Algo trading software development
Backtest, paper trade, and go live on the same code.
Algo trading software turns a strategy into orders automatically: research and backtesting on historical ticks, paper trading, then live execution with risk checks. Acasa Labs builds it. The options platform we built for a confidential client runs research, paper, and live on one code path, so a strategy behaves the same way at every step.
The problem
The backtest was great. Live was different.
Many algo strategies die in the gap between backtest and live. The backtest ran on different data, through different code, with fills that were too kind. Then the live version is rewritten for the broker, and the two drift apart with every fix. When a loss comes in, the trader cannot tell whether it came from the strategy or from the plumbing. The fix is one code path and one dataset, from the first test to the first live order.
What we build
- 01
A tick archive for research
Every tick stored in Parquet and queried with DuckDB and Polars, so research runs on the data the live system recorded.
- 02
Backtests and run comparison
Strategies replayed over the archive, with a dashboard for comparing runs side by side.
- 03
Paper trading on the live path
Paper runs go through the same engine and risk checks as live, against real market data.
- 04
Visual strategy building
A strategy builder in the dashboard, so the desk can assemble and test ideas without waiting for a developer.
From production
How the options platform handles strategies.
One code path, three modes
Options trading platformThe client's core requirement was one system for research, paper trading, and execution, with the same code path in each, so a strategy is never rewritten on its way to production.
Live orders behind several gates
Options trading platformPosition caps, a per-bar guard, and a kill switch sit in front of execution, and live orders must pass several independent gates. The kill switch stops new risk without dropping the data stream.
Order logic that can move to Rust
Options trading platformOrder logic is kept free of engine imports, so a strategy proven in Python can be ported to Rust without untangling it from the engine first.
Signal to order in about 0.5 ms
Options trading platformMeasured in tape replay, the platform goes from signal to order created in about half a millisecond, with latency tracked across 12 stages.
Honest trade-offs
When you do not need custom algo software
If your strategies run on one broker and its API does what you need, a retail algo platform or the broker's own tools may be enough, and cheaper. Custom software earns its cost when you trade several strategies, need tick-level backtests you trust, want paper and live to match exactly, or have risk rules that off-the-shelf tools cannot enforce. We do not write strategies; we build the system yours run on.
What we reach for
- Python
- Rust-core event engine
- Parquet
- DuckDB
- Polars
- Redis
- Next.js
- Prometheus, Grafana
Proof
Shipped, not pitched.
01Client work
Options trading platform
Microsecond market-data handling for index options, with risk checks that never sleep.
A private trading platform for index options on an Indian exchange: tick capture, research and backtesting, paper and live execution, and a control dashboard. Built for a client's internal use; details are anonymized.
Trading engine · Realtime gateway · Web dashboard
Questions
Algo trading software questions we get asked.
Do you write trading strategies?
No. We build the platform: data, backtests, paper and live execution, and risk controls. The strategies are yours, and so is the code that runs them.
Can you backtest on tick data?
Yes. The platform archives ticks in Parquet and runs research with DuckDB and Polars over that archive.
How do you stop a runaway strategy?
Position caps, a per-bar guard, and a kill switch that stops new risk while the data keeps flowing, plus several independent gates before any live order.
Does a strategy change between paper and live?
Not on one code path. Paper and live run through the same engine and risk checks, so the strategy code is the same in both.
Do you work with options and futures?
The platform we built trades index options on an Indian exchange. The same engine design applies to futures and equities; the broker adapter and the instruments change.
Algo trading software development
Show us the gap between your backtest and live.
Tell us how strategies move from research to orders today. We will tell you where the two drift apart and how we would close it.