[BostonCard]

Since 2011, Stanford is 9-3 (.750) after a bye versus 70-19 (.789) overall. Against the spread during the same time frame, Stanford is 7-5 (.583) versus 54-34 overall (.614). Note that Stanford is the second-best team against the spread overall since 2011 versus somewhere in the 30s against the spread after the bye. Small sample size on the data-slicing, of course, but perhaps an indicator of underperformance.

No, it is not an indicator of anything.  The two records after a bye (straight up and against the spread) are perfectly compatible with being a subset of the overall records.  They are about as close to even as you could possibly make them (if you change 9-3 to 10-2, it's an .833 record, which is farther away from .789 than 9-3 is; if you change 7-5 to 8-4, it's a .666 recordm which is further from .614 than .583).  In fact, what is uncanny is how closely it hews to the overall record.

If you want to be statistically rigorous about it, the best record that he could have had that would have indicated a statistically significant (p < .05) underperformance in games after a bye week compared to games overall would have been 5-7 overall, 3-9 against the spread.  Which is to say that "small sample size on data-slicing" is the correct answer.

BC

[BostonCard]

Since 2011, Stanford is 9-3 (.750) after a bye versus 70-19 (.789) overall. Against the spread during the same time frame, Stanford is 7-5 (.583) versus 54-34 overall (.614). Note that Stanford is the second-best team against the spread overall since 2011 versus somewhere in the 30s against the spread after the bye. Small sample size on the data-slicing, of course, but perhaps an indicator of underperformance.

No, it is not an indicator of anything.  The two records after a bye (straight up and against the spread) are perfectly compatible with being a subset of the overall records.  They are about as close to even as you could possibly make them (if you change 9-3 to 10-2, it's an .833 record, which is farther away from .789 than 9-3 is; if you change 7-5 to 8-4, it's a .666 recordm which is further from .614 than .583).  In fact, what is uncanny is how closely it hews to the overall record.

If you want to be statistically rigorous about it, the best record that he could have had that would have indicated a statistically significant (p < .05) underperformance in games after a bye week compared to games overall would have been 5-7 overall, 3-9 against the spread.  Which is to say that "small sample size on data-slicing" is the correct answer.

BC