Trading Strategy Engine
Write the strategy with AI. Money still loves silence.
Trading Strategy Engine is a high-performance backtesting and trading engine for C++ and Python, delivered as a licensed binary library with examples and documentation. It is the harness around the model: position accounting through a reversal, exposure at the instant of the fill, the order refused before it is sent, and one yardstick applied to every candidate whatever wrote it. It runs on your own premises, so the model, the data and orders logic never leave them.
Who this is for
Trading Strategy Engine is a component, not a platform. It is bought by people who write their own strategies, own their own data and venue connections, and run their own machines — and who need the layer underneath all of that to be right.
Proprietary trading desks
Teams of two to twenty trading their own capital, writing their own strategies in C++ or Python, already holding the venue connections and the market data. What is missing is not another place to log in — it is the position accounting, the risk gate and the statistics underneath, built so they can be trusted.
Systematic funds running their own stack
Firms for whom research and production must agree: a number that appears in a backtest report has to hold when the same code trades. One codebase covers both here, and it deploys where your compliance already lets you deploy — inside your own perimeter.
Teams generating strategies with AI
Anyone now producing candidates faster than they can be trusted. Generation got cheap; verification did not. Every candidate is measured by the same yardstick here — the same statistics for a two-line crossover as for a deep learning model — which is what makes a hundred of them comparable at all.
Engineers who would otherwise build it in-house
Groups entirely capable of writing this themselves, who have counted what it costs to finish it and then keep it correct for years. Some are already writing one and run this underneath in the meantime, so the research does not stop while the in-house version is finished. This is the part nobody sees, everybody has to get right, and no one gets credit for.
Why not LEAN, and why not build it yourselfAudit trail: what the engine records
What the engine does
Being right about the money
A backtest that is wrong still prints a profit, and nothing on the surface tells you which one you are looking at. That is why the invariants underneath are not only implemented but tested one by one, each with a test that fails the moment it breaks. Correctness you cannot see from the outside has to be demonstrated.
Your data and your strategy stay yours
Fully on-premise: strategy logic, market data and order flow never leave your perimeter, and you decide which AI to connect — external or local. Inside the perimeter the boundary holds too: the decision module is separate from the order-routing pipeline, so a strategy plugs in as an opaque callable whose logic the engine itself never sees.
From Python to C++ in two calls
Research develops and backtests a strategy in Python, where the model tooling lives; technology trades it from C++. The strategy crosses in two calls — a save in Python and a load in C++ — with its contracts, inputs, patterns, rules and robot intact; only your own data processors are registered again under the same names. The teams hand over a saved recipe, not a system to rebuild, and the operational management of that handoff shrinks to a minimum.
High-performance C++ engine
The industry standard for professional trading, not a retail stack. Large datasets, complex multi-currency portfolios, and detailed behavioural statistics for every backtest and live run.
Every candidate measured the same way
The same statistics for a two-line crossover as for a deep learning model, and the per-period features a selection model ranks them with. That comparability is what begins to matter once candidates are generated by the thousand.
ML and AI ready
The same robot scaffold serves a century-old technical indicator and a neural network alike — only the data processor changes. Plug in GBM/XGBoost or call out to an external AI service.
Accounting is not your edge
A trading team is paid for two things: finding strategies that hold up, and pricing the risk it takes. The ledger underneath — position accounting, cost basis, exposure, the statistics that make two candidates comparable — is indispensable, invisible, and the same at every firm that has ever built one. Nobody ever won a mandate for having a better ledger, and building one in-house is months of work and a six-figure budget before it sees a live test. Spend your people on what the market actually pays for.
What the engine does not do
No charting. You take the data the library records and visualize it in the tools you already use.
This is not a product for someone looking for a strategy to run. There are no signals here, no managed accounts, no ready-made portfolios and no advice — the strategy is yours to write. It is not a hosted research environment: there is no browser IDE, no charting and nothing to log into, because the library is driven from your own code. And it is not for discretionary, manual trading: everything here assumes a program decides.
Solo 76 sells software, not financial services. The Trading Strategy Engine is a licensed binary library that runs on your own machines. Solo 76 is not a broker, an exchange, a fund or an investment adviser. It never takes custody of money or assets, never holds a customer account, never places an order on anyone's behalf, and gives no investment advice. It holds no financial-regulator licence because it provides no financial service: the software is a tool you run yourself, against venues and data feeds you choose and contract with directly.
Supported platforms
- Windows (x64)
- Ubuntu 26.04 LTS (other Linux on request)
Read the API overview Browse robot examples See the measured numbers