DEV Community

Igor Ganapolsky
Igor Ganapolsky

Posted on • Originally published at github.com

AI Trading: LL-124: Phil Town CSP Strategy Not Executing Trades

#ai #trading #python #machinelearning

LL-124: Phil Town CSP Strategy Not Executing Trades

Incident Date

January 9, 2026

Severity

CRITICAL - Paper account sat idle for 4 days with $5,000, 0 trades

Summary

The Phil Town Cash-Secured Put (CSP) strategy was NOT executing trades due to two bugs:

  1. Wrong condition in rule_one_trader.py: Only traded when stock was BELOW MOS (defeats the purpose!)
  2. Misleading counters in orchestrator: puts_executed counter incremented on logging, not execution

Root Cause Analysis

Bug 1: Inverted Trade Logic

The code in scripts/rule_one_trader.py lines 310-320 had:

if "STRONG BUY" in recommendation and "Below MOS" in recommendation:
    # Only traded here
Enter fullscreen mode Exit fullscreen mode

Problem: Phil Town's strategy is to SELL PUTS when stock is ABOVE MOS to "get paid to wait" for your price. The code only traded when stock was already below MOS - which defeats the entire purpose!

Correct Logic:

  • Stock BELOW MOS → Buy shares directly (it's on sale!)
  • Stock ABOVE MOS but BELOW Sticker → SELL PUT at MOS (getting paid to wait) ← THIS IS THE STRATEGY
  • Stock ABOVE Sticker → Don't trade (overvalued)

Bug 2: Dishonest Counters

The orchestrator's run_options_strategy() had:

results["puts_executed"] += 1  # MISLEADING!
Enter fullscreen mode Exit fullscreen mode

This counter incremented when LOGGING signals, not when executing orders. It made the system appear to be trading when it wasn't.

Evidence

  • Paper account: $5,000 equity, 0 positions, 0 orders for 4+ days
  • Alpaca dashboard showed: "No orders. Place a trade..."
  • Workflow logs showed "puts_executed: 3" but no actual orders in Alpaca

Fixes Applied

Fix 1: Corrected Trade Logic (rule_one_trader.py) 

# NOW TRADES WHEN:
elif "BUY" in recommendation:
    # Stock is above MOS but below Sticker = "getting paid to wait"
    logger.info(f"  🎯 ACTIONABLE: {symbol} above MOS, below Sticker - Selling CSP!")
    trade = execute_phil_town_csp(client, symbol, analysis)
Enter fullscreen mode Exit fullscreen mode

Fix 2: Honest Counters (orchestrator/main.py) 

results["puts_logged"] = 0      # HONEST: we log, not execute here
results["calls_logged"] = 0     # HONEST: we log, not execute here
results["note"] = "Signals logged here; execution in rule_one_trader.py"
Enter fullscreen mode Exit fullscreen mode

Verification

After fix:

  • Stocks trading between MOS and Sticker will trigger CSP orders
  • Counters accurately reflect what the code does
  • Actual execution happens in rule_one_trader.py which submits real orders

Prevention

  1. Counter naming: Use _logged or _analyzed for non-execution counters
  2. End-to-end testing: Verify orders appear in Alpaca, not just logs
  3. Daily trade activity monitor: Alert when $5K account has 0 trades for 3+ days

Related Files

  • scripts/rule_one_trader.py - Fixed trade condition
  • src/orchestrator/main.py - Fixed misleading counters
  • .github/workflows/daily-trading.yml - Calls rule_one_trader.py

Phil Town Strategy Reminder

From "Payback Time" by Phil Town:

"We sell puts on wonderful companies at their Margin of Safety price. If the stock drops to our price, we get to buy it at a huge discount. If it doesn't, we keep the premium. Either way, we win."

The key insight: You sell puts ABOVE your target price to GET PAID WHILE WAITING.

Tags

  • phil-town
  • options
  • csp
  • execution-failure
  • paper-trading
  • critical-bug

This lesson was auto-published from our AI Trading repository.

More lessons: rag_knowledge/lessons_learned

Top comments (0)