Daily Alpha Brief — 2026-09-02 — BEAR regime across majors

Bayesian crypto brief for 2026-09-02: BEAR regime, strategy performance, weight shifts, and quant-vs-crowd sentiment across 9 symbols.

TL;DR

• Dominant regime: BEAR (4 bullish / 5 bearish of 9 symbols).
• Top signal: ema_cross on ETH cut 0.255 (0.51 → 0.26).
• Shadow portfolio: $730 (-27.03%), 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 59.6% signals balanced.

🌐 Market Overview

Economic (FRED):

Stablecoin dominance 9.8% (risk-on); yield curve regime Flat.

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

Economic Calendar: Next FOMC: in 14 days | Next CPI: in 9 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 bearish: 4 BULL / 5 BEAR of 9 symbols.

🔗 Correlation Signals

Stable pairs (r > 0.70): XRP/DOGE: r=0.90 ✅ | SUI/AVAX: r=0.89 ✅ | XRP/SUI: r=0.89 ✅ | DOGE/SUI: r=0.88 ✅ | DOGE/AVAX: r=0.88 ✅ | BTC/ETH: r=0.87 ✅

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

ETH/USDT (BULL):

ema_cross and walk_forward_ema are the worst performers: both show 7d win-rates well below 30% with deeply negative Sharpes (-3.47 and -4.24) and negative pnl, so they are cut hard per rules 2-3 — this is a conflict-rule (rule 6) situation where the BULL regime would normally favour momentum but the data overrides it. rsi_macd also carries a negative 7d Sharpe (-1.75) and sub-30% effective win rate so it is reduced modestly. Bollinger is the standout paradox in a BULL regime: a 5.17 Sharpe and +199 7d pnl warrant holding it near current weight despite the band-walk risk, and it is only trimmed slightly from 0.573 to 0.58 (rule 4 applied cautiously given the regime conflict). volume_spike holds near neutral on positive 7d metrics; tsmom and post_shock_reversal have zero trades (rule 1 thin-data guard, held within ±0.05 of current); sentiment_strategy gets a mild haircut from 0.555 toward 0.52 because 24h coverage is zero and the 7d composite is near-flat (+0.02), indicating weak signal; correlation_strategy and macro_strategy remain near their current weights given no actionable data.

WIF/USDT (BEAR):

rsi_macd and tsmom have zero trades on both windows so rule 1 applies and weights are held at current. ema_cross has only 3 7d trades (rule 1 low-sample guard) but its metrics are exceptionally strong (67% win, Sharpe 7.04, +198 pnl) and BEAR regime favours momentum trend-followers, so a small nudge up to 0.490 is warranted. volume_spike has ample 7d data showing 30% win-rate (borderline rule 2) and a severe 24h collapse (-50 pnl, -3.52 Sharpe), signalling recent deterioration, so it is cut meaningfully to 0.420 despite the strong 7d pnl figure — the 24h signal of -3.52 Sharpe is a clear warning. walk_forward_ema shows a problematic 33% 7d win-rate with negative-Sharpe-adjacent risk-adjusted profile once the 24h data is considered, so it is trimmed slightly to 0.400; post_shock_reversal stays at current given thin 7d trades and mixed signals. sentiment_strategy is reduced to 0.400 from 0.433 because 7d confluence is 0% and coverage is sparse (sources=2, avg_composite near zero), indicating contradictory or missing data per the sentiment weighting rules; correlation_strategy and macro_strategy remain at current with no actionable data.

BNB/USDT (BULL):

walk_forward_ema is the standout performer with 7d Sharpe=5.31, 43% win-rate and +89.16 pnl, earning a meaningful raise; rsi_macd retains a moderate weight on its positive 7d Sharpe (1.45) and pnl despite a brutal 24h, though its elevated current weight is trimmed given the 7d win-rate of only 38%. bollinger, volume_spike, tsmom, and post_shock_reversal all show 7d win-rates below 30% with large negative Sharpes and pnl — rules 2-3 trigger cuts toward 0.10-0.20, and ema_cross is similarly cut for negative 7d Sharpe and sub-35% win-rate despite BULL regime alignment (conflict rule 6 overrides the momentum prior). sentiment_strategy is nudged down slightly from 0.509 toward 0.46 because 7d confluence is 0% and composite is essentially flat, signalling contradictory/sparse sources; correlation_strategy and macro_strategy remain near current weight under the thin-data guard as they show zero trades/signals in both windows.

📡 Signal Outlook

💼 Shadow Portfolio

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

📡 Discourse Radar

Sentiment analysis pipeline launching soon.

Academic Footnote

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

Disclaimer

This is algorithmic analysis only, not financial advice.