Daily Alpha Brief — 2026-09-21 — BULL regime across majors

Bayesian crypto brief for 2026-09-21: BULL regime, strategy performance, weight shifts, and quant-vs-crowd sentiment across 9 symbols.

TL;DR

• Dominant regime: BULL (6 bullish / 2 bearish of 9 symbols).
• Top signal: tsmom on AVAX raised 0.338 (0.30 → 0.64).
• Shadow portfolio: $728 (-27.17%), 0 open, 35% win rate.

Shadow portfolio value since inception (dashed line = $1,000 starting capital).

Current Bayesian reliability weights, averaged across the universe.

Detected regime per symbol (EMA200 + ADX); % = lead-strategy conviction.

🌐 Market Indices

Crowd reads greed at 72/100; BTC dominance 58.6% signals balanced.

🌐 Market Overview

Economic (FRED):

Stablecoin dominance 9.3% (risk-on); yield curve regime Flat.

Funding Rates (contrarian — heavy positive = crowded longs to fade):

Economic Calendar: Next FOMC: in 37 days | Next CPI: in 22 days | position-size modifier 1.00×.

📊 Regime Analysis

Regime per symbol from EMA200 + ADX on daily candles, with the fusion signal.

The book is broadly bullish: 6 BULL / 2 BEAR of 9 symbols.

🔗 Correlation Signals

Stable pairs (r > 0.70): ETH/SOL: r=0.86 ✅ | BTC/ETH: r=0.85 ✅ | SOL/DOGE: r=0.82 ✅ | ETH/DOGE: r=0.81 ✅ | BTC/SOL: r=0.81 ✅ | DOGE/SUI: r=0.80 ✅

Divergence alerts (|z| > 1.5):

Index signals: BTC dominance NEUTRAL → no rotation; Fear & Greed contrarian NONE.

⚙️ Strategy Performance

Aggregate 7-day performance per strategy (replayed across the universe). Edge = win-rate × avg P&L per trade — the table is sorted by it.

All P&L figures normalized to a 100 USDT position size for comparability (not summed across symbols).

🧠 Bayesian Weight Shifts

AVAX/USDT (BULL):

rsi_macd and ema_cross both have only 1 and 11 trades respectively on 7d — rsi_macd triggers rule 1 (thin data, hold near current), while ema_cross shows a 36% win-rate and negative 7d Sharpe (-1.35) in a BULL regime, a direct conflict per rule 6, so it is cut modestly. bollinger is the standout data winner (69% win, Sharpe 16.69, +1007 pnl on 7d) and earns a raise per rule 4 despite being a mean-reversion strategy in a BULL regime — the data overrides the regime prior per rule 6. volume_spike has a strong pnl but only 37% win-rate on 7d, so it is trimmed slightly below its current weight; tsmom’s 7d win-rate (47%) fails the rule-4 threshold and it is nudged up only marginally given its positive Sharpe; post_shock_reversal’s 7d win-rate of 33% in a BULL regime triggers a modest cut. walk_forward_ema mirrors ema_cross metrics exactly (36% win, -1.35 Sharpe, negative pnl) and is already below neutral, so it is held near its floor. sentiment_strategy has full 3-source coverage and 100% confluence on both windows, justifying a maintained high weight with a small trim from the current elevated level. correlation_strategy and macro_strategy both show zero activity on all windows, so they are held near their current weights with no movement.

ETH/USDT (BULL):

ema_cross and walk_forward_ema are the clearest cuts: both show 7d win-rates well below 30% with deeply negative Sharpes and large negative pnl (-192 and -140), triggering rules 2-3 hard despite the BULL regime — a clear conflict-rule 6 situation where data overrides the regime prior. tsmom has only 3 trades on 7d (rule 1 low-sample guard) so it is held within ~0.05 of current, pulled down modestly from its elevated 0.657 given the negative metrics rather than left unchanged. post_shock_reversal is the standout performer (7d: 46% win, Sharpe 14.45, +938 pnl) and earns a maintained high weight; bollinger and volume_spike have mixed 7d metrics (positive Sharpe but sub-50% win-rates) so they are held near neutral with only modest adjustments; sentiment_strategy retains a high weight given 100% confluence across 3 sources and a consistently positive composite, trimmed slightly from 0.733 given the near-neutral avg_composite signal; correlation_strategy and macro_strategy both report zero activity so they are held near their current weights per the non-price voter guidance.

XRP/USDT (BULL):

tsmom has only 3 trades on 7d (rule 1 low-sample guard) but both windows show deeply negative Sharpe and pnl (-99.97), so I reduced it meaningfully from its inflated 0.677 toward a more cautious 0.45 — the thin data prevents a full cut but the consistent losses justify a clear downward move. post_shock_reversal is the standout performer with 61 7d trades, 41% win-rate, Sharpe 10.57, and +787 pnl, earning a raise to 0.60 despite being a mean-reversion strategy in a BULL regime (data wins over regime per rule 6). volume_spike and walk_forward_ema are both cut: volume_spike has 7d win-rate of 33% with barely positive pnl (rules 2-3), and walk_forward_ema has a negative 7d Sharpe and negative pnl on 23 trades (rule 3); sentiment_strategy holds at 0.75 given strong 3-source coverage and 100% confluence across both windows, while the inactive correlation and macro strategies remain near their current neutral-low weights.

📡 Signal Outlook

💼 Shadow Portfolio

$1,000 USDT paper portfolio, 1% risk/trade, MEXC fees modelled. No real capital is deployed.

📡 Discourse Radar

Crowd-vs-quant sentiment from CryptoCompare news, crypto RSS, the macro feeds (FT / Fed / CNBC), and the Fear & Greed index. (Reddit retired — Responsible Builder Policy restrictions, 2026.)

• Fear & Greed: +0.50 (bullish crowd).
• Macro signal (FT/Fed/CNBC): +0.03 (neutral) — “Federal Reserve Board announces approval of the application by Banco Santander, S.A. and Santander H”.

Most-talked alt: BNB/USDT at +0.50 (bullish crowd lean).

📰 Today’s Key Reads

Curated from 3 sources monitored today:

🟢 Bitcoin ETFs eke out positive week with $433 million Friday inflow as ether funds snap four-week inflow streak — theblock — Score: +0.56

🔴 Ethereum confirms Glamsterdam dates, but warns ‘fake’ builders could stall the chain — coindesk — Score: -0.39

🟢 JPMorgan says bitcoin could get more support than gold if ETF hedging eases — theblock — Score: +0.36

⚪ Bank of Japan raises interest rates by 25 basis points. Bitcoin tops $77,000 — coindesk — Score: +0.19

⚠️ TotalEnergies faces fresh scrutiny over response to Mozambique attack — ft_markets — Score: -0.17

📈 System Architecture

The four-panel Bayesian fusion engine behind these daily decisions (shown Mondays). See Methodology.

Academic Footnote

Care & Cumming (2024) — algorithmic trading and crypto market quality.

Disclaimer

This is algorithmic analysis only, not financial advice.