Evolutionary Quant Arena

Strategies that evolve.
Only the fittest trade.

Darquant breeds, mutates and backtests thousands of trading strategies every generation. Statistical gates kill the lucky ones. Survivors compete in a live arena — and the strongest lineage takes capital.

Arena · Leaderboard LIVE
generation #1,284 · 9,600 genomes evaluated
01trend-hma-0x7F10.96
02mean-rev-0xA3C0.91
03vol-breakout-0x2E0.88
04regime-tsi-0xB90.84
05carry-supertrend-0x440.79
01The evolution loop

A genetic algorithm that runs your strategy book like a species.

Every cycle, code-level strategies are crossed and mutated, stress-tested out-of-sample, and ranked by multi-objective fitness. Weak lineages die. Strong ones reproduce and graduate to live capital.

→ GEN

Generate

LLM-driven crossover & mutation produce complete, novel strategy code each generation.

→ TEST

Backtest

Backtrader runs every genome with walk-forward windows and lookahead-bias audits.

→ GATE

Validate

PSR / DSR significance gates reject strategies that are merely lucky, not skillful.

→ SELECT

Select

NSGA-II ranks survivors on return, drawdown & robustness — a Pareto front, not one metric.

→ PROMOTE

Promote

The fittest lineage enters the live arena under shadow mode, then real capital.

02The Arena

Strategies don't get deployed. They earn it.

The Arena is a continuous tournament. Capital is normalized across competitors, exposure is aggregated, and a deterministic curator retires underperformers with a full soft-delete audit trail. No black-box deletions — every promotion and culling is logged.

Circuit breakers, slippage tracking and collision detection keep the live book honest. Metrics stream to Prometheus & Grafana in real time.

Capital allocation · liveLIVE
notional normalized · 5 active lineages
trend-hma-0x7F142%
·mean-rev-0xA3C27%
·vol-breakout-0x2E18%
·regime-tsi-0xB913%
03The engine

Quant rigor under the hood.

🧬

Multi-objective GA

NSGA-II evolves a Pareto front across return, risk and robustness — never overfitting to a single number.

NSGA-II · Pareto
📊

Statistical validity gates

Probabilistic & Deflated Sharpe Ratios separate genuine edge from multiple-testing luck.

PSR · DSR

Walk-forward & OOS

Rolling out-of-sample validation with in-sample/out-of-sample consistency checks on every genome.

walk-forward
🌐

Regime-aware

Hidden Markov regime detection seeds the population so strategies adapt to changing market states.

HMM · 6–8 regimes
🛡️

Safe live trading

Shadow mode, circuit breakers, aggregate-exposure limits and collision detection before a cent is risked.

shadow · breakers
🗂️

Deterministic curation

Scheduled promotion & retirement with soft-delete audit logs — reproducible, never a black box.

Celery · audit
0+
genomes / generation
0%
PSR significance gate
24/7
continuous arena
0%
auditable decisions
Private beta

Put natural selection to work on your book.

Darquant is opening a small private beta for systematic traders and quant researchers. Request access below.

No spam · early cohorts onboarded first

Darquant is research & trading-automation software. Nothing here is financial advice. Algorithmic trading involves substantial risk of loss; past or simulated performance does not guarantee future results.