Key Takeaways: Chapter 18

Core Principles

1. Behavioral biases are systematic, not random. Unlike random noise, cognitive biases produce predictable patterns of mispricing that persist over time and across markets. This predictability is what makes biases exploitable.

2. The favorite-longshot bias is the most well-documented bias in betting and prediction markets. Longshots (low-probability events) are systematically overpriced, and favorites (high-probability events) are systematically underpriced. The bias is driven by probability weighting (overweighting small probabilities) and is well-described by prospect theory's weighting function with gamma approximately 0.65.

3. Anchoring causes slow price adjustment. Traders anchor to initial prices, round numbers, and external forecasts, adjusting only 40-70% of the way to the information-justified price. This creates predictable under-reaction to new information.

4. Overconfidence manifests as overprecision. The most costly form of overconfidence for traders is overprecision — confidence intervals that are too narrow. This leads to excessive position sizing and insufficient hedging.

5. Herd behavior creates momentum and reversal patterns. Short-term positive autocorrelation (momentum) followed by medium-term negative autocorrelation (reversal) is the empirical signature of herding. Contrarian strategies can profit from herding episodes, but distinguishing herding from genuine information flow is the key challenge.

6. The availability heuristic overweights vivid, recent events. Contracts related to dramatic, emotionally salient, and heavily covered events tend to be overpriced. This is amplified by media feedback loops.

7. Prospect theory's loss aversion produces the disposition effect. Traders sell winners too early (risk-averse in the gain domain) and hold losers too long (risk-seeking in the loss domain). The loss aversion parameter lambda is approximately 2.25.

8. Confirmation bias is insidious and amplified by echo chambers. Traders seek and remember confirming evidence while ignoring contradictions. Online prediction market communities often become echo chambers that amplify this bias, creating collective overconfidence in popular positions.

9. Recency bias and the narrative fallacy create predictable drift-and-reversal patterns. Prices drift gradually as narratives build and correct sharply when narratives break. Contrarian traders can profit at narrative reversal points.

10. Multiple biases compound. When several biases point in the same direction, the combined mispricing can be substantial (10-15 cents), even though individual biases typically produce only 2-5 cents of edge.

Practical Trading Rules

• Sell longshots, buy favorites as a baseline strategy (FLB exploitation).
• Trade toward new information when the market has not fully adjusted (anti-anchoring).
• Fade extreme prices when you detect crowd overconfidence (anti-overconfidence).
• Trade against momentum when it occurs without new information (anti-herding).
• Sell vivid-event contracts that are priced above base rates (anti-availability).
• Use structured decision processes (checklists, pre-mortems, decision journals) to mitigate your own biases.
• Combine multiple bias signals for higher-confidence trades.

Quantitative Benchmarks

Debiasing Essentials

1. Knowledge of biases is necessary but not sufficient. You must implement structured processes.
2. Estimate before you look at the market price to avoid anchoring.
3. Consider the opposite before every trade.
4. Pre-commit to exit rules and do not modify them (to avoid disposition effect).
5. Keep a decision journal and review it monthly.
6. Train your calibration with regular practice.
7. Red team your analysis by arguing the other side.
8. Track your biases quantitatively with the tools from this chapter.

Common Mistakes to Avoid

• Assuming you are immune to biases because you know about them (bias blind spot).
• Treating all price movements as herding when some reflect genuine information.
• Overestimating the magnitude of biases (they exist but are typically small in prediction markets).
• Ignoring transaction costs, which can eat the entire bias-exploitation edge.
• Overfitting bias models to historical data and failing to validate out-of-sample.
• Using too many bias signals without a disciplined framework for combining them.
• Neglecting position sizing — even a correct bias detection is worthless if you bet too much or too little.