There are limitations to the study, discussed within it, but it doesn't seem to me the the data would be as it is if the effect were not real and rather large.
Article about it: http://www.economist.com/blogs/gametheory/2011/09/statistical-research-baseball
Original article, with methodological detail: http://www.economist.com/blogs/gametheory/2011/09/statistical-research-baseball
Article about it: http://www.economist.com/blogs/gametheory/2011/09/statistical-research-baseball
Original article, with methodological detail: http://www.economist.com/blogs/gametheory/2011/09/statistical-research-baseball