Prometheus Gen 1

Premium | By Prometheus Neural Systems ⚡ 50 Founding Copies — Price Rises 50% After Copy 50. Permanently.

The market doesn't repeat. Neither does Predator Pulse.

Most indicators were calibrated in the past and applied to the future, hoping nothing changes. Predator Pulse AI Trader was built on a different conviction entirely: that a signal engine serious about edge must not only read the market — it must learn from it, in real time, without stopping.

This is not an upgrade to Predator Technican Version, our free community tool. This is a different class of system.

Two Engines. One Signal.

At the base sits a 60-feature institutional microstructure engine built on the Jane Street production framework and the Lopez de Prado methodology. Sixty distinct, zero-redundancy market features — spanning order flow imbalance, VPIN toxic flow probability, Kyle's Lambda price impact, Hurst exponent, realised volatility, Volume GARCH clustering, Amihud illiquidity, and deep institutional microstructure decomposition — are scored across 7 dynamically weighted signal groups every bar. Cross-calibration EMAs adjust the relative influence of each group in real time as market conditions shift. The result is a composite that reflects what the market is actually doing, not what it was doing when the model was last tuned.

Running alongside it is Prometheus — an AI layer that changes everything.

Prometheus: What It Actually Does

Most "AI" indicators run a fixed trained model. Prometheus is not a model. It is a living intelligence architecture built from four interconnected systems that together do something genuinely different: they discover, evaluate, evolve, and remember.

The Pattern Library

Prometheus maintains a library of active trading patterns — but not candlestick patterns or preset chart shapes. These are symbolic mathematical expressions, parsed and evaluated as expression trees, each encoding a unique relationship between market features and expected returns. Each pattern carries its own structure, its own constants, and crucially, its own attention weights — a per-feature mask that tells the system which of the 60 market features this specific pattern is actually sensitive to. Patterns with diffuse attention are penalised; patterns that have learned to focus sharply on the right features are rewarded. The library self-selects for precision.

Before any pattern is added, it passes a polarity rejection test: expressions that produce always-positive or always-negative outputs across 20 random feature vectors are rejected as static biases rather than signals. Only patterns that genuinely encode directional intelligence enter the library.

The Math Engine and Feature Interactions

When a pattern is evaluated, Prometheus does not simply feed raw features into it. The Math Engine first generates second-order polynomial interaction features — cross-products between the base 60 features — expanding the signal space before the expression tree is applied. This means Prometheus can discover non-linear relationships between features that the base static engine cannot express. The attention mask then weights this expanded feature set, and the result is passed through the pattern's expression tree. Signal output is clamped to prevent overflow. Realised Sharpe is computed from actual forward returns, not simulated ones.

Online Learning: The Outcome Feedback Loop

Every time a signal closes in real trading, Prometheus records the actual normalised return. This is not a simulation — it is live outcome data from real trades on your chart. The pattern that contributed to that signal receives a Bayesian performance update using Welford's online algorithm: mean fitness and fitness variance are updated incrementally without storing all historical data. The learning is continuous, lightweight, and mathematically sound.

The Confidence-Fitness Function

This is where Prometheus becomes genuinely adaptive. Each pattern in the library carries a confidence-fitness score that combines multiple components: the evolution Sharpe from historical evaluation, the live trade Sharpe from actual closed outcomes, a consecutive loss penalty that penalises patterns on losing streaks, a consecutive win bonus that rewards patterns that are currently working, a staleness penalty that reduces the score of patterns not recently traded, a complexity penalty proportional to the expression tree depth (preventing overfitting to elaborate structures), and an attention bonus that rewards focused, high-entropy attention distributions.

This composite score is what determines which patterns survive and which are replaced. It is regime-aware: every pattern also tracks separate performance statistics for each of the five detected market regimes, so a pattern that works in trending markets is not unfairly penalised by its performance during consolidation.

Thompson Sampling Ensemble

When multiple patterns contribute to a signal, they are not simply averaged. Prometheus uses Thompson Sampling — a Bayesian probability matching algorithm — to assign ensemble weights. Each pattern draws a fitness sample from its uncertainty distribution, and weights are assigned proportionally. Patterns with high-confidence fitness dominate; patterns with high uncertainty are given less weight but remain in the ensemble in case their edge re-emerges. Correlation between patterns is also penalised: two patterns that encode similar views of the market are downweighted relative to a diverse ensemble, because diversity of perspective reduces variance in the composite signal.

Prometheus Evolution: Survival of the Fittest

Every 50 bars, Prometheus audits the entire pattern library against a minimum Sharpe threshold. Patterns that no longer meet the bar are pruned and replaced. Replacement candidates are tested for polarity before admission. The library self-purges and self-renews without any intervention from the trader.

Prometheus Persistence: Your AI Remembers

Pattern state, constants, attention weights, performance history, and regime centroids are saved to disk per symbol and per timeframe after every evolution cycle. When MT5 restarts, Prometheus loads the last saved state and continues from where it left off. The patterns learned on EURUSD H1 are never mixed with GBPUSD H4. Each instrument on each timeframe has its own evolved intelligence that compounds over time.

Signal Integrity

Before a signal reaches your chart it passes through a full chain: composite score floor, confidence floor, saturation decay gate, a circuit breaker that raises the confidence bar during loss streaks, a per-timeframe cooldown, and a minimum price move filter. Every layer exists to protect you from the noise that most indicators fire on without hesitation.

TP and SL levels are ATR-calibrated per timeframe — not fixed multiples. Three risk modes (Tight, Normal, Wide) give you further control. TP1 and TP2 are drawn live alongside the entry line.

The on-chart Canvas dashboard displays session stats, win/loss ratio, net pips, TP1 and TP2 hit counts, active signal status, and the live Prometheus equation for the currently active pattern — so you can see, in real terms, what the AI is acting on. Every signal logs automatically to CSV. Alerts run across popup, sound, push, email, and Telegram.

The Founding Price. 50 Copies. Then It's Gone.

Predator Pulse AI Trader is available right now at the founding price. When copy 50 is claimed, the price rises by 50% — permanently and immediately. No countdown. No second window.

If you have been running Predator Technican Version — our free gift from The Sons — you have already seen what institutional signal intelligence looks like on a chart. Prometheus is what comes next: a system that doesn't just read the market but builds its own understanding of it, bar by bar, trade by trade, and gets measurably better over time.

50 copies at the founding price. The only question is which side of copy 50 you land on.

The model comes with a trainer EA as an early access premium. On purchase message Humphrey for additional guide and information. Happy trading!!

MetaTrader 5 | All symbols and timeframes | All required libraries included in the package

This indicator provides signals and analysis for informational purposes. Past performance is not indicative of future results. Trade responsibly.