Friday, Feb. 1, 16:28 EDT
So Arvind Parthasarathi gets the prize for the coolest application of Sherlock mentioned at our new machine’s launch event (note: the video starts a little after 28:20 on my computer, so you may need to scroll to get it to start).
The president of YarcData, which built Sherlock, Arvind noted that the batting and pitching records of major league baseball are just a big ol’ graph problem. We have data on pitchers’ performance throwing fastballs, curves, and for those few weirdoes, knuckle balls — and how often they’re hit, and by whom. You could even work in data about individual pitchers’ and batters’ performance as the game progresses.
Normally it would be like picking through so much cooked spaghetti to optimize all these relationships. Calculable, but not all that quickly. But thanks to Sherlock’s specialized architecture, which can investigate numerous possibilities in parallel, we can do this in real time. Think: which pitcher should I start given the opponent’s opening lineup? Should I relieve him now, or wait for a different part of their rotation? With whom?
Pirates, we are looking for projects to work on Sherlock. You may want to think about making use of every local resource available. I’m just sayin’, is all …