By Kelly Galarneau. Don’t Attack Me Until the End.  My probability model is not perfect, and never will be.  We will discuss the assumptions and limitations.

Slides:



Advertisements
Similar presentations
NCAA Mens Basketball Tournament The 64 teams that play in the elimination rounds are divided into four groups of 16 teams. The four groups represent the.
Advertisements

Chapter 6 – Normal Probability Distributions
Rules for Means and Variances
Sampling Distributions Welcome to inference!!!! Chapter 9.
 These 100 seniors make up one possible sample. All seniors in Howard County make up the population.  The sample mean ( ) is and the sample standard.
Other Inventory Models 1. Continuous Review or Q System 2 The EOQ model is based on several assumptions, one being that there is a constant demand. This.
Simulation Operations -- Prof. Juran.
Session 7a. Decision Models -- Prof. Juran2 Overview Monte Carlo Simulation –Basic concepts and history Excel Tricks –RAND(), IF, Boolean Crystal Ball.
Continuous Random Variables. L. Wang, Department of Statistics University of South Carolina; Slide 2 Continuous Random Variable A continuous random variable.
Basic Business Statistics, 10e © 2006 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Confidence Interval Estimation Basic Business Statistics 10 th Edition.
1 The Basics of Regression Regression is a statistical technique that can ultimately be used for forecasting.
1 Hypothesis Testing In this section I want to review a few things and then introduce hypothesis testing.
NFL Football Mark: The History of the NFL andSherrie: How the game is played, the players, and teams.
Discrete Probability Distributions
8-1 Copyright ©2011 Pearson Education, Inc. publishing as Prentice Hall Chapter 8 Confidence Interval Estimation Statistics for Managers using Microsoft.
Copyright ©2011 Pearson Education 8-1 Chapter 8 Confidence Interval Estimation Statistics for Managers using Microsoft Excel 6 th Global Edition.
 Binomial distributions for sample counts  Binomial distributions in statistical sampling  Finding binomial probabilities  Binomial mean and standard.
First we need to understand the variables. A random variable is a value of an outcome such as counting the number of heads when flipping a coin, which.
BCOR 1020 Business Statistics Lecture 11 – February 21, 2008.
Simulation.
Statistics for Managers Using Microsoft Excel, 4e © 2004 Prentice-Hall, Inc. Chap 7-1 Chapter 7 Confidence Interval Estimation Statistics for Managers.
Binomial PDF and CDF Section Starter Five marbles are on a table. Two of them are going to be painted with a “W” and the rest will be painted.
Application of Random Variables
CHAPTER 2 COMPARING TWO PROPORTIONS Objectives: Students will be able to: 1) Test a difference in proportions 2) Use technology to simulate a difference.
HAIL TO THE REDSKINS! REDSKINS 13 RAVENS 12. Hail to the REDSKINS!! Game Schedule Week 1–Sept. 16, 9:00am, Field #1 WIN REDSKINS vs. Chiefs Week 2–Sept.
Basic Business Statistics, 11e © 2009 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Confidence Interval Estimation Basic Business Statistics 11 th Edition.
Ch 8 Estimating with Confidence. Today’s Objectives ✓ I can interpret a confidence level. ✓ I can interpret a confidence interval in context. ✓ I can.
Confidence Interval Estimation
Basic Business Statistics, 10e © 2006 Prentice-Hall, Inc. Chap 8-1 Chapter 8 Confidence Interval Estimation Basic Business Statistics 11 th Edition.
Monte Carlo Simulation and Personal Finance Jacob Foley.
PARAMETRIC STATISTICAL INFERENCE
Follow-Up on Yesterday’s last problem. Then we’ll review. Sit in the groups below Brendan and Tim are playing in an MB golf tournament. Their scores vary.
Continuous Random Variables Continuous Random Variables Chapter 6.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
From Theory to Practice: Inference about a Population Mean, Two Sample T Tests, Inference about a Population Proportion Chapters etc.
P. STATISTICS LESSON 8.2 ( DAY 1 )
14.3 Simulation Techniques and the Monte Carlo Method simulation technique A simulation technique uses a probability experiment to mimic a real-life situation.
LECTURE 14 THURSDAY, 12 March STA 291 Spring
Special Topics. Mean of a Probability Model The mean of a set of observations is the ordinary average. The mean of a probability model is also an average,
© 2005 McGraw-Hill Ryerson Ltd. 5-1 Statistics A First Course Donald H. Sanders Robert K. Smidt Aminmohamed Adatia Glenn A. Larson.
N.F.C. teams : Dallas Cowboys Washington Redskins New York Giants Philadelphia Eagles Minnesota Vikings Green Bay Packers Chicago Bears Detroit Lions.
By: Jim Ogle Charles Henkel Kyle Marcangelo.  We are trying to find what team would win in a hypothetical game based on their stats for the
7.2 Means and variances of Random Variables (weighted average) Mean of a sample is X bar, Mean of a probability distribution is μ.
8.2 The Geometric Distribution. Definition: “The Geometric Setting” : Definition: “The Geometric Setting” : A situation is said to be a “GEOMETRIC SETTING”,
Chap 8-1 Chapter 8 Confidence Interval Estimation Statistics for Managers Using Microsoft Excel 7 th Edition, Global Edition Copyright ©2014 Pearson Education.
Area Under the Curve We want to approximate the area between a curve (y=x 2 +1) and the x-axis from x=0 to x=7 We want to approximate the area between.
The Practice of Statistics Third Edition Chapter 7: Random Variables Copyright © 2008 by W. H. Freeman & Company Daniel S. Yates.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
Simulations and Probability An Internal Achievement Standard worth 2 Credits.
Football originated during the 1823 (it was called rugby) in England. Rugby was then introduced to the North Americans by the British army. It was introduced.
David Cleveland.  The BCS is how college football decides who is going to be in the national championship every year  BCS stands for Bowl Championship.
One loss ruins a season in current system It’s what the fans want Every conference is represented --All conferences receive a share of the revenue --All.
Simulation Chapter 16 of Quantitative Methods for Business, by Anderson, Sweeney and Williams Read sections 16.1, 16.2, 16.3, 16.4, and Appendix 16.1.
Sampling Distributions: Suppose I randomly select 100 seniors in Anne Arundel County and record each one’s GPA
Binomial Distributions Chapter 5.3 – Probability Distributions and Predictions Mathematics of Data Management (Nelson) MDM 4U Authors: Gary Greer (with.
6.3 Binomial and Geometric Random Variables
Rick Walker Evaluation of Out-of-Tolerance Risk 1 Evaluation of Out-of-Tolerance Risk in Measuring and Test Equipment Rick Walker Fluke - Hart Scientific.
Statistics 16 Random Variables. Expected Value: Center A random variable assumes a value based on the outcome of a random event. –We use a capital letter,
The NFL is split into two CONFERENCES. And each conference is broken down into four DIVISIONS. American Football Conference (AFC) National Football Conference.
By: Carlos Licon. A Football Life  It had been more than a decade since Ray Lewis won his first and only Super Bowl.  Every year he has had a single.
Aaron Rodgers Clay Matthews. Who started the Packers? Curly Lambeau and George Whitney started the Green Bay Packers in August 11, The Indian Packaging.
Lecture #14 Thursday, October 6, 2016 Textbook: Sections 8.4, 8.5, 8.6
Computer Simulation Henry C. Co Technology and Operations Management,
Chapter 7 Confidence Interval Estimation
Statistical Modelling
Greenbay Packers Official Site
Predicting NFL Game Outcomes: Back-Propagating MLP
Example Project Presentation
Sampling Distributions (§ )
Presentation transcript:

By Kelly Galarneau

Don’t Attack Me Until the End.  My probability model is not perfect, and never will be.  We will discuss the assumptions and limitations of this model.  History and Concept  NCAA tournament  Rating Systems

Jeff Sagarin’s Ratings  Published in the USA Today since  NCAA men’s basketball tournament selections  Bowl Championship Series (BCS) selections  Secret Formula  Each team’s rating  Home field advantage (constant for all teams)

Any Given Sunday  We know that a team’s performance is not always constant. On some days they play better, others worse.  Many things naturally vary according to the normal distribution, so perhaps we can assume NFL ratings do the same.  In an , Sagarin suggested that I use a standard deviation of 15 or 16 (perhaps 15.5). I chose 16.

Normal Distribution  Here is a graph for the theoretical performance of the Tennessee Titans, with a mean of 32.7 and standard deviation of 16.

Normal Curves  Now compare two teams, Titans with average of 32.7 and the Lions with an average of We’ll pretend the Lions are playing at home and add in Sagarin’s prescribed 3.02.

Monte Carlo Simulation  Used for simulating probability situations.  Mostly used for business and finance applications.  Allows us to vary a parameter according to whatever distribution we choose (uniform, normal, Poisson, exponential, etc.)  As the parameter is changing, we can observe the effect on other variables.  I am using free software from (Rutgers University)

How each trial works  We let the computer pick a random “performance rating” from the normal distribution for each team.  We see which one is greater and tally that as a win for that team.  Then repeat 1000 times. This graph shows the results of one trial, with a win going to the Titans.

The graph on the left represents a game in which the Lions perform better than average, the Titans perform worse than average, but the Titans still get the win. The graph on the right represents a game in which the Lions pull off the upset. Then we repeat… …1000 times.

Here is a graph of 1000 trials

Here is my actual Excel programming: ABCDEFG 1 TeamRatingS.D. NormalIf/thenOutput 2 Titans =gennormal(C2,D2)=if(E2>E3,1,0)=simoutput(F2) 3 Lions =gennormal(C3,D3)=if(E3>E2,1,0)=simoutput(F3) ABCDEFG 1 TeamRatingS.D. NormalIf/thenOutput 2 Titans Lions Here is one trial: This matchup gives us: Titans 88.8% (1000 trials) Lions 11.2%

Do we need a simulation?  I tried to approach the probability calculation from a theoretical perspective.  My thought was to assign each team a random variable X and Y and vary them according to a normal distribution. So: X ~ N(μ₁, σ) and Y ~ N(μ₂, σ).  A statistics textbook led me to the idea of multiplying them to get a 3-D probability distribution [X, Y, and P(X,Y)].

A 3-D Probability Distribution Here, X ~ N(35.72,16) and Y ~ N(10.84,16) Where is X > Y? YX Titans 88.8% Lions 11.2%

Integrals  To find the area under this surface, we need to evaluate the double integral.  Mathematica will not evaluate the exact integral, but it will give us a decimal approximation.

Comparing the results  Mathematica Input : NIntegrate[NIntegrate[(1/((16^2)*2*Pi))*(E^((- 1/2)*((x-35.72)/16)^2))*(E^((-1/2)*((y- 7.82)/16)^2)),{y,x,300}],{x,-300,300}]  Mathematica gives us: Titans 89.1% Lions 10.9%  The Monte Carlo with 1000 trials gave us: Titans 88.8% Lions 11.2%

The Bracket  6 teams from each conference (AFC and NFC) make the playoffs. The division winners are seeded 1-4 by record, the wild card teams are seeded 5 and 6 by record.  In each game the higher seed gets home field advantage.  In the “Wild Card” round of the playoffs:  The 3 seed plays the 6 seed  The 4 seed plays the 5 seed.  The 1 and 2 seeds get a first round bye.  In the “Divisional” round the 1 seed gets to play the lower seed of the two advancing teams.

So that means… or seed wins 6 seed wins

Possibilities…  In a playoff bracket with 12 teams, there are 11 games to be played, with a total of 2^11=2048 possibilities.  In my Excel spreadsheet I have made use of the complement rule and referencing so that I only need to calculate 66 matchups.

Assumptions and Limitations  We are using offensive and defensive statistics from the regular season, we are assuming that teams continue similar play into the post-season.  We are assuming that overall performance is normally distributed.  We are assuming that home field advantage is the same constant for all teams.  We are not taking anything into account for weather, injuries, matchups that might be significant, or any other variables.

2007 trial run (AFC) SeedTeamRating Div. Game Champ. Superbowl Superbowl Win 1pats % 73.9% 48.3% 32.4% 2colts % 58.0% 25.1% 13.7% 3chargers % 28.6% 11.9% 6.5% 4steelers % 12.8% 4.3% 1.9% 5jags % 19.2% 8.0% 4.0% 6titans % 7.4% 2.5% 1.0% %

2007 trial run (NFC) SeedTeam Rating Div. Game Champ. Superbowl Superbowl Win 1cowboys % 62.6% 34.9% 14.5% 2packers % 65.9% 35.6% 15.6% 3seahawks % 18.4% 7.3% 2.3% 4bucs % 12.0% 4.2% 1.2% 5giants % 26.4% 12.7% 5.1% 6redskins % 14.7% 5.2% 1.7% %

2008 Playoff Prediction  My OFFICIAL prediction won’t be available until after week 17 (end of regular season is Sunday, Dec. 28)  Over Christmas break you can access it at: kgalarneau.wikispaces.com/NFL+Prediction kgalarneau.wikispaces.com/NFL+Prediction  Feel free to me with your comments or suggestions:

2008 Prediction (As of Dec. 15) SeedTeamRating Div. Game Champ. Superbowl Superbowl Win NFC1Giants % 63.6% 37.7% 19.0% 2Panthers % 57.5% 28.1% 13.5% 3Vikings % 25.4% 11.8% 5.3% 4Cardinals % 16.3% 5.8% 2.2% 5Cowboys % 20.0% 8.4% 3.6% 6Bucs % 17.2% 8.2% 3.7% sum % AFC1Titans % 64.6% 39.6% 22.0% 2Steelers % 65.1% 31.8% 16.9% 3Jets % 14.7% 5.1% 2.0% 4Broncos % 9.7% 2.8% 1.0% 5Colts % 24.3% 10.8% 5.4% 6Ravens % 21.5% 9.9% 5.3% sum % standard dev. = 16 home field advantage = 2.58

QQuestions? Comments?

But Wait… …There’s More!

History of Seeds  The NFL has been using the 12 team playoff system since SeedDivisionChamp.SuperbowlSuperbowl Win sum

By Percentage: SeedDivisionalChampSuperbowlSuperbowl Win 138.9%50.0%44.4% 237.5%27.8% 369.4%9.7%5.6% 466.7%6.9%11.1% 533.3%5.6%2.8%5.6% 630.6%1.4%2.8%5.6% sum2111

Superbowl Winners by Seed since 1990 First 10 years – 7 #1 seedsLast 8 years – Only 1 #1 seed 9/10 are 1 or 2 seeds4/8 are 1 or 2 seeds

The End. Special thanks to: Harry Geiser Steve Havlichek