COT Signal Backtesting

The Idea

Every week, the CFTC reports how different groups of traders are positioned in futures markets. Sometimes a group reaches an extreme - for example, commercial hedgers might be the most net-long they've been in 6 months. Historically, these extremes have often preceded price moves. But how often? That's what this page answers.

Step 1: Find Extreme Positioning (with Example)

For each weekly COT report, we look at the trader's net position (longs minus shorts) and compare it to recent history. Here's a simple example with Gold commercials over 10 weeks:

Week 1:  +10,000
Week 2:  +15,000
Week 3:  +12,000
Week 4:   +8,000  ← Lowest in window
Week 5:  +20,000
Week 6:  +14,000
Week 7:  +11,000
Week 8:   +9,000
Week 9:  +13,000
Week 10: +25,000  ← This week's COT report just came out

If this week were +8,500 instead, the percentile would be (8,500 - 8,000) / 17,000 = 2.9th percentile - that's below 10 (which is 100 minus the 90 threshold), so it would be a bearish signal. Commercials have almost no long exposure.

The Threshold filter controls the cutoff. At 90, only the top 10% and bottom 10% trigger signals. At 80, the top 20% and bottom 20% count (more signals, less extreme).

Step 2: Enter on the Signal, Then Watch What Happens

Important: We don't look at what price already did. We look at what price does NEXT. The signal fires the moment the COT report shows extreme positioning. That's the entry. Then we wait and measure:

Week 10: COT report → 100th percentile → BULLISH SIGNAL!
         Gold price right now = $2,000 ← ENTRY PRICE (we "buy" here)

Week 11: Gold = $2,020  → 1W return: +1.0%
Week 12: Gold = $2,040  → 2W return: +2.0%
Week 14: Gold = $2,080  → 4W return: +4.0% ← WIN! Price went up.
Week 18: Gold = $2,050  → 8W return: +2.5%

The percentile measures positioning (how many contracts traders are holding), not price. Price could be doing anything when the signal fires. The bet is: "these traders know something, and price will eventually follow their positioning."

For bearish signals, it works in reverse: we check if price went DOWN after the signal. If it did, that's a win.

Step 3: Calculate Accuracy

We go through every signal in history, check if price moved the right way, and count up wins vs losses. If 62% of bullish signals were followed by price going up within 4 weeks, that's a 62% win rate. Above 50% means the signals have predictive value — they beat a coin flip.

What Do the Stat Cards Mean?

• Extreme Signals Found: How many times positioning hit an extreme level across all markets in the selected time period.
• 4-Week Accuracy: Of all those signals, what percentage correctly predicted direction 4 weeks later. This is the main number to look at.
• Avg Move After 4 Weeks: The average price move (in the signal's direction) after 4 weeks. Positive = profitable on average.
• Best Signal: The single signal with the largest correct price move at 4 weeks.

Tips for Using This

• Institutions (commercial hedgers) are often the best traders in commodity markets — they're hedging actual business exposure and tend to be right at extremes.
• Fund Managers may have better signals in financial futures (indices, currencies).
• A higher threshold (95) gives fewer signals but they tend to be more accurate. Lower (80) gives more signals for larger sample sizes.
• Look at the 4W and 8W columns — those tend to show the clearest edge. The 1W column is often noisy.
• Click any market name to see a price chart with the signal points overlaid — this makes the pattern much easier to visualize.