Walsh|AAPLBUY✓ ApprovedWeighted score=1

Walsh Demo

Multi-agent stock research pipeline · Pre-generated runs (no live API calls)

The Jan–Dec 2023 backtest returned +75.5% vs +119.8% passive buy-and-hold on AAPL/MSFT/NVDA, underperforming by ~44 pp. The primary cause is that the volatility damper (vol_cap=40%) zeroed out NVDA allocation — the period's biggest winner — while the mean-reversion agent added conflicting SELL signals in a one-directional trend year. The immediate next step is replacing the hard vol_cap=40% binary cutoff in RiskManager.adjust() with continuous position-size scaling.

Analyst Signals

📊Fundamental

BUY

65% confidence

PE=22, revenue_growth=8%, profit_margin=26%, D/E=0.6 → score=1

📈Technical

STRONG BUY

60% confidence

RSI=42, MACD_hist=0.150, golden_cross=False → score=2

🗞️Sentiment

BUY

54% confidence

news=0.40, social=0.30, insider_buying=False → score=1

🌍Macro

BUY

50% confidence

GDP=2.7%, fed_rate=5.3%, inflation=3.1%, sector=neutral → score=1

Decision

🧠Portfolio Manager Synthesis

Weighted score=1.26 → BUY. Agents: [fundamental=BUY(65%), technical=STRONG_BUY(60%), sentiment=BUY(54%), macro=BUY(50%)]

🛡️Risk Manager Decision

✓ APPROVED·Final: BUY

No risk flags

📉 Backtest ResultsReal Data

Fixed universe: AAPL · MSFT · NVDA — numbers do not change with ticker selection above.

Jan – Dec 2023 · AAPL · MSFT · NVDA · Rule-based approximation — deterministic SMA/momentum agents, not the LLM-powered agents above.

Sharpe Ratio

2.36

vs 2.86 B&H −0.5

Total Return

+75.52%

vs +119.75% B&H −44.2 pp

Win Rate

76.2%

21 trades · Max DD −13.50%

Walsh (+75.5%)B&H (+119.8%)Gap (−44 pp)

Portfolio value (% of initial $100k) · monthly snapshots

Per-agent calibration metrics pending ≥50 completed LLM-backed pipeline runs.