Chris Witt’s adjusted pitching grades for 1901

For a good while now, there’s been a flurry of comments on one of Scott Fennessy’s columns and a good APBA/deadball discussion is brewing.  There was talk of season compatibility and even adjustments to pitchers’ ratings to make them more accurate and realistic. 

Chris Witt provides a tangible outcome from that discussion.  He has passed on his suggested pitchers’ ratings for the 1901 season. 

I’ll let Chris describe it for himself:

In brief…  I took out a new set of 1906 cards (copyright date was 2012 on them) and tried to "crack the code".

I got a spreadsheet of all pitching stats and started trying to figure how APBA was coming up with their pitching grades.

In short, I compared strikeout rates for each pitcher against their league’s strikeout rates and, sure enough, when you sorted them from top to bottom you saw the pitchers with an XY, followed by X, followed by Y, followed by nothing.

Also straight-forward was the control rates.  I compared walk rates for each pitcher against their league’s walk rates and sorting them from best to worst got me ZZ, then Z, then nothing, then W.

For both the 1906 AL and 1906 NL, I was able to find certain "cut-off" points that were fairly consistent for when the game company decided to award each grade.

The grades were a little trickier.  I took a guess that it based purely on hit rates allowed, but that didn’t hold up.  With the basic game in particular, you need that grade to help get the pitcher close to their overall ERA.  Some pitcher might give up fewer hits but have a higher ERA because they allowed more HR, for example.  So that alone didn’t do it.  The correlation between hit rate allowed and pitcher grade was about 0.8.

I tried to see if there was a direct correlation between ERA and grade.  Again, there wasn’t.  Little stronger though.  The correlation was about 0.8.

Lastly, I looked to see if there was a connection between Wins and grade, knowing that’s how the company used to do things.  But that didn’t turn out directly related, either.  The correlation was about 0.6.

So what did I do?  Basically smashed the 3 things together and came up with a formula.  What I found was not perfect, but I did find that I was shooting 99% on guessing when they’d award an A&C, 95% on when they’d award an A, 90% on when they’d award a B.  Cs and Ds were a bit trickier, but I was right on about 75% of them and within one letter grade on 100%, so let’s call that "close enough"!

So these are definitely not official, but they are using formulas that got me pretty darned close to what they seem to be using for another recently issued Deadball set of cards.

I have a database query I can run that generates all rating for any pitcher from 1871 to the present.  But I really wanted to just open this up more for discussion.  Looking these over, I don’t see anything that doesn’t make me think "ya, that makes more sense than what the official 1901 set says."  There are some ratings that are on that set which make absolutely no sense, but as others have said, it looks like APBA used to have set numbers that a pitcher had to hit to earn a certain letter rather than basing them on a pitcher’s numbers relative to the rest of their league.

Hoping this makes a nice little conversation piece if nothing else.  Or maybe a nice add-on for those who might want to play with older sets and get (hopefully) more realistic numbers.  I rolled way too many games with that 1901 set that were resulting in maybe 1.5 K/9 being thrown per game when the numbers should have been more than twice that high.

If anybody likes these, I’m totally willing to open up my code for all the APBA community to have a look at.

Take care and great job on the site!

Here are Chris’ suggested grades for the 1901 season.  They include all pitchers with 25 or more innings pitched. 

For those who are looking for a more accurate experience in your replay, this may help.  If you’re just looking to have fun… well, let’s face it, people don’t replay the 1901 season “just to have fun”.  :)

Thanks Chris!

[photo credit]