AI summary by TwinMind
Odds and Log(Odds), Clearly Explained!!!
StatQuest with Josh Starmer
11m 30s
Introduction and Basic Odds Concept
- Josh Starmer introduces odds and log odds concepts
- Odds are ratio of something happening to something not happening
- Example 1: Team winning odds 1 to 4
- 5 games total: 1 win, 4 losses
- Written as fraction: 1/4 = 0.25
- Example 2: Team winning odds 5 to 3
- 8 games total: 5 wins, 3 losses
- Written as fraction: 5/3 = 1.7
Odds vs Probability Distinction
- Odds ≠ probabilities (critical distinction)
- Odds: ratio of something happening to something NOT happening
- Probability: ratio of something happening to EVERYTHING that could happen
- Example comparison:
- Odds in favor: 5 to 3 (5/3 ratio)
- Probability of winning: 5/8 = 0.625
- Probability of losing: 3/8 = 0.375
Calculating Odds from Probabilities
- Two methods yield same results:
- From raw counts: direct ratio
- From probabilities: P(win)/P(lose) or P/(1-P)
- 50% of time odds calculated from counts, 50% from probabilities
- Simplified formula: P/(1-P) where P = probability of winning
Log Odds Concept and Problem Solving
- Odds range issues:
- Against winning: 0 to 1 (approaching zero for worse teams)
- For winning: 1 to infinity (increasing for better teams)
- Asymmetry makes comparison difficult
- Example asymmetry problem:
- Odds against 1:6 = 1/6 = 0.17
- Odds for 6:1 = 6/1 = 6.0
- Magnitudes appear very different
Log Odds Solution
- Taking log creates symmetry
- Example with 1:6 vs 6:1 odds:
- Log(1/6) = log(0.17) = -1.79
- Log(6/1) = log(6) = +1.79
- Equal distance from zero (origin)
- Same calculation methods work: counts or probabilities
Statistical Applications
- Log odds ratio = logit function
- Forms basis for logistic regression
- Random pairs adding to 100 create normal distribution histogram
- Useful for binary outcome problems (win/lose, yes/no, true/false)
Key Distinctions and Summary
- Odds vs odds ratio are different concepts (future topic)
- Log odds provides:
- Symmetrical representation
- Easier interpretation
- Better for statistical analysis
- Core concept: log odds = log of the odds ratio
