[Pacoheadley]

Originally Posted by Matt Arnold

There's different ways to calculate how correct we are, whether you look at IFH/GB or IFH/SI. Whenever I run the numbers, we tend to be a hare low if anything.

I think it's the sort of thing that you do tend to notice more often than not when playing out a game. Partly because our animations of infield hits are... not great. I think the 3 most common plays you see are slow rolling balls that the infielder has no real chance at, balls that are hit a little harder, that the infielder might take an extra fraction of a second to get to, and the runner just beats out at the base, or hard hit balls that get classified as infield hits because the fielder cuts it down, but if they didn't, it would get through them for more.

And in all 3 cases, those are like the 3 worst plays we handle in the 3d. The first case, it's really hard to get balls to get that sort of slow bounce that looks realistic. The second one is a timing issue, so trying to line up the ball getting to the fielder plus the fielder's throw to arrive just a hare slower than the batter getting to the base is just extra tricky. And the 3rd case we've only really "added" somewhat more recently, tend to kind of look like an error for us that gets called a hit.

They're all always on my list to improve, but because of how fast the plays run, they are the hardest ones to look correct. But whenever we run our internal tests, the numbers come out as expected. I have some test data, and it comes out to around 1650-1850 infield singles out of about 13000-14000 singles total, which is pretty close to MLB numbers which are 1746 infield hits in 12668 total singles this year (and has consistently been in around that range for the last few years now).

Thanks a ton for the transparency with this type of thing.