A Simpler Football Simulation: A New Paradigm That Re-Frames the G.O.A.T. Debate

By Andrew R. Crawford

The book takes the concept of computer-based simulations for predicting American football gameslike the Madden video games for exampleand attempts to devise a simpler method. With just a word processor, some spreadsheets, a fundamental understanding of percentages and probabilities, and a few hours of work, anyone can achieve results that are remarkably similar to those of the more sophisticated computer models. Chapters 1 to 5 outline the method and its rationale. Chapters 6 to 9 detail an actual simulated contest play-by-play. Chapters 10 to 11 and the appendices utilize the method described to determine which professional football team of the past fifty years was truly the best.

• Language: English
• Publisher: Xlibris US
• Release date: Oct 19, 2017
• ISBN: 9781543459210

Andrew R. Crawford

Andrew R. Crawford has been a football fan since watching his first Super Bowl in January, 1977, but he has been a consumer of sports statistics since buying his first pack of baseball cards several months before that. His love of both watching and writing about sports, as well as an aptitude for mathematics, have been a mutually beneficial symbiosis for over four decades. He and his wife of 23 years live in Athens, Ohio, where he works as a sportswriter and as a high school track and field coach.

Book preview

Copyright © 2017 by Andrew R. Crawford.

ISBN:                      Softcover                        978-1-5434-5922-7

                                eBook                              978-1-5434-5921-0

All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from the copyright owner.

Any people depicted in stock imagery provided by Thinkstock are models, and such images are being used for illustrative purposes only.

Certain stock imagery © Thinkstock.

Rev. date: 10/19/2017

Xlibris

1-888-795-4274

www.Xlibris.com

765816

CONTENTS

Acknowledgments

Introduction

Chapter 1     Relevant Statistics

Chapter 2     Limitations of the Process

Chapter 3     Setting Up a Simulated Game

Chapter 4     Preparing for Kickoff

Chapter 5     The In-Game Tracker

Chapter 6     Game Time!

Chapter 7     The Second Simulation

Chapter 8     Interpreting the Results

Chapter 9     Final Considerations

Chapter 10   The All-Time Tournament

Chapter 11   Concluding Remarks

Appendix A   The Final Rankings

Appendix B   Season Summaries

Acknowledgments

I developed this method over many years, perfecting it as much as possible by running simulations involving the Super Bowl champions. When faced with a dilemma for which the numbers could not adequately account, I tweaked the underlying premises and began anew. These trials would eventually become the tournament that is described in chapter 10.

Two websites were invaluable in helping me assemble the season statistics of the participating teams. I used pro-football-reference.com for the vast majority of this work, and when I could not find the information there, I went to footballdb.com.

Introduction

It all started—for me, anyway—with Dream Season, the series that ESPN first aired in 1989.

Of course, Greatest of All Time debates have been engaging sports fans for years, but this was different. It was the first time that such an argument was going to be settled not by columnists relying on the eye test and citing anecdotal evidence, but by unbiased statistics fed into a computer. The fact that the combined muscle of ESPN and NFL Films was behind Dream Season lent it an undeniable air of gravitas.

Until 1989, football statistics had never risen above the status of curiosities on the backs of trading cards. Suddenly there arose a credible way to take those numbers and create a new paradigm in which to frame a Greatest of All Time debate. That idea fascinated me, and I’ve never lost that fascination. If anything, it has grown over the years as the computer models have become more sophisticated. One need only take a mental step back to 1989 to be in awe of where the technology has brought us. An annual video game can now contain the equivalent of ESPN’s original opus.

With this fascination, however, has come equal parts frustration. I’ve always been a person who is more interested in the premise than in the conclusion. When the original Dream Season ended with the 1978 Steelers beating the 1972 Dolphins in the mythical championship game, I wanted to know how the result was reached.

Every computer simulation contains a kind of disclaimer telling us that the figures have been run ten thousand times or more, but the formulas themselves are guarded like secret recipes. As those of us who used to follow the computer rankings from the old Bowl Championship Series well remember, drastically different results can be achieved depending on which statistics one uses and how much importance is placed on each.

I wrote this book to give stat geeks who are not necessarily computer geeks an alternative. My goal was to develop a simple system by which anyone could run a simulated contest between two football teams and achieve a plausible result. Rather than relying on a high-powered computer, this method requires nothing more than a few spreadsheets and a word processor. Instead of spending days on computer programming, you’ll be able to accumulate the relevant season statistics in just a few hours. You’ll need to run only two simulations—not ten thousand. And most importantly, instead of having to trust the conclusion without knowing the premise, you’ll see the play-by-play for each game develop right in front of you.

This method isn’t perfect. No computer model is. That’s why every expert seems to have his or her own favorite. I have tried to be forthright in pointing out where the flaws in mine might be. But I do believe that the scores and statistics that result from using this method will often be very close to those of the more sophisticated ones and achieved at a fraction of the effort.

Chapter 1

Relevant Statistics

This chapter addresses eight statistical categories. The first five require that we compute totals for a team’s offense, as well as opponents’ statistics versus its defense. With the final three categories, two of which are special teams’ metrics, we need know only the statistics of the team in question.

1. Rushing

• rushing attempts, resulting yards gained, and fumbles lost while on offense

• rushing attempts by opponent, resulting yards allowed to opponent, and fumbles recovered while on defense

2. Passing

• passing attempts, completions, and resulting yards gained while on offense, and interceptions by opponent

• passing attempts and completions by opponent, resulting yards allowed to opponent, and interceptions made while on defense

Here’s the first of many controversial ideas that I will propose. For our purposes, touchdowns scored are irrelevant! After all, a touchdown is merely the result of a team’s matriculatin’ the ball down the field, in the words of Hank Stram. Gain enough yards at the right time, and you’re in the end zone. The teams that score the most touchdowns are the teams that gain the most yards. This may not always be true when looking at small sample sizes such as individual games, but it is certainly borne out over the long term.

3. Quarterback sacks

• sacks allowed and yards lost while on offense

• sacks achieved and resulting yards earned while on defense

4. Penalties

• number of penalties called on and yards penalized while on offense

• number of penalties called on and yards penalized of opponent

This category is unique. It can reasonably be argued that better teams force their opponents into committing more penalties. My system, however, measures penalties as a normalization technique, rather than as a reflection of a team’s defensive prowess. Simply put, there are more penalties called in today’s games than in earlier