Fix backtesting crash for non-SMA strategies, division by zero, and alert logic

[devin-ai-integration]

Summary

Fixes three bugs in app.py:

1. Backtesting crash (critical): Selecting "RSI Oversold/Overbought", "Bollinger Band Breakout", or "Custom" in the backtesting module caused a KeyError on df['Position'] because only the SMA Crossover strategy was implemented. Now all four strategy options work:

   • RSI: Buys when RSI drops below oversold, sells when RSI rises above overbought. Uses simple rolling-window RSI.
   • Bollinger Band: Buys on upper band breakout, sells on lower band breakout.
   • Custom: Falls back to buy-and-hold with a warning.

2. ZeroDivisionError in MAE/MFE analysis: avg_mfe / avg_mae crashed when avg_mae was 0. Added a guard.

3. Alert threshold only fired on gains: daily_pnl > threshold never alerted on large losses. Changed to abs(daily_pnl) > threshold.

Review & Testing Checklist for Human

• Verify RSI calculation correctness: The RSI implementation uses a simple rolling mean (SMA) for average gain/loss rather than Wilder's exponential smoothing. This is a valid approach but will produce slightly different values than most charting platforms. Decide if this is acceptable for your use case.
• Test all 4 backtesting strategies end-to-end in the Streamlit UI: Select each strategy type, run a backtest with real ticker data, and confirm the equity curve and metrics render without errors.
• Verify the alert behavior change is desired: The alert threshold now fires on large losses too (via abs()). Previously it only alerted when P&L was positive and exceeded the threshold. Confirm this is the intended behavior.
• Check signal persistence logic for RSI/BB: Both strategies use ffill() to hold a position until the opposite signal fires. Verify this matches your intended trading logic (vs. exiting to flat when the signal is neutral).

Notes

• No automated tests exist in this repo, so all verification must be manual via the Streamlit UI.
• The avg_loss.replace(0, np.nan) in RSI guards against division by zero when computing relative strength.

Link to Devin session: https://app.devin.ai/sessions/a97d0c2bf8584905b198c44220a50eac
Requested by: @Philip-2007

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring