臺灣博碩士論文加值系統

本文研究目的在於預測12-13年NBA總冠軍,蒐集12-13球季NBA常規賽數據,將每場球賽的數據整理分類,主要研究方法是使用B-Spline為基底建模,配適出每個球隊的實力曲線,再根據三種不同概念的指標,DIFF(Difference):比賽得失分差。GmSr(Game Score):自身球隊的效率分數,表示球隊攻擊能力。AntiGmSr(Anti-Game Score):比賽對手的效率分數,表示壓制對手的能力。最後,使用排名積分法和最適加權法來融合這三種不同指標分數,預測當年度總冠軍。此外,本文也將02-12年總共十個球季的總冠軍賽果加入預測判斷條件,使之更能加強對總冠軍預測的準確性。排名積分法預測2013年總冠軍為熱火隊、最適加權法預測為灰熊隊。

We apply a B-Spline approach to predict the playoff results of National Basketball Association (NBA). Here the features are composed of the regular seasons. Moreover, we use DIFF(diffentials),GmSr(gamescore) and AntiGmSr(the rival's gamescore) to rank to predict the final result. With 10-year records, the predict result is that for the best chance of 2013 NBA Championship is Heat.

1 緒論........................4
1.1 研究動機與目的..............4
1.2 本文架構...................4

2 文獻探討.....................5
2.1 NBA數據分析起源和發展........5
2.2 預測NBA冠軍的相關文獻........6

3 研究方法.....................7
3.1 建構曲線模型假設............7
3.2 knots個數選擇..............8
3.3 指標分數計算................9
3.4 總冠軍預測方法..............10

4 2002-2013年~NBA球隊數據分析...11
4.1 02-12年之總冠軍分析.........12
4.2 12-13年度總冠軍預測分析......15

5 總結與討論....................17

附錄...........................18

[1] Martin Manley's Basketball Heaven: The 1990 Edition. Doubleday, 1989.
[2] Dave Heeren. Basketball Abstract 1989-90. Prentice Hall, 1989.
[3] John Hollinger. Pro Basketball Forecast: 2005-06. Potomac Books Inc,
USA, 2005.9.30.
[4] Feifang Hu and James V. Zidek. Forecasting nba basketball playo out-
comes using the weighted likelihood. Institute of Mathematical Statistics,
45:385{395, 2004.
[5] Michael Lewis. Moneyball: The Art of Winning an Unfair Game. W W
Norton &; Co Inc, 2011.
[6] Mary J. Lindstrom. Penalized estimation of free-knot splines. Journal of
Computational and Graphical Statistics, pages 8:333{352, 1999.
[7] Bernard Loeelholz, Earl Bednar, and Kenneth W. Bauer. Predicting nba
games using neural networks. Journal of Quantitative Analysis in Sports,
5(1), 2009.
[8] Dean Oliver. Basketball On Paper: Rules And Tools For Performance
Analysis. Potomac Books Inc, 2005.
[9] David Ruppert. Selecting the number of knots for penalized splines. Journal
of Computational and Graphical Statistics, 11(4):735{757, 2002.
[10] Bob Ryan, Sam Smith, Leonard Koppett, Bob Bellotti, and Kenneth A.
Shouler. Total Basketball: The Ultimate Basketball Encyclopedia. Natl
Book Network, 2003.
[11] Thomas W. Sederberg. An introduction to b-spline curves. 3 2005.
[12] Na Wei. Predicting the outcome of nba playos using the nave bayes algo-
rithms. 2010. Department of Biomedical Engineering, College of Engineer-
ing, University of South Florida, Tampa, FL 33620, USA.
[13] Jackie B. Yang and Ching-Heng Lu. Predicting nba championship by learn-
ing from history data. 2012 Spring.