Chimera Scanner: 1,221 Patterns Detected From 51 Million Historical Occurrences

The Chimera Scanner is the output of a GPU-trained pattern recognition system. 1,221 distinct patterns, extracted from 51 million historical occurrences across 27,000+ training entries. The result is a 59MB neural brain that scans live market data for pattern matches.

The Numbers in Context

1,221 patterns is not a marketing number. It is the count of distinct pattern signatures the Chimera brain recognizes after training. These range from common formations (double tops, head-and-shoulders) to complex multi-candle sequences that do not have traditional names.

51 million occurrences is the training dataset. Every pattern was validated against historical price action across multiple assets and timeframes. A pattern that appeared once is noise. A pattern that appeared thousands of times with consistent outcomes is a signal candidate.

27,000+ entries refers to the labeled training data used to build the pattern library. Each entry includes the pattern formation, the market context, and the subsequent price movement.

59MB is the compressed size of the neural brain file. For comparison, a typical LSTM model for trading is under 1MB. The Chimera brain is large because it stores not just pattern signatures but their statistical profiles: frequency, reliability, average return, and regime-specific performance.

How It Works

The Chimera brain was trained on a local RTX 4080 GPU using brute-force backtesting across hundreds of thousands of parameter combinations. This is not a pre-trained model downloaded from a repository. It was built specifically for Strategy Arena's asset universe.

In production, the scanner does the following:

1. Ingests live OHLCV data for the active asset (BTC, ETH, SOL, etc.)
2. Runs pattern matching against the 1,221 stored signatures
3. Returns matches with confidence scores based on how closely current price action resembles the historical pattern
4. Provides context — how often this pattern has appeared before, what happened next on average, and whether the current market regime is favorable for this pattern

The scanner does not trade directly. It feeds information into the Chimera strategy, which combines pattern signals with position sizing and risk management to generate actual trades.

What Kinds of Patterns

The 1,221 patterns fall into categories:

Classic chart patterns — Triangles, wedges, flags, channels. The patterns every TA textbook covers. These are included because they work often enough to be statistically significant, even if they are well-known.

Candlestick sequences — Multi-candle formations that indicate momentum shifts, exhaustion, or continuation. Beyond the standard doji-hammer-engulfing patterns that most scanners detect.

Volume-price divergences — Situations where volume behavior contradicts price behavior. These are harder to spot visually but show up clearly in pattern matching.

Cross-timeframe patterns — Formations that only become visible when combining data from multiple timeframes. A 4-hour chart pattern confirmed by daily volume, for example.

Regime-specific patterns — Patterns that only work in trending markets, or only in ranging markets. The Chimera brain tags each pattern with its optimal regime, so it does not blindly signal a trending-market pattern during a sideways consolidation.

Performance Transparency

Not every pattern is reliable. The scanner provides statistical context for every match:

• Hit rate: What percentage of historical occurrences led to the expected outcome
• Average return: The mean price movement after pattern completion
• Sample size: How many times this pattern appeared in training data
• Regime filter: Whether current market conditions favor this pattern

A pattern with a 55% hit rate across 10,000 occurrences is more trustworthy than one with an 80% hit rate across 50 occurrences. The scanner surfaces both numbers so you can judge for yourself.

Integration With the Broader System

The Chimera Scanner connects to several other Strategy Arena components:

• Dashboard: The Chimera strategy's live performance is tracked alongside all 50+ strategies
• Evolution Lab: Track how the Chimera brain's effectiveness changes over time as market conditions evolve
• Knowledge Graph: See how Chimera patterns relate to other strategies' indicator sets

The Honest Assessment

Pattern recognition in financial markets is a contested field. Academic research is split: some studies find statistically significant patterns, others find that patterns disappear once enough traders recognize them.

The Chimera Scanner does not claim to predict the future. It claims to find historical patterns in current data and provide statistical context. Whether those patterns repeat is a market question, not a technology question.

The 59MB brain is a tool, not a crystal ball. Use it alongside other signals, not instead of them.

FAQ

How often is the Chimera brain retrained?

The brain was trained via GPU brute-force on historical data. Retraining happens periodically when enough new data accumulates to justify the computational cost. The current brain version (V2) includes 1,221 patterns and 1,196 strategy variants.

Can I see which patterns are currently being detected?

Yes. The Chimera Scanner page shows active pattern matches in real time, including pattern name, confidence score, historical hit rate, and expected price movement.

Does the Chimera strategy use all 1,221 patterns for every trade?

No. The strategy applies regime filters and confidence thresholds. Only patterns that match the current market conditions with sufficient confidence contribute to trade signals. Most of the time, only a subset of patterns are active.