Baseball Prospectus is one of the pre-eminent sabermetric websites, with a long history stretching back over ten years. They have revealed some of the most influential statistics known to baseball fans, including Pythagenport, VORP, and WARP. However, what makes them special is the PECOTA projection system, a unique proprietary design by Nate Silver. While VORP and WARP have undergone gargantuan criticism, PECOTA is a favored baseball projection system.
PECOTA is empirically derived, and is not merely based upon statistics. It uses phonotypical characteristics of players, in order to make a system that is able to include a much larger set of data, and able to predict further into the future. This allows those farther away from the mean to be projected more accurately, which is always a great boon. Nate Silver has since moved onto political projections, utilizing the same principles that are in PECOTA.
However, Baseball Prospectus spurred much controversy through its headliner statistics, such as WARP and VORP. The concept of the replacement level is at the crux of the statistical analysis of baseball. It is the point where the talent distribution meets the number of major league spots. Essentially, a player that is at replacement level has tens of other comps that could take over for him without any change in talent level. For many years, Baseball Prospectus set its replacement level at an appallingly low level. It is generally accepted that replacement level is around 48 wins at a team level, while BP set it at around 35 wins. This led to comparisons between players being way off, because instead of eight wins to three wins, it is 11 wins to six wins. 1.83 instead of 2.6. When you are talking in a context where one win means millions of dollars, this is an immense error.
Ultimately, Baseball Prospectus has been at the forefront of sabermetrics, its stats are too wild, and that BP has to focus on joining the rest of the sabermetric front in a unified effort to have the best information.