| Symbol | Direction | Date/Time (EST) | Entry | TP | SL | Confidence | Engine | Status |
|---|---|---|---|---|---|---|---|---|
| Loading picks... | ||||||||
| Symbol | Direction | Opened (EST) | Closed (EST) | Entry | Exit | Result | P/L % | Engine |
|---|---|---|---|---|---|---|---|---|
| No closed picks yet | ||||||||
GENESIS evolves mathematical formula trees using strongly-typed genetic programming. Each individual is an expression tree where leaf nodes are market indicators (RSI, MACD, ATR, volume ratios, price deltas) and internal nodes are operators (+, -, *, /, max, min, conditional). Trees are initialized via ramped half-and-half (depth 2-6), then evolved through:
Population of 200, 50 generations per run. Elite preservation (top 5%). Bloat control via depth limit of 12 and parsimony pressure.
ATLAS maintains a discretized archive (behavior map) where each cell holds the highest-fitness strategy discovered for that behavior niche. The behavior space is defined by two axes:
This gives a 10x10 = 100-cell archive. New strategies are generated by mutating archive occupants. If a new strategy's fitness exceeds the current occupant of its behavior cell, it replaces it. Coverage = filled cells / 100. This ensures diversity: ATLAS finds good scalpers AND good swing traders AND everything in between.
NEXUS evolves softmax weight vectors over the pool of all validated strategies from every engine. Each individual is a weight vector w[i] where the ensemble signal = sum(softmax(w) * signal[i]). Evolution uses CMA-ES (Covariance Matrix Adaptation) with:
Walk-forward validation: 70% train / 30% test, rolling 6-month windows. Only strategies with positive Sharpe on test set survive the audit.
LEGION coevolves teams of 3-7 strategies simultaneously. Each individual is a team composition (which strategies to include + allocation weights). Two populations evolve in parallel:
Fitness is a multi-objective: maximize Sharpe, minimize max drawdown, maximize win rate. NSGA-II non-dominated sorting selects Pareto-optimal teams. Coevolutionary pressure ensures strategies that work well together are preserved even if individually mediocre.
PHOENIX analyzes closed losing picks to diagnose failure modes: stop-hunted (SL hit then price reversed), late-entry (signal lagged the move), overconfident (high confidence but wrong), trend-reversal (fought the trend), and regime-blind (wrong market conditions). For each failure pattern, it evolves corrective variants with targeted mutations — wider stops for stop-hunted, momentum filters for late-entry, confidence caps for overconfident. Variants are backtested and only promoted if they outperform the parent strategy.
GENESIS is our formula invention lab. It starts with random mathematical formulas combining market indicators, then breeds and mutates the best-performing ones over many generations — like natural selection for trading strategies. After 50 generations, only the most profitable formulas survive. It discovers patterns no human would think to look for.
ATLAS is our diversity engine. While other engines optimize for a single "best" strategy, ATLAS finds the best strategy for every type of market behavior — fast trades, slow trades, calm markets, volatile markets. It fills a map of possibilities so we always have the right tool for the current conditions. Think of it as building a complete toolkit, not just the sharpest knife.
NEXUS is our portfolio architect. It takes all the strategies found by GENESIS, ATLAS, and external sources, then figures out the optimal way to combine them. How much weight should each strategy get? NEXUS evolves the answer, ensuring the combined portfolio outperforms any individual strategy while reducing risk through diversification.
LEGION builds dream teams. It evolves groups of 3-7 strategies that complement each other — when one zigs, another zags. The result is teams with smoother returns, smaller drawdowns, and higher consistency than any single strategy. LEGION finds synergies between strategies that only emerge when they trade together.
PHOENIX is our failure analyst. When strategies lose money, PHOENIX doesn't just discard them — it performs a post-mortem to understand why they failed. Was the stop too tight? Was the entry too late? Was the model overconfident? Then it creates improved versions that specifically address each weakness. Think of it as a coach reviewing game tape to fix what went wrong.
Imagine you want to breed the ultimate racing dog. You start with 200 random dogs, race them, and let the fastest ones have puppies. Then you race the puppies, breed the winners again, and repeat 50 times. After all that, you get some seriously fast dogs that nobody could have designed on purpose. GENESIS does exactly this, but with trading formulas instead of dogs. It breeds math equations and keeps the ones that make money.
Instead of finding just one strong Pokemon, ATLAS tries to collect the best one of every type — the best fire-type, the best water-type, the best grass-type. In trading terms, it finds the best strategy for quick trades, the best for slow trades, the best for crazy markets, the best for boring markets. That way, no matter what the market throws at us, we have the right strategy in our collection ready to go.
You know how in fantasy sports, having the best individual players doesn't guarantee the best team? You need balance — a good QB AND good receivers AND a solid defense. NEXUS looks at all our strategies and figures out the best "roster" — how much to rely on each one. It keeps tweaking the lineup until the combined team performs better than any star player alone.
A great band isn't just 4 amazing soloists — it's musicians who sound great together. LEGION holds auditions: it throws random groups of strategies together, sees which groups harmonize, then mixes and matches until it finds the ultimate supergroup. Sometimes a mediocre strategy becomes amazing when paired with the right partners, and LEGION discovers those hidden synergies.
After a football team loses, the coaches watch the game film to figure out exactly what went wrong. Was it a bad play call? Did the defense break down? PHOENIX does the same thing for our trading strategies. When a strategy loses money, PHOENIX watches the replay — was the stop too close? Did we enter too late? Were we too confident? Then it creates a new version of the strategy with those specific problems fixed. It literally rises from the ashes of failure to become stronger.