University:
Manchester University
Course:
MATH 105 | Basic Algebra
Academic year:

2024
Views:

183

Pages:

19
Author:

larbiwxcvbn

Predicting Outcomes of NCAA Basketball Tournament Games Project Goals y Collect data from NCAA Men’s Division I Basketball Tournaments y y y y y 2004-2007 Compute the relative strength for all the rating systems of the favored team for each game in the tournament. Compute relative strength by both ratio and difference to test the accuracy of the predictions of each rating system. Are all five rating systems statistically significant predictors of outcomes? Does one rating system predict the outcome better than the others? Is the relative strength a good predictor of the probability that the favored team wins that particular game? Rating Systems y RPI – Ratings Percentage Index y Pomeroy y Sagarin y Massey y Tournament Seeds Comparing the Systems Margin Home of Victory Court Strength of Schedule Strength of Record Opponents (win and Schedule losses) Range of Ratings RPI No Yes Yes Yes Yes 0.0 to 1.0 Pomeroy Yes Yes Yes No Yes 0.0 to 100 Sagarin Yes Yes Yes No Yes 0 to 100 Massey Yes Yes No No Yes -4.0 to 4.0 Tournament Seeds y The team’s actual seed is determined by a committee that has access to the other rating systems. y Does that mean that the tournament seeds should be a better predictor than the other systems? Any Problems? y Massey 2005 y Pomeroy 2007 (Instead of 252, we had 189) y Ties? - One for RPI - Two for seeds Rating Systems’ Predicting Ability Rating System Success rate for predicting winner Seeds 73.2% Pomeroy 73.0% Massey 72.6% Sagarin 72.2% RPI 70.9% Statistically Significant? y Hypothesis Test for One Sample (RPI) H 0 : p = .50 H A : p > .50 System is ineffective System is efficient in predicting where p is the population proportion of correct prediction After calculating we obtain a z-score = 6.62. Any z-score > 3 will result in a p-value that is approaching zero. Thus our p-value < 0.05, so we can conclude our alternate hypothesis. Which one is better? y Hypothesis Test for two samples (Seeds and RPI) H 0 : p1 = p2 No difference between seeds and RPI H A : p1 ≠ p2 There is a difference between the two After calculating we obtain a z-score = .58. Using a statistical table we obtain a p-value = .562 . Thus our p-value is not less than 0.05, so we can conclude that our alternate hypothesis is false. There is not significant difference between the two, seeds and RPI. Predicting Game Outcomes Using Relative Strength y Logistic Regression Model + Log of odds is a linear function of relative strength p (x) = Predicting the probability of winning β0 = 0 p (x) = Logistic Curve P(x) x Sagarin RS Df Seeds RS Df Example y Sagarin RS-Df 2004 Elite 8 Connecticut vs. Alabama Connecticut’s Rating = 90.74 Alabama’s Rating = 84.46 Relative Strength of the Favored Team = 6.28 *Look at our Graph* Connecticut had approximately an 80% chance of winning according to Sagarin’s RS_Df. β1 Sig. (P-Value) Cox and Snell % of games fav. team won Sagarin RS_Df .206 .000 .312 72.2 Sagarin RS_Rt .930 .000 .202 72.2 RPI RS_Df 24.068 .000 .271 70.9 RPI RS_Rt .864 .000 .181 70.9 Pomeroy RS_Df .219 .000 .297 73.0 Pomeroy RS_Rt .953 .000 .217 73.0 Massey RS_Df 3.078 .000 .278 72.6 Massey RS_Rt .867 .000 .231 72.6 Seed RS_Df .200 .000 .301 73.2 Seed RS_Rt .472 .000 .293 73.2 R2 Top 3 y Sagarin_Df- p(x) = y Seeds_Df- p(x) = y Pomeroy_Df- p(x) = What happens when the relative strength increases 1 unit? = β1x where relative strength = x y Then by raising both sides to the power of e, and we obtain: where relative strength = x y Thus when we adding one to the relative strength we obtain: y ln * Conclusions y All 5 rating systems were statistically significant. y Not one rating systems more accurate than the others. y Was the prediction about seeds correct? y Ratio vs. Difference y Top 3 - Sagarin RS_Df - Pomeroy RS_Df - Seeds RS_Df