Erasmus university rotterdam
Research question & hypotheses
Download 1.62 Mb.
|
- Navigate this page:
- Methodology
Research question & hypothesesAfter reviewing the existing literature in section 2, you can conclude that there are a lot articles about the favorite longshot bias in different kinds of sports, but there is not much according to tennis. A larger and more recent research on the favorite longshot bias in tennis will contribute to the existing literature. The research question that is addressed in the thesis is: ‘Is there a favorite longshot bias in ATP and WTA tennis?’ Besides the favorite longshot bias in general, this study will look at the favorite longshot bias in other aspects in tennis; like gender, surface and years. The following four hypotheses will be investigated. H1: There exists a favorite longshot bias in ATP and WTA tennis. H2: The favorite longshot bias is stronger in women tournaments. H3a: The favorite longshot bias is becoming stronger over the years. H3b: The favorite longshot bias is becoming stronger over the years in ATP tournaments. H3c: The favorite longshot bias is becoming stronger over the years in WTA tournaments. H4a: At gravel tournaments occurs a stronger favorite longshot bias than in hard-court tournaments. H4b: At gravel tournaments occurs a stronger favorite longshot bias than in hard-court tournaments in ATP tournaments. H4c: At gravel tournaments occurs a stronger favorite longshot bias than in hard-court tournaments in WTA tournaments. The expectation of the first hypothesis is based on the outcomes Forrest & Mchale (2007). This article is discussed in section 2.3. The outcome of this research was that there exists a favorite longshot bias in tennis. Based on this article, the expectation is the same for this study, which means that a favorite longshot bias in ATP and WTA tennis exists. For the other tree hypothesis, it is hard to have an expectation, because there are no articles that looked at the favorite longshot bias for men and women, the surface or the trend over the years before. So the hypotheses are not based on literature, but on own experience and interest. The expectation of the second hypothesis is that a favorite longshot bias exist in men and women tournaments, but in women tournaments the bias will be stronger. This expectation is based on the fact that the differences in level in women tennis are bigger than in men’s tennis. This means that favorites are playing much better tennis than the underdogs and the gap between favorites and underdogs is bigger in women tennis. The favorites in women tennis are more constant and there is less chance to have a ‘surprising’ win of an underdog in WTA tennis. The third hypothesis tests the trend of the favorite longshot bias over the years. The last years the favorites (top 10 players) are getting better and better. In Appendix C to F the results of the 4 grand slam tournaments from 1950 to 2014 are presented. For example looking at the ATP winners of Roland Garros you see that Rafael Nadal won from 2005 till 2014, except of 2009. In the years before there was no one who won so much Roland Garros titles in a row. This is also the case with ATP Wimbledon – Federer / Sampras, WTA Wimbledon – Williams, WTA Roland Garros – Henin-Hardenne, ATP US Open Federer and WTA US Open Williams. The trend is not visible in every category in every tournament, but if you look at the overall results, you see that in the recent years there are more the same winners than in the earlier years. This means the favorites are getting better and the difference in level with the underdogs are getting bigger. This is the reason why the expectation that the favorite longshot bias is becoming stronger over the years. The last hypothesis tests the favorite longshot bias at different kind of surfaces. Most tournaments investigated in this study, are played on gravel (30%) and hard-court (59%). Only 11% of the tournaments in this study are played on grass, that why this surface is excluded in the hypothesis. The expectation is that in gravel tournaments there is a stronger bias than in hard-court tournaments (indoor and outdoor). The expectation is also based on the historical results in Appendix C to F. Comparing Roland Garros (gravel) with Australian Open and US Open (both hard-court), we see more the same winners in Roland Garros. This can be an indication that favorite players who prefer gravel are more constant at winning games, than players who have a preference for hard-court. This means that the gap in level between gravel players is larger, than the gap between hard-court players is smaller. The favorite longshot bias in gravel tournaments should be stronger, because the chance of a ‘surprising’ win of an underdog is lower at this surface. A reason for the gap in level can be that there are more player who love to play on hard-court, than players who prefer gravel. This means there is more competition in the top favorites players on hard-court. However, it is also possible that it is harder to become very good at playing on gravel and it is easier to become a favorite on hard-court. The reason for this is not very important for the hypothesis.
The first subject which is discussed in this section is the dataset. Then the methodology is explained by working out the statical research and statical tests which are used for the results of this thesis.
This study will contain data of all ATP and WTA tournaments over the past 5 years (2009-2013). In Appendix A and B an overview of all those tournaments is attached. For every tournament the official tournament name, country, surface and number of matches is included. There are in total 18.610 matches included in this study. These are only single matches, so every match contains two players. Those two players each got an odd, so this study contains 37.220 odds. That is a lot if you compare this to other studies; Forrest & Mchale (2007) 8.500 matches/17.000 odds, Cain et al (2003) 91 matches/182 odds. Only the official tournament matches are included, there are no qualification matches in this study. The data is collected from: http://www.oddsportal.com/results/#tennis. The data will be tested for four different subjects namely; totally, men vs women, the surface and the trend over the years. In the table below, an overview of those categories is given. Table 5: Data distribution per hypotheses subject
| ||||||||||||||||||||||||||||||||||||||