Behavioral Model Interactive

Anchoring Effects

People rely too heavily on the first piece of information they see. Opening lines, season averages, and round numbers all create anchoring biases.

⚓ The Anchoring Bias

How It Works

1. Person sees initial value (anchor)
2. They adjust from the anchor toward their estimate
3. Adjustment is typically insufficient
4. Final estimate is biased toward anchor

Famous Study

Tversky & Kahneman (1974): Spun a wheel showing 10 or 65, then asked "What % of African countries are in the UN?"

• Anchor of 10 → average guess: 25%
• Anchor of 65 → average guess: 45%
• Random anchor still affected judgment!

Player Props Example

15 35

0.1 0.8

Lower adjustment = stronger anchoring bias

📊 Bias Analysis

True Value 25

Anchor 30

Public Estimate 28.0

Mispricing +3.0 pts

Anchoring Visualization

True: 25

Anchor: 30

Public: 28.0

Public estimate is pulled toward the anchor, creating exploitable bias.

Opening Line Anchoring

18 28

Opening

23.5

Public Expects

23.1

True Value

Line opened too high. Public still anchors high → take UNDER.

🎯 Common Anchoring Points

Opening Line

Public doesn't adjust enough from opener

Edge: Fade line movement in wrong direction

Season Average

Ignores recent context/matchup

Edge: Value in contextual adjustments

Last Game

Overweight recent performance

Edge: Fade outliers (regression)

Round Numbers

Cluster bets at 25, 30, etc.

Edge: Value at 24.5, 30.5

💰 Pricing Implications

For Line Setting

→ Opening line sets the public's anchor
→ Even if you move the line, public adjusts slowly
→ Strategic opener can capture value

For Finding Edge

→ Look for lines that opened wrong
→ Fade insufficient adjustments
→ Value often at non-round numbers

R Code Equivalent

# Anchoring model
estimate_with_anchor <- function(anchor, true_value, adjustment = 0.4) { 
  # People start at anchor and adjust toward true value
  # But adjustment is typically insufficient
  public_estimate <- anchor - (anchor - true_value) * adjustment
  mispricing <- public_estimate - true_value
  
  list(
    anchor = anchor,
    true_value = true_value,
    public_estimate = public_estimate,
    mispricing = mispricing
  )
}

# Example
result <- estimate_with_anchor(30, 25, 0.4)
cat(sprintf("Anchor: %.1f, True: %.1f\n", result$anchor, result$true_value))
cat(sprintf("Public estimate: %.1f\n", result$public_estimate))
cat(sprintf("Mispricing: %+.1f pts\n", result$mispricing))

✅ Key Takeaways

• People anchor on first information seen
• Adjustment from anchor is insufficient
• Opening lines, season averages are common anchors

• Creates exploitable mispricing
• Fade lines that opened wrong
• Value at non-round numbers (24.5 vs 25)