14.07.2013 Views

Globalisation and Efficiency in the Fixed-odds Soccer Betting ... - Istituti

Globalisation and Efficiency in the Fixed-odds Soccer Betting ... - Istituti

Globalisation and Efficiency in the Fixed-odds Soccer Betting ... - Istituti

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>and</strong>om process). Estimat<strong>in</strong>g <strong>the</strong> regression equation once only would <strong>the</strong>refore be mislead<strong>in</strong>g<br />

because of <strong>the</strong> unreliability of <strong>the</strong> st<strong>and</strong>ard errors.<br />

Accord<strong>in</strong>gly, we estimate each equation twenty times us<strong>in</strong>g twenty different<br />

r<strong>and</strong>omisations. The statistical significance of <strong>the</strong> coefficients can <strong>the</strong>n be exam<strong>in</strong>ed us<strong>in</strong>g a<br />

procedure outl<strong>in</strong>ed <strong>in</strong> Snedecor <strong>and</strong> Cochran (1967). We count <strong>the</strong> number of cases <strong>in</strong> <strong>the</strong><br />

twenty trials where a particular coefficient is significant at <strong>the</strong> 5% level. A normal<br />

approximation can be used to test <strong>the</strong> null hypo<strong>the</strong>sis that <strong>the</strong> “true” proportion of cases where<br />

<strong>the</strong> coefficient is not equal to zero is 5%. If <strong>the</strong> null is true, <strong>the</strong> observed proportion of<br />

rejections, R1 is distributed approximately normally with mean r <strong>and</strong> st<strong>and</strong>ard deviation<br />

s = (r (1-r)/n) ½ , where n is <strong>the</strong> sample size, here 20. The normal deviate, with a correction for<br />

cont<strong>in</strong>uity, is z = (|R – r| - (2n) -1 )/s. The critical value for this test statistic is 2.33 at a<br />

conservative 1% significance level. If <strong>the</strong>re are four significant (at 5%) coefficients out of<br />

twenty <strong>in</strong> our trials, <strong>the</strong> value of z is 2.57 which exceeds our 1% critical value, so where <strong>the</strong>re are<br />

four or more significant coefficients amongst twenty trials, we conclude that <strong>the</strong> particular<br />

coefficient is significantly different from zero (or one as appropriate).<br />

6. Results<br />

Table 1A reports our results for <strong>the</strong> whole sample <strong>in</strong> each of our four seasons. The values shown<br />

are mean coefficient estimates across twenty trials. The figures <strong>in</strong> paren<strong>the</strong>ses <strong>in</strong>dicate <strong>in</strong> how<br />

many trials <strong>the</strong> particular coefficient estimate was significantly different from <strong>the</strong> value specified<br />

by <strong>the</strong> null (zero or one). Where this number is four or more, we can reject <strong>the</strong> null hypo<strong>the</strong>sis.<br />

19

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!