OPN-201 · TRADING METHODOLOGY

Zenith 4.0
Trading Methodology

A fully autonomous AI agent for crypto derivatives trading. Zenith perceives market state, reasons about regime and risk, makes its own trading decisions, and learns from every outcome — running continuously without human intervention.

What Makes Zenith an Agent

The difference between a trading bot and a trading agent is autonomy of judgment.

A bot executes rules: if indicator crosses threshold, buy. If stop-loss hit, sell. The logic is static. A human wrote the rules, and the bot follows them. When market conditions change, the bot doesn't know — it keeps executing the same rules until a human updates them.

Zenith operates differently. It continuously perceives its environment across multiple data streams — price action, technical indicators, social sentiment, news, liquidity depth. It forms its own assessment of what kind of market it is in. It decides whether to act or to stand aside. It manages positions through their full lifecycle. And after every trade, it feeds the outcome back into a reinforcement learning layer that adjusts its own decision parameters — getting better at distinguishing conditions that favour action from conditions that favour restraint.

This is what makes it agentic: Zenith does not wait for instructions. It sets its own operational rhythm, makes its own risk judgments, and improves its own performance over time. The system has been running autonomously since January 2026 — every trade on the live dashboard was identified, sized, entered, managed, and exited by the agent without human input.

Three Core Pillars
01
Regime Classification
Determining market state before trading.
+
The foundation of every decision Zenith makes is a regime classification: is the current market trending or mean-reverting? The answer determines the entire strategy — trend-following setups in trending markets, reversion entries in range-bound conditions.

This classification is built from price action analysis and a battery of technical indicators evaluated across multiple timeframes. Zenith does not use a single indicator or a fixed rule set — it synthesises directional, momentum, and volatility signals to form a probabilistic regime assessment that updates continuously.
Multi-timeframe price action analysis for directional structure
Momentum and volatility indicators for regime probability scoring
Dynamic regime switching — the system adapts its strategy type as market conditions evolve
02
Sentiment & Macro Screening
Filtering trades against real-world risk.
+
Technical signals operate in a vacuum. Markets do not. Before committing to any position, Zenith screens real-time sentiment from social channels, news sources, and on-chain activity to assess whether macro conditions support the trade thesis.

This is a risk filter, not a signal generator. Zenith uses sentiment to avoid bad trades — identifying macro headwinds, sudden narrative shifts, or event-driven risk that technical analysis alone would miss. If sentiment contradicts the technical setup, the system stands aside.
Real-time sentiment analysis across social media, news feeds, and crypto-native channels
Macro risk detection — regulatory events, major liquidation cascades, narrative regime breaks
Directional sentiment cross-check — ensures technical signals are not fighting the prevailing macro environment
03
Liquidity-Aware Risk Management
Structuring every trade around defined risk.
+
Every position Zenith takes has a pre-defined stop-loss, take-profit, and position size — determined before entry, not adjusted after the fact. These parameters are calculated from technical levels, risk-reward ratios, and liquidity heatmap data.

Liquidity heatmaps show where resting orders cluster — where stops are likely stacked, where large bids or asks sit. Zenith uses this information to place stops at structurally sound levels rather than arbitrary percentages, and to target exits where liquidity is sufficient to absorb the position.
Stop-loss and take-profit levels set by technical structure, not fixed percentages
Risk-reward ratio evaluation before every entry — minimum threshold required to proceed
Liquidity heatmap integration — position sizing and exit targeting informed by observable order depth
Per-trade risk budgeting — no single position exceeds defined exposure limits
Reinforcement Learning — How the Agent Learns

The difference between automation and agency is the ability to improve.

A static system trades the same way on day one and day three hundred. Zenith does not. A reinforcement learning layer continuously evaluates trading outcomes and adjusts the agent's decision parameters — not the core logic, but the calibration of that logic against current market reality.

After every trade, the RL layer asks: did this parameter configuration produce a good risk-adjusted outcome? It rewards configurations that worked and penalises those that didn't. Over time, this shapes the agent's behaviour — it learns which regime signals to weight more heavily, how aggressive to size in different conditions, how sensitive to make the sentiment filter, and how strong a signal needs to be before committing capital.

Signal weightsWhich indicators matter more under current conditions
Entry thresholdsHow strong a signal must be before the agent commits capital
Sizing parametersHow aggressively to size given recent regime stability
Filter sensitivityHow much weight sentiment carries relative to technical conviction

This is what separates Zenith from a trading bot with fixed parameters. The core logic — regime classification, sentiment screening, structured risk management — remains stable and interpretable. But the calibration of that logic evolves continuously, driven by the agent's own experience. No human recalibration required.

The Agent Loop

Zenith runs a continuous perception-judgment-action-learning cycle. This is not a pipeline that fires once per signal — it is an ongoing cognitive loop that the agent runs 24/7, deciding at every iteration whether to act, hold, or stand aside.

01Perceive
Continuously ingest price action, technical indicators, sentiment data, and liquidity depth across all traded assets.
02Judge
Classify market regime (trending or mean-reverting). Screen sentiment for macro risk. Determine whether conditions warrant a new position, require adjusting an existing one, or favour doing nothing.
03Size
If acting: calculate position size, stop-loss, and take-profit from technical levels, risk-reward ratio, and liquidity heatmap.
04Execute
Route order autonomously. Monitor the position through its full lifecycle — adjusting or closing based on evolving conditions and predefined exit logic.
05Learn
Feed the trade outcome back to the RL layer. Update parameter weights. The agent's next iteration of this loop is informed by everything it has done before.

The loop runs continuously. Most iterations result in no action — the agent is perceiving and judging but choosing to wait. This is a feature, not a limitation. The ability to decide not to trade is as important as the ability to decide when to trade.

Development Timeline
Jun — Aug 2024Prototype
Core technical analysis engine built. Backtesting framework. Rule-based entries with basic risk management.
Sep — Nov 2024v1.0
Regime classification introduced — system determines market state (trending vs mean-reverting) before selecting strategy. First live trades.
Dec 2024 — Feb 2025v2.0
Sentiment screening layer added. Technical signals filtered against social and news sentiment. Liquidity heatmap integration for stop/target placement.
Mar — Nov 2025v3.0
Reinforcement learning refinement layer added. System self-optimises signal weights, entry thresholds, and sizing parameters against live performance. Expanded asset coverage.
Jan 2026v4.0
Public launch on Operon Network. Full autonomous operation with live dashboard and transparent trade history.

Zenith was built from scratch starting in June 2024. Each version addressed a specific limitation of its predecessor — from static rules to regime-aware strategy selection to RL-driven adaptation.

What Zenith Trades
BTC
Perpetual
ETH
Perpetual
XRP
Perpetual
SOL
Perpetual
BNB
Perpetual

Strategy period: January 2026 — present. All trades executed autonomously. Full trade history with entry/exit prices available on the live dashboard.

← Back to Zenith