Sports Metric Forecasting

By William Mallios

Outcomes of major league gameswinning/losing margins and total points scored relative to the odds makers lines in baseball, basketball and footballare graphed in terms of sports metric candlestick charts and then forecast in terms of adaptive drift modeling. The charts are constructed to reveal ad hoc forecasting patterns that may contribute to effective forecasting. These patterns are then included with variables contained in major sports data bases. The augmented data bases then provide input variables in the drift modeling forecasts.

• Language: English
• Publisher: Xlibris US
• Release date: Nov 17, 2014
• ISBN: 9781503517325

Book preview

SPORTS METRIC

FORECASTING

Guides to Beating the Spreads

in Major League Baseball,

Basketball, and Football Games

William Mallios

www.MalliosAssociates

Copyright © 2014 by William Mallios.

ISBN:      eBook      978-1-5035-1732-5

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: 11/14/2014

Xlibris

1-888-795-4274

www.Xlibris.com

636643

Contents

1. Introduction

1.1 Forecasting Scenarios: Overview

1.2 Review of Sports Metric Publications

1.3 Sports Pundits: Commentaries from the Past

1.4 Online Sports Gambling

1.5 Sports Hedge Funds

1.6 Sports Big Data and Meta-Analyses

2. NFL Sports Metric Charts

2.1 Seattle Seahawks in 2012-14: Leaders in Beating the Difference Spreads

2.2 Denver Broncos: 2013–2014

2.3 Baltimore vs. San Francisco: The 2012–13 Finale

2.4 New York Giants: The 2011–12 and 2007–08 Super Bowls

2.5 New England in 2007-08 and 2011-12: Misery for the Patriot Faithful

2.6 San Diego in 2012–13: A Lack of Motivation?

2.7 Houston Texans 2012–14: The Woeful Years

3. NBA Sports Metric Charts

3.1 Miami Heat: 2011–13: Two in a Row

3.2 San Antonio Spurs 2012–13: Unexpected Finalists

3.3 Indiana 2012–2013: Almost Titlists

3.4 Oklahoma City Thunder 2012–13: for the Loss of a Horse, the War was Lost

3.5 Los Angeles Lakers 2008–2010: A Tradition

3.6 Dallas 2005–07: The Playoff Games

3.7 Los Angeles Clippers 2013–14: Difficult Scenarios

4. MLB Sports Metric Charts

4.1 San Francisco Giants 2012: Sweeping the Favored Tigers

4.2 Boston Red Sox 2013: Banner Year

4.3 New York Yankees 2012: Playoff Blues

4.4 Houston Astros 2013: Tracking Losers for Winning Bets

5. Sports Databases: Listings of Performance Variables

5.1 NFL Databases

5.2 NBA Databases

5.3 MLB Databases

6. Reducing Database Dimensions

6.1 Reducing the Variable Dimensions of NFL Databases: Factor Analyses of the Baltimore Ravens Database

6.2 Reducing the Variable Dimensions of NBA Databases: Factor Analyses of the San Antonio Spurs Database

6.3 Reducing the Variable Dimensions of MLB Databases: Factor Analyses of the New York Yankees 2012 Database

7. NFL Forecasting Applications

7.1 The 2012–2013 Baltimore Ravens

7.2 The 2011–2012 New York Giants

7.3 The 2011–13 New England Patriots

8. NBA Forecasting Applications: The 2012–2013 Finals

8.1 Miami and San Antonio: Forecasts of Winning/Losing Margins and Total Points Scored

8.2 Forecasting Per Game Points Scored byLebron James in 2012–2013

9. MLB Forecasting Applications

9.1 New York Yankees: 2012 Playoff Game Forecasts

9.2 Forecasts of Runs Allowed: CC Sabathia

9.3 Boston Red Sox: 2013 Playoff Game Forecasts

10. Dynamic Cointegrated Time Processes

10.1 Modeling Cointegrated Time Processes

10.2 Modeling Dynamic Cointegrated Time Processes: Adaptive Drift Modeling

10.3 Adaptive Drift Modeling in Sports Gambling Markets

10.4 Risk Assessment: Modeling Volatility

Appendix: Modeling Parallels between Financial and Sports Gambling Markets

A.1 The Genesis of Sports Metric Analyses: Financial Market Forecasting

A.2 In the Matter of Insider Trading

A.3 Charting Daily Equity Price Movements during Periods of Insider Trading:

References

1. INTRODUCTION

1.1 Forecasting Scenarios: Overview

The focus is on forecasting major league game outcomes—winning/losing margins and total points scored—relative to the oddsmakers’ lines (spreads). Forecasts of individual player performances are included. How many innings will CC Sabathia last in his next start and how many earned runs will he allow? How many points will LeBron James score in his next game?

Forecasting proceeds through two stages: (1) sports metric graphics and (2) stochastic modeling. The graphics are hybrids of the Japanese candlestick charts (JCC) used in monitoring price movements in financial markets. The JCC are illustrated in the appendix with reference to recent cases of insider trading.

Successive game outcomes for individual NBA, NFL and MLB teams are graphed in terms of sports metric candlestick charts. The charts are constructed to reveal ad hoc forecasting patterns. These patterns—combined with variables in sports databases—then become input variables in stochastic modeling procedures. Input variables are evaluated regarding their statistical effects, if any, on game outcomes. The modeling procedures are dynamic to allow model structures to adapt and drift in the process of accommodating inevitable changes in an individual team’s performances over time. Accordingly, the modeling procedure is termed adaptive drift modeling.

A sports metric candlestick chart is a depiction of successive game outcome variables for team i, say, i = San Francisco 49ers (SF). Each game is represented by a vertical candlestick containing (1) a body and (2) a wick that extends above or below the body. The upper and lower values of the body are determined by the values of D and LD where D denotes the SF winning/losing margin and LD the line on D. If D > LD, the body is white, with D and LD defining the body’s maximum and minimum values. If LD > D, the body is black, with LD and D defining the body’s maximum and minimum values. The body reduces to a horizontal straight line if D = LD. See Figure 1.1.1 for illustrations. The length of the candlestick wick is determined by the value of GST, the difference between total points scored (TP) in the game and the corresponding line on TP (LT), i.e., GST = TP-LT. The wick extends above the body when GST > 0 and below the body when GST < 0. There is no wick when GST=0.

1.JPG

Figure 1.1.1. Sports metric candlestick illustrations

Figure 1.1.2 presents the SF candlestick chart for the 2012–2013 season. The nineteen regular season and playoff games begin with the 30–22 win over Green Bay and end with the 31–34 Super Bowl loss to Baltimore (BAL). Regarding ad hoc patterns, note that a SF loss or tie is always followed by a win. (Game 9 was a 24–24 tie against St. Louis.) Note also that two successive white bodies occur on five occasions. Following each of these successive white bodies, SF lost and failed to beat the spread, i.e., two successive white bodies are always followed by a black body loss. The SF Super Bowl loss to BAL was preceded by the two successive green body wins in playoff games against Green Bay (week 17) and Atlanta (week 18). This ad hoc pattern correctly suggested a forthcoming black body Super Bowl loss for SF (a 4.5-point favorite). In §7.1, the corresponding adaptive drift modeling (ADM) forecast has SF losing to BAL by two points. (A preliminary note on NFL database sample sizes: The ADM forecast for the SF vs. BAL game outcome is based on all SF regular season and playoff game performances from the first regular season game in 2009–2010 season through the week 18 of the 2012–2013 season. In general, all forecasting models, both for regular season and playoff games, are undated following each successive game played so as to allow for model changes, if any, in predictor variables and/or predictor variable effects. Sample sizes for all forecasts are sufficiently large.)

2.JPG

Figure 1.1.2 San Francisco 49ers candlestick chart for the 2012–13 regular season and playoffs ending with a 31–34 loss to the Baltimore Ravens in the Super Bowl. Note that two successive white bodies are always followed by a black body loss. Moreover, a loss/tie is always followed by a win. Game 9 was a 24–24 tie against St. Louis.

Forecasting a game outcome relative to its spread is, conceptually, a matter of modeling interactive types of human behavior. Realistic modeling of such behavior is not parsimonious and has been said to require analyses of Big Data, a term that is fashionably used to describe massive databases that can be stored online at low cost. This writing illustrates that human factor effects in sports competition can be estimated in terms of smaller sets of alias variables. Specifically, game outcomes are primarily determined by game participants (players and coaches) whose present performances are influenced by numerous aspects of their past performances. Measures of these past performances, when properly quantified, become aliases of human factor variables that affect subsequent performances and game outcomes. Consider a simple illustration. Regarding team performances relative to the spreads, gambling shocks—defined as differences between game outcomes and corresponding spreads—are measures that partially reflect human factor variables affecting subsequent game outcomes. If a heavily favored team loses to the underdog, the corresponding negative gambling shock reflects motivational variables that influence player/team performances in their next game and in their subsequent rematch against the underdog. Results of the adaptive drift modeling applications provide support for such conjectures.

Conventional spreads/odds on outcomes are based on the gambling public’s expectations of game outcomes. Given that spreads/odds are chosen to approximate these expectations and then adjusted to approximately divide the gambling public’s monies on each side of the spread, losing bets pay off winning bets while bookies take commissions (usually about 10 percent) on each bet. When spreads are sufficiently close to outcomes, the markets are said to be efficient—in which case the gambling public’s expectations are said to be rational. Spreads are difficult to beat in efficient markets. When spreads are not sufficiently close to outcomes—which is the case for most games—the markets are said to be inefficient and the gambling public’s expectations irrational. When markets are inefficient, rules for profitable forecasting exist, but innovative modeling is usually necessary.

In the early 1990s, the writer interviewed Roxy Roxborough, founder of Las Vegas Sports Consultants. He said that professional gamblers dominated NBA betting—except for the playoffs—whereas the public dominated NFL betting. Because of the large number of games, the public thought that regular season NBA games were not significant. Per NFL team, there were relatively few regular season games. For NBA games, professional gamblers profited through arbitrage when there was a sufficient discrepancy in point spreads. Consider, for example, a Bulls-Knicks game where Chicago bookies favor the Bulls by 3.5 points, and the New York bookies favor the Bulls by 1.5. An arbitrageur’s betting strategy: In NY, take the Knicks +3.5 points, while in Chicago, take the Bulls -1.5 points. If the Knicks lose by 2 or 3, both bets win. Otherwise, it’s a standoff except for the bookies’ fees. Given the rise and immense popularity of the sports media markets—fantasy sports, ESPN sports, and the social media—Roxborough’s 1990s scenario has undergone changes. However, the extent of such changes remains unclear given that roughly 99 percent of U.S. sports gambling is illegal as of 2014. (Asher, 6/26/12)

The preponderance of gambling market inefficiency is illustrated in terms of recent playoff game outcomes:

1. The Denver Broncos were favored by 1.5 points and lost to Seattle 43–8 in Super Bowl 2014.

2. The San Francisco 49ers were favored by 4.5 points and lost to Baltimore 34–31 in Super Bowl 2013.

3. The New England Patriots were favored by 4 points and lost to the New York Giants 21–17 in Super Bowl 2012.

4. The Detroit Tigers were favored in all four games of the 2012 World Series and were swept