternatives compared
Details below were checked against each project's own pages in November 2026.
| Tool | Type | Query language | Best for | Licence |
|---|---|---|---|---|
| kdb+ and KDB-X | Time-series database | q, Python, SQL | Very large tick data | Free Community Edition with limits, paid above |
| QuestDB | Time-series database | SQL | Market data with SQL | Apache 2.0, paid enterprise edition |
| ClickHouse | Columnar database | SQL | General analytics at scale | Apache 2.0, paid cloud and enterprise |
| DolphinDB | Time-series database | Own scripting language | Data and order-book backtests in one system | Proprietary, free community edition with limits |
| TimescaleDB | PostgreSQL extension | SQL | Teams already on PostgreSQL | Apache 2.0 core, Timescale License for other features |
| Parquet with DuckDB | Files with an in-process SQL engine | SQL, Python | Plain files with no database to run | Open source |
| Reamer Research | Backtesting engine, not a database | Python, C++ or C | Backtests of strategies on bars | $20 per seat per month |
Databases
- kdb+ and KDB-X from KX have long been used for tick data in capital markets. The free KDB-X Community Edition is limited to 16 GB of memory for q. See What are the best kdb+ alternatives in 2026?
- QuestDB is open source and names capital markets as a main use. It has SQL, nanosecond timestamps, order book analytics and materialized views from ticks to OHLCV bars.
- ClickHouse is a general columnar database, so market data functions and conventions are yours to add.
- DolphinDB is a database with its own language and a backtest plugin with a matching-engine simulator. See What are the best DolphinDB alternatives in 2026?
- TimescaleDB extends PostgreSQL, so existing tools and skills carry over.
- Parquet files with DuckDB are the closest match to ArcticDB's serverless shape. DuckDB runs inside your process and queries Parquet files with SQL, but versions and snapshots are yours to manage.
If you use ArcticDB to backtest
A database stores and reads your data. It does not decide how orders fill. ArcticDB has no fill model, cost model or order handling, so a backtest on it is your own code.
Reamer Research is an engine for that part, for strategies on time, tick, volume, dollar or custom bars. Bid and ask, spread, slippage, commission, swap, margin and order lifetimes are in place under a public execution specification. It stores no data, so you can keep ArcticDB or any database here and pass in bars. It works on bars, not ticks, and runs on Linux, macOS and Windows through WSL2. See How does Reamer Research compare with ArcticDB?
How to choose
| If you want | Look at |
|---|---|
| Versioned dataframes from Python | Stay on ArcticDB |
| Very large tick data, with q skills on the team | kdb+ or KDB-X |
| Open source with SQL, for market data | QuestDB |
| General analytics at scale | ClickHouse |
| A database and order-book backtests in one system | DolphinDB |
| To stay on PostgreSQL | TimescaleDB |
| Plain files, no database to run | Parquet with DuckDB |
| Modelled fills and costs for strategies on bars | Reamer Research, with any database |
See Who are Reamer Research and Reamer Server for? Reamer Research is $20 a month, cancel any time; see pricing.