Last week I discussed how bad the Sox decline in winning percentage was in the second half of the season. This week I take a quick look at what could have caused it. That might involve how well they played in each half. One way to estimate what a team's winning percentage should be is based on their OPS differential (OPS is on-base percentage plus slugging percentage). Maybe the Sox really did play worse in the second half by this measure. If so, that would account for the fall in pct. Was their change in their OPS differential enough to explain the collapse?
First, last week I looked at 2nd half pct vs. 1st half pct. But here, since I used data from ESPN, I use pre-All Star and post-All Star stats. So the first "half" actually covers 88 games.
Second, I used the OPS differential since I showed in earlier research that it does a decent job of explaining winning percentage. The relationship I found was
PCT = 1.21*OPSDIFF + .5
Basically, since OPS includes the ability to get on base (OBP) and the ability to hit for power (SLG), the OPS differential tells us how much better a team was at these crucial tasks than their opponents.
The Sox had a .073 OPSDIFF in the first "half" and a -.001 in the second. So the change was a -.074. That is not the worst decline in the last 5 years, but it is pretty bad (actually the 15th worst, as you can see in the table below). So it is in the bottom 10%.
The formula predicts that the Sox would have a .588 pct in the first "half" and .499 in the second. Their actual numbers were .648 and .446, respectively. So the Sox were lucky to have such a good percentage in pre-All Star action while they were a little unlucky afterwards. For the season, the formula predicts that the Sox should have won 88.84 games, almost exactly what they did win. Maybe the good and bad luck evened out. Going just by the pct predicted by OPSDIFF, the Sox drop off was just .089. They did play worse in the second half, but maybe things did not decline as much as at first glance.
The table below shows all the changes in OPSDIFF over the last 5 years.