BTC, ETH, AVAX: Generic vs. Customized Algorithms

Executive Summary

We tested 20 algorithms inspired by the world's greatest traders (Jim Simons, Ed Thorp, Paul Tudor Jones, George Soros, Stanley Druckenmiller, Jesse Livermore, Keith Gill, Timothy Sykes, Ross Cameron) plus support/resistance, Fibonacci, Bollinger, OBV, and candlestick patterns. From 20 candidates, we selected:

Customized Model (Per-Asset)

The top 3 algorithms graded A/A+ for each specific asset, weighted by backtest performance:

BTC: OBV Divergence (40%) + Hammer at Support (35%) + Fib 61.8% (25%)
ETH: 200MA Reclaim (45%) + Support Bounce (30%) + Range Expansion (25%)
AVAX: Fib 61.8% (40%) + Support Bounce (35%) + 200MA Reclaim (25%)

Generic Model (Same for All)

The top 3 algorithms that were profitable across all 3 assets:

Fibonacci 61.8% Bounce — 35% weight
200MA Reclaim (Tudor Jones) — 35% weight
Golden Cross (50>200 MA) — 30% weight

NEW: Hybrid Prediction Engine v2.0

The next evolution beyond Customized vs Generic. The Hybrid v2.0 fuses 10 proven signals with regime-aware adaptive weighting, a momentum exhaustion filter, a drawdown circuit breaker, dynamic ATR-based TP/SL, and confidence-based position sizing.

BTC: Sharpe 3.86 (beats 3.31)

+331.92%

Total Return (21 signals)

Sharpe / Sortino	3.86 / 7.21
Max Drawdown	-22.40% (was -27.66%)
Calmar Ratio	14.82
Win Rate 30d	61.9%
Bull Run Sharpe	10.44 (100% WR, 7 signals)
vs Buy & Hold	6.6x outperformance

Architecture (9 Innovation Layers)

32 engineered features (price, momentum, volume, volatility, pattern)
6-regime detection (Trending Up Strong/Up, Down, Slow Bleed, Mean-Reverting, Volatile)
Regime-adaptive signal weighting per market condition
Rolling 60-day adaptive reweighting based on signal win rates
Confluence gate: 2+ signals in unfavorable regimes, 1 high-confidence in favorable
Dynamic TP (2.5×ATR, min 8%) / SL (1.5×ATR, min 5%) with trailing stop
Position sizing: confluence × regime × inverse-volatility
Momentum exhaustion scoring (distribution phase detection)
Drawdown circuit breaker (pauses after 3 consecutive losses)

Cross-Asset Results

Asset	Hybrid Sharpe	Return	Sortino	Max DD	Prev Best Sharpe	Buy & Hold
BTC	3.86	+331.92%	7.21	-22.40%	3.31 (Custom v1)	+50.21%
AVAX	2.50	+272.65%	4.16	-52.24%	3.64 (Custom v1)	-49.36%
ETH	2.00	+112.07%	3.53	-47.04%	4.29 (Generic v1)	+28.49%

Note: Hybrid v2.0 obliterates BTC Sharpe (3.86 vs 3.31, +16.6% improvement). AVAX and ETH trail the per-asset-optimized v1 models because extended altcoin downtrends reduce the effectiveness of general regime detection. However, AVAX Hybrid delivers +272.65% vs -49.36% buy-hold (5.5x outperformance).

Top Performing Signals (BTC)

Signal	Fires	7d Win Rate
OBV Divergence (Smart Money)	5	80%
Range Expansion (Tudor Jones)	3	66.7%
Support Bounce (3+ touches)	9	55.6%
Hammer at Support / 200MA Reclaim	4	50%
EMA50 Bounce in Uptrend	5	40%

Live Hybrid Predictions → | Raw BTC Backtest → | API Status →

Methodology

Data Source & Timeframe

Parameter	Value
Source	Kraken Public OHLC API (`api.kraken.com/0/public/OHLC`)
Interval	Daily candles (1440 minutes)
Candles Available	~721 per asset
Date Range	Feb 25, 2024 – Feb 14, 2026
Assets	BTCUSD, ETHUSD, AVAXUSD (Kraken pairs)
Walk-Forward Rule	At each bar t, only data [0..t] visible. Zero look-ahead bias.
Debounce	5-day minimum gap between same-model signal fires
Signal Threshold	Weighted score ≥ 25 (out of 100)

Market Regime Periods (Out-of-Sample Validation)

Period	Start	End	Regime
Period 1	2024-02-25	2024-09-30	Consolidation / Pre-Bull (BTC $50-67k)
Period 2	2024-10-01	2025-01-31	Bull Run (BTC $60k → $126k ATH)
Period 3	2025-02-01	2025-07-31	Correction + Recovery (BTC $126k → $78k → $115k)
Period 4	2025-08-01	2026-02-14	Late Cycle / Mixed (BTC $115k → $68k)

Outcome Measurement

Returns: Close-to-close at +7 days, +14 days, +30 days after signal
Max Gain/Drawdown: Measured from intraday highs/lows (not just closes)
TP/SL System: +10% take-profit vs -5% stop-loss. Whichever is hit first determines WIN/LOSS. If neither hit within 30 days: NEITHER.
Equity Curve: Starts at $100. Each signal's 30-day return compounds.
Statistical Tests: Welch's two-sample t-test on 30-day returns (customized vs generic). 95% confidence intervals on average returns.

Assumptions & Limitations

Execution at daily close price — no slippage model applied
No transaction fees included (Kraken fees: 0.16-0.26% per trade)
Full position on each signal (no fractional Kelly sizing)
~2 years of data: limited sample size for rare signals (Golden Cross fires only ~2x per asset)
Daily timeframe only: intraday patterns not captured
No on-chain data, no sentiment data, no order book data — Kraken OHLCV only
Survivorship bias: only testing assets that still trade on Kraken
Walk-forward on single continuous window (no expanding-window cross-validation)

Backtesting Results

Loading BTC backtest from Kraken API... (first load ~10s)

Current Live Predictions

Loading live signals...

DISCLAIMER: This research is for educational and analytical purposes only. It is NOT financial advice. Past performance does not guarantee future results. Cryptocurrency trading carries substantial risk of loss. All data sourced from Kraken public API. No guarantees of accuracy, completeness, or timeliness. Always do your own research and consult a qualified financial advisor before making investment decisions.

Source Code & API

Hybrid v2.0 Backtest (BTC): hybrid_predictor.php?action=backtest&asset=BTC
Hybrid v2.0 Live Predictions: hybrid_predictor.php?action=predict_all
v1 Backtest API (BTC): prediction_model.php?action=backtest&asset=BTC
v1 Compare All: prediction_model.php?action=compare_all
v1 Live Predictions: prediction_model.php?action=predict_all
20-Algo Legend Backtest: legend_backtest.php?action=backtest&asset=BTC
12-Signal Spike Detector: spike_detector.php?action=scan