Architecture guide
Inside a simulated matching engine.
Key takeaways
- Determinism matters more than raw speed: the same inputs must produce the same outputs, reproducibly.
- Every fill should be reconstructible months later from retained inputs, or disputes cannot be settled.
- Ask for a demonstration of reconstruction, not a description of the architecture.
Why simulated execution is a product decision
Evaluation programs do not route trader orders to a live venue; they match against a simulated engine using licensed market data. That is not a shortcut — it is what allows a firm to offer an evaluation at all without taking on the obligations of a broker.
But it makes the engine's behaviour part of your product. If a trader believes a fill was wrong, the firm is defending its own software, not a third-party venue's.
Determinism, and why it beats latency claims
A deterministic engine produces the same result given the same inputs, every time. That property is what makes a dispute resolvable: you can replay the moment and show what happened.
Vendors more often lead with latency figures, which are easier to market and harder to interpret. A fast engine whose behaviour cannot be reproduced is worse for an evaluation business than a slower one that can be. Read how to measure latency before accepting any published number.
Properties worth testing
Reproducibility
Replay a historical session and confirm the fills match what the trader originally received.
Order-type behaviour
Confirm how stops, limits and brackets behave at gaps and around the open, where evaluations are most often contested.
Input retention
Confirm the market data that produced a fill is retained long enough to reconstruct it within your dispute window.
Failure behaviour
Confirm what happens to working orders and open positions during an incident, and how that is communicated.
Keep reading
Back to hub
Infrastructure & data
How prop firm infrastructure is architected: tenancy boundaries, matching and execution, market data licensing and how performance should be measured.
ReadRelated
Single-tenant architecture
What tenancy actually determines in a prop firm platform: failure blast radius, data boundaries, release control and which risks isolation does not solve.
ReadRelated
Measuring latency
A methodology for measuring platform latency honestly: what to instrument, why percentiles beat averages, and how to read a vendor's published numbers.
ReadNext step
Nexus Trading
Explore Nexus Trading: a white-label platform with advanced charts, simulated order execution and connected risk controls for funding evaluation programs.
ReadTest reconstruction against a real session.
Technical validation calls are with engineers, and reconstruction is the demonstration worth asking for.