aidensystematic Live AI

The mutation engine

We breed strategies, then we try to kill them.

A genetic algorithm searches the strategy space like evolution searches biology — generate, score, select, mutate, repeat — across 25 markets. Millions of simulated trades every sweep — and we publish the real, current count below. The other half of the system exists to stop us fooling ourselves.

Loading live counts…

1 · Genome

Each strategy is a set of genes — entry, stop, target, session, scoring weights. One fully-specified rule.

2 · Fitness

Backtested causally; scored on risk-adjusted return minus a complexity penalty — we punish knobs, because knobs overfit.

3 · Selection

Tournament selection: winners breed, the weak die. Elitism preserves the best each generation.

4 · Crossover

Two parents swap genes — combining one's entry timing with another's exit management.

5 · Mutation

Failure-targeted: diagnose the champion's dominant failure mode and bias mutation toward the genes responsible.

6 · Repeat

Dozens of generations, many runs, many markets — converging on high-fitness regions of the space.

⚠️ The honest part nobody else tells you

A mutation engine is an overfitting machine by default — enough genomes and it always finds something that looks perfect on the past. That's true of every optimizer, including our competitors'. The discipline is what happens after the search:

A champion with a flawless record over three trades on one day is the engine finding luck, not edge. See how we treat champions →