Quant Companion MCP

# Momentum Plus Strategy Backtest Report Symbol: SPY (S&P 500 ETF) Period: 2015-01-01 to 2025-12-01 (~11 years) Run Date: 2025-12-06 --- ## Executive Summary Initial Capital: $10,000 Final Capital: $23,094.57 Total Return: +130.95% CAGR: 8.82% Sharpe Ratio: 0.715 Sortino Ratio: 0.950 Max Drawdown: 19.56% Win Rate: 100% (5/5 trades) Avg Holding Period: 595 days --- ## Strategy Logic The momentum_plus strategy is an enhanced dual-timeframe momentum system: ### Entry Conditions (ALL must be true) 1. 12-month momentum > 0% — Long-term trend is up 2. 3-month momentum > 0% — Short-term confirmation 3. 20-day annualized volatility < 30% — Avoids choppy/crash markets ### Exit Conditions (ANY triggers exit) 1. 12-month momentum < -5% — Long-term trend broken 2. 3-month momentum < -10% AND 12-month momentum < 3% — Early warning crash signal ### Key Design Choices - Asymmetric exits: Harder to exit than enter (reduces whipsaw) - Vol filter on entry only: Lets winners run even in rising vol - Dual timeframe confirmation: Requires both long and short term agreement I tried symmetric entry/exit thresholds initially. Disaster — got chopped up in every sideways market. The asymmetry feels wrong intuitively but the numbers dont lie. --- ## Data Pipeline ### How Data Was Pulled 1. MCP Tool: run_backtest called with params: symbol: "SPY", start: "2015-01-01", end: "2025-12-01", strategy: { type: "momentum_plus" } 2. Data Provider: YahooFinanceProvider (default) Endpoint: getHistoricalOHLCV() Interval: Daily (1d) Fields: timestamp, open, high, low, close, volume 3. Data Flow: Yahoo Finance API -> OHLCV array -> Candle[] conversion -> Backtest engine ### Provider Fallback Chain If POLYGON_API_KEY environment variable is set: Polygon.io (primary) -> Yahoo Finance (fallback) Otherwise: Yahoo Finance only --- ## Trade-by-Trade Analysis ### Trade 1: Post-2015 Recovery Bull Run Entry: 2016-04-01 @ $206.92 Exit: 2018-12-14 @ $260.47 Shares: 48 P&L: +$2,557.90 (+25.60%) Holding: 987 days (2.7 years) Context: Entered after the early 2016 correction stabilized. Strategy correctly identified the sustained bull market and held through minor corrections. Exited during Q4 2018 selloff when 3-month momentum crashed below -10%. --- ### Trade 2: 2019 Recovery into COVID Entry: 2019-02-06 @ $272.74 Exit: 2020-03-09 @ $274.23 Shares: 45 P&L: +$54.71 (+0.44%) Holding: 397 days Context: Re-entered after December 2018 crash recovery. Held through 2019's strong year but got stopped out at the start of COVID crash (March 9, 2020 - just days before the bottom). Small profit but avoided the -34% COVID drawdown. --- ### Trade 3: COVID Recovery Bull Market Entry: 2020-05-28 @ $302.97 Exit: 2022-05-09 @ $398.17 Shares: 41 P&L: +$3,886.88 (+30.90%) Holding: 711 days Context: Re-entered after vol normalized post-COVID crash (Fed intervention). Captured the massive 2020-2021 bull run. Exited in early May 2022 as the bear market began, avoiding the bulk of 2022's -20%+ drawdown. --- ### Trade 4: 2023 AI Rally Entry: 2023-05-10 @ $412.85 Exit: 2025-04-04 @ $505.28 Shares: 39 P&L: +$3,585.06 (+21.80%) Holding: 695 days Context: Entered as the 2023 AI-driven rally gained momentum. Held through 2024. Exited in early April 2025 when momentum conditions reversed. --- ### Trade 5: 2025 Re-entry (Active) Entry: 2025-05-29 @ $590.05 Exit: 2025-11-28 @ $683.39 (EOY) Shares: 33 P&L: +$3,080.22 (+15.39%) Holding: 183 days Context: Re-entered late May 2025. Still holding at end of backtest period. --- ## Risk Analysis ### Drawdown Periods Avoided 1. Q4 2018 (-20%): Exited Dec 14, avoided worst 2. COVID 2020 (-34%): Exited Mar 9, avoided -28% of the drop 3. 2022 Bear (-25%): Exited May 9, missed most of the drawdown the COVID exit timing was honestly lucky - strategy happened to trigger right before the bottom. Wouldnt count on that level of precision in the future ### Time Out of Market - Total days: ~2,744 trading days - Days in market: ~2,973 calendar days across 5 trades - Days flat: Significant periods in 2018-2019, 2020, 2022-2023 ### Max Drawdown Experienced 19.56% — This is the strategy's max DD, NOT SPY's. The strategy avoided the worst of SPY's drawdowns. --- ## Strategy Parameters ```typescript createMomentumPlusSignal( longLookback: 252, // 12-month momentum lookback shortLookback: 63, // 3-month momentum lookback entryThreshold: 0, // min momentum to enter (0%) exitThreshold: -0.05, // exit if 12mo < -5% shortExitThreshold: -0.10, // exit if 3mo < -10% maxVolThreshold: 0.30 // don't enter if vol > 30% ) ``` --- ## Code Reference Strategy implementation: packages/quant-core/src/backtest.ts lines 292-349 ```typescript // Entry logic if (currentPosition === "flat") { if (longMomentum > entryThreshold && shortMomentum > entryThreshold && annualizedVol < maxVolThreshold) { return "long"; } } // Exit logic if (currentPosition === "long") { if (longMomentum < exitThreshold) return "flat"; if (shortMomentum < shortExitThreshold && longMomentum < 0.03) return "flat"; } ``` --- ## Files Generated 1. momentum_plus_SPY_2015-2025.json — Full raw results 2. momentum_plus_SPY_2015-2025_trades.csv — Trade log for spreadsheets 3. momentum_plus_SPY_2015-2025_REPORT.md — This report --- ## Comparison to Buy & Hold Momentum Plus: Total Return +130.95%, Max Drawdown 19.56%, Sharpe 0.715, 5 Trades SPY Buy & Hold: Total Return ~180%*, Max Drawdown ~34% (COVID), Sharpe ~0.55*, 1 Trade *Estimates for comparison **Verdict:** Lower total return but significantly better risk-adjusted returns and much smaller drawdowns. This is a defensive momentum strategy. --- ## Notes - should probably run this on more symbols to check if it generalizes - the 100% win rate is misleading, its only 5 trades. not statistically significant - need to add transaction cost modelling