Contents

Preface

Purpose

Audience

Prerequisites

Teaching from This Book

Acknowledgments

Chapter 1. Introduction to Ranking

Social Choice and Arrow’s Impossibility Theorem

Arrow’s Impossibility Theorem

Small Running Example

Chapter 2. Massey’s Method

Initial Massey Rating Method

Massey’s Main Idea

The Running Example Using the Massey Rating Method

Advanced Features of the Massey Rating Method

The Running Example: Advanced Massey Rating Method

Summary of the Massey Rating Method

Chapter 3. Colley’s Method

The Running Example

Summary of the Colley Rating Method

Connection between Massey and Colley Methods

Chapter 4. Keener’s Method

Strength and Rating Stipulations

Selecting Strength Attributes

Laplace’s Rule of Succession

To Skew or Not to Skew?

Normalization

Chicken or Egg?

Ratings

Strength

The Keystone Equation

Constraints

Perron–Frobenius

Important Properties

Computing the Ratings Vector

Forcing Irreducibility and Primitivity

Summary

The 2009–2010 NFL Season

Jim Keener vs. Bill James

Back to the Future

Can Keener Make You Rich?

Conclusion

Chapter 5. Elo’s System

Elegant Wisdom

The K-Factor

The Logistic Parameter ξ

Constant Sums

Elo in the NFL

Hindsight Accuracy

Foresight Accuracy

Incorporating Game Scores

Hindsight and Foresight with ξ = 1000, K = 32, H = 15

Using Variable K-Factors with NFL Scores

Hindsight and Foresight Using Scores and Variable K-Factors

Game-by-Game Analysis

Conclusion

Chapter 6. The Markov Method

The Markov Method

Voting with Losses

Losers Vote with Point Differentials

Winners and Losers Vote with Points

Beyond Game Scores

Handling Undefeated Teams

Summary of the Markov Rating Method

Connection between the Markov and Massey Methods

Chapter 7. The Offense–Defense Rating Method

OD Objective

OD Premise

But Which Comes First?

Alternating Refinement Process

The Divorce

Combining the OD Ratings

Our Recurring Example

Scoring vs. Yardage

The 2009–2010 NFL OD Ratings

Mathematical Analysis of the OD Method

Diagonals

Sinkhorn–Knopp

OD Matrices

The OD Ratings and Sinkhorn–Knopp

Cheating a Bit

Chapter 8. Ranking by Reordering Methods

Rank Differentials

The Running Example

Solving the Optimization Problem

The Relaxed Problem

An Evolutionary Approach

Advanced Rank-Differential Models

Summary of the Rank-Differential Method

Properties of the Rank-Differential Method

Rating Differentials

The Running Example

Solving the Reordering Problem

Summary of the Rating-Differential Method

Chapter 9. Point Spreads

What It Is (and Isn’t)

The Vig (or Juice)

Why Not Just Offer Odds?

How Spread Betting Works

Beating the Spread

Over/Under Betting

Why Is It Difficult for Ratings to Predict Spreads?

Using Spreads to Build Ratings (to Predict Spreads?)

NFL 2009–2010 Spread Ratings

Some Shootouts

Other Pair-wise Comparisons

Conclusion

Chapter 10. User Preference Ratings

Direct Comparisons

Direct Comparisons, Preference Graphs, and Markov Chains

Centroids vs. Markov Chains

Conclusion

Chapter 11. Handling Ties

Input Ties vs. Output Ties

Incorporating Ties

The Colley Method

The Massey Method

The Markov Method

The OD, Keener, and Elo Methods

Theoretical Results from Perturbation Analysis

Results from Real Datasets

Ranking Movies

Ranking NHL Hockey Teams

Induced Ties

Summary

Chapter 12. Incorporating Weights

Four Basic Weighting Schemes

Weighted Massey

Weighted Colley

Weighted Keener

Weighted Elo

Weighted Markov

Weighted OD

Weighted Differential Methods

Chapter 13. “What If . . .” Scenarios and Sensitivity

The Impact of a Rank-One Update

Sensitivity

Chapter 14. Rank Aggregation–Part 1

Arrow’s Criteria Revisited

Rank-Aggregation Methods

Borda Count

Average Rank

Simulated Game Data

Graph Theory Method of Rank Aggregation

A Refinement Step after Rank Aggregation

Rating Aggregation

Producing Rating Vectors from Rating Aggregation-Matrices

Summary of Aggregation Methods

Chapter 15. Rank Aggregation–Part 2

The Running Example

Solving the BILP

Multiple Optimal Solutions for the BILP

The LP Relaxation of the BILP

Constraint Relaxation

Sensitivity Analysis

Bounding

Summary of the Rank-Aggregation (by Optimization) Method

Revisiting the Rating-Differential Method

Rating Differential vs. Rank Aggregation

The Running Example

Chapter 16. Methods of Comparison

Qualitative Deviation between Two Ranked Lists

Kendall’s Tau

Kendall’s Tau on Full Lists

Kendall’s Tau on Partial Lists

Spearman’s Weighted Footrule on Full Lists

Spearman’s Weighted Footrule on Partial Lists

Partial Lists of Varying Length

Yardsticks: Comparing to a Known Standard

Yardsticks: Comparing to an Aggregated List

Retroactive Scoring

Future Predictions

Learning Curve

Distance to Hillside Form

Chapter 17. Data

Massey’s Sports Data Server

Pomeroy’s College Basketball Data

Scraping Your Own Data

Creating Pair-wise Comparison Matrices

Chapter 18. Epilogue

Analytic Hierarchy Process (AHP)

The Redmond Method

The Park-Newman Method

Logistic Regression/Markov Chain Method (LRMC)

Hochbaum Methods

Monte Carlo Simulations

Hard Core Statistical Analysis

And So Many Others

Glossary

Bibliography

Index

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset