1901 APBA: New Board Strikeout Ratings

As an addendum to an earlier post, I understand that some folks wouldn’t be terribly keen on changing pitcher grades.

This is particularly true for those who think that pitchers can “pitch to the score” (despite evidence to the contrary) and therefore the pitcher win statistic is somehow meaningful.

Those folks believe a 20-10, 4.50 ERA pitcher who throws for a team that scores 5 runs per game deserves an A while a 15-10, 2.50 pitcher who throws for a team that scores 2 runs per game doesn’t.

Obviously you can figure out where I stand on that.

But I digress.

At the simplest, I’d recommend making the following changes to strikeout ratings for the 1901 APBA cards when bringing them over to the newest board version:

  • National League
    • Chesbro – add (Y)
    • Dinneen – add (Y)
    • T. Hughes – replace (Y) with (X)
    • Leever – add (Y)
    • Pittinger – add (Y)
    • Tannehill – add (Y)
    • Townsend – add (Y)
    • Waddell – replace (Y) with (X)
    • Willis – add (Y)
  • American League
    • Garvin – add (Y)
    • Patten – add (Y)
    • Young – add (Y)

This is to account for the fact that when the 1901 cards were issued, A pitchers earned about 1 strikeout per game because of PRN 9 rolls with the Bases Empty.

That’s no longer the case with the new boards and we need to find those strikeouts somewhere.

An A is roughly equivalent to a (Y) and an A(Y) is roughly equivalent to an (X).

Among pitchers who saw significant playing time, the cards as issued had 2 A(Y), 6 A without a (Y), and 4 non-A (Y) pitchers in the NL. That boiled down to needing to find a place to assign 2 (X) and 10 (Y) ratings.

Keeping their (Y) were W. Donovan, Hahn, Mathewson and White.

You may notice one additional (Y) in there and that’s Townsend, who had decent strikeout rates but is the #5 starter for the Phillies. He’s not likely to get a whole lot of appearances and throwing him a (Y) seems acceptable here.

Over in the AL, there was 1 A&C and 2 A pitchers with no (Y) ratings. So we basically needed to find place for 3 (Y) ratings.

That was easy enough as we distributed them out to the 3 pitchers with the best strikeout rates – Garvin, Young and Patten.

Advertisements

1901 APBA Baseball: Re-Grade

I’ve dove back into a replay of the 1901 season using APBA Baseball.

I have a long, horrible history with this project.

I tried to do it once and quickly found something was amiss with the cards.

For example, of the 32 pitchers who are members of the 4-man rotations in the National League, 29 have a (Z) and 1 has a (W).

How can 29 of them have a (Z)?

Only 6 of the 32 NL starting pitchers have a (Y), while none of the 32 American League pitchers have a strikeout rating.

I entered every result of every card into a spreadsheet, weighting them by their plate appearances that season and I couldn’t find any way that strikeout totals would come anywhere close to historically accurate.

So I set off on an adventure of re-grading and re-rating every card from that set based on an elaborate set of equations using the card data.

However, it was only recently that a thought occurred to me.

I’d been trying to play these cards with the latest set of boards.

The 1901 set was issued in 1988, before things like (ZZ), (K) or (R) ratings had been issued.

And, more to the point, when this set was published, the board included Strikeout readings for some pitchers on those hits that get taken away due to being an A or B pitcher.

Oh dear…

So, for example, my concern about a guy like Christy Mathewson not being able to come anywhere near his real-life strikeout rates with only a (Y) on his card were somewhat unfounded. After all, by my math, that’s good for maybe one extra strikeout per game.

But giving him a Strikeout with the Bases Empty when a batter rolls up a Play Result Number of 9? That’s good for another additional strikeout per game.

So now I realize that him being an A with a (Y) was actually good for about 2 extra strikeouts per game over a pitcher who was neither an A nor a (Y).

Given the National League average that year was 3.8 strikeouts per 9 innings and Matty whiffed 5.9 per 9 innings – a difference of 2.1 – that suddenly seemed pretty valid.

In short, I started replaying the season by “retro-fitting” my current boards so that they more closely resemble the boards at the time that the 1901 card set was first released.

(You can find PDF files online that document the history of APBA board changes to accomplish this.)

And, as low-fidelity as APBA can be, I’ve been enjoying it so far. I’ve rolled about 35 games and it’s been fun.

However, there are a few things that have still been bothering me.

Guys like Rube Waddell (6.2 SO/9) and Tom Hughes (6.6 SO/9) had higher strikeout ratings than Mathewson that season, but they are both B(Y)(Z) pitchers.

As B pitchers, they don’t get the benefit of strikeouts on a PRN of 9 with the Bases Empty, so even though their strikeout rates were noticeably higher that that of Mathewson, they will actually strikeout fewer batters than him.

My obsessive compulsive tendencies over these games started to kick in.

As I have been rolling these games and comparing my sim rates to historical rates, they’re really not all that far off.

So I know that as a whole there are probably a fairly correct number of (Y), (Z) and (W) ratings out there.

But where I’d still find a point of contention is how they are distributed.

So I got to work.

I plugged every carded pitcher into a spreadsheet, calculating how many total batters faced are given to each grade and each rating.

For the purpose of bringing these ratings over to the modern board, an A pitcher was the equivalent of a (Y). An A(Y) was the equivalent of an (X).

Doing this, I came up with the National League having 2,851 Batters Faced allocated to an (X) – this was contributed by Mathewson and “Wild Bill” Donovan, each of whom are A(Y) pitchers in the original set. 11,784 Batters Faced belong to (Y) pitchers.

Then I re-distributed everything.

By a point of example, I sorted all carded pitchers by strikeout rate (strikeouts per batters faced). I started giving out (X) ratings until I had given them out to 2,851 totals Batters Faced. Then I gave out (Y) ratings until I had given out some type of rating to a total of 14,635 Batters Faced – this is 2,851 plus 11,784 from above.

So what I did was just try and ensure that the chances of a plate appearance being affected by a pitcher grade, strikeout rating or control rating were no different than before.

All I did was move things around.

One other note about this – the re-distribution didn’t penalize pitchers with small sample sizes. Much like the newly issued sets of cards from the APBA Game Company, these formulas are assuming that you will use pitchers close to their historical usage. So a pitcher could get an A even if he only made one appearance. It’s expected that the re-player won’t cheat the system and use this guy as a regular member of the rotation.

Okay, that’s the backdrop. What changed?

Well, surprisingly little. I’ll call some things out alphabetically as I go through the list.

Name Old New Notes
Jack Chesbro A(Z) A(Y)(Z) A touch above league average in K/9
Roger Denzer D(Z) B(Z) Close to league average ERA in 62 IP
Bill Dinneen B(Z) B(Y)(Z) A touch above league average in K/9
Ed Doheny C(Z) B(Z) Close to league average ERA in 150 IP
Jack Harper B(Z) C(Z) ERA was 29th out of 34 qualifying pitchers
Tom Hughes B(Y)(Z) B(X)(Z) Led league with 6.6 SO/9
Brickyard Kennedy D(Z) B(Z) Nice 110 ERA+ in 85 IP
Bob Lawson D(W) B(W) 110 ERA+ in 46 IP
Sam Leever C(Z) B(Y)(Z) Finished 10th in ERA and 9th in SO/9 but had just 20 GS
Gene McCann D B Right around league average ERA but just 34 IP
Mike O’Neill D(Z) A(Z) 1.32 ERA in 41 IP
Togie Pittinger B(Z) B(Y)(Z) A touch above league average in K/9
Ed Poole D C Only 80 IP, but really only slightly below average in ERA
Jack Powell B(Z) D(Z) ERA was 26th out of 34 qualifying pitchers
Jesse Tannehill A(Z) A(Y)(Z) A touch above league average in K/9
Happy Townsend C C(Y) A touch above league average in K/9
George Van Haltren D(W) A(W) 3.00 ERA in 6 IP
Rube Waddell B(Y)(Z) B(X)(Z) 6.2 SO/9 was 2nd best in NL
Vic Willis A(Z) A(Y)(Z) Above league-average strikeout rate

Now, how about the AL, where we originally had 1 A&C, 2 A and no strikeout ratings?

Remember from above that I am converting this to the equivalent of 3 full-time (Y) pitchers.

Name Old New Notes
Bill Bernhard C(Z) D(Z) 17 Wins but ERA was 32nd out of 34 qualifying pitchers
Pete Dowling C D ERA was 28th out of 34 qualifying pitchers. On the cusp here.
Jack Dunn D B ERA was just better league average but only 60 IP
Chick Fraser B C Another downgrade for Philly. ERA worse than league average.
Ned Garvin B B(Y) League-best strikeout rate deserves one of the (Y) ratings
Clark Griffith A(Z) B(Z) 4th-best ERA in the league, but if we’re only giving out 3 As or better…
Bill Hart D C 3.77 ERA was only 0.08 worse than league average.
Zaza Harvey D B ERA was a little better than league average in 92 IP
Ed High D(Z) B(Z) 3.50 ERA is better than average. 18 IP.
Watty Lee C(Z) D(Z) ERA was 31st out of 34 qualifying pitchers
Ted Lewis B(Z) C(Z) ERA a touch better than league average, but no world-beater
Jack McAleese D D(Z) Walked 1 of 17 batters he faced.
Harry McNeal D D(Z) BB rate in line with other full-time pitchers who have a Z
Fred Mitchell D C In 109 IP, his ERA was barely worse than league average
Jerry Nops D(Z) C(Z) ERA was 26th out of 34 qualifying pitchers
Frank Owen D C ERA not great, but not bad enough for a D
Casey Patten C C(Y)(Z) 3rd best strikeout rate in the AL deserves one of the Y ratings
Wiley Piatt D C ERA was 27th out of 34 qualifying pitchers
George Prentiss D A(W) 1.80 ERA and 5.4 BB/9 in just 10 IP
Bill Reidy C(Z) D(Z) ERA was 29th out of 34 qualifying pitchers
Crazy Schmit D(W) A(W) 1.99 ERA in just 23 IP
John Skopec D(W) B(W) 111 ERA+ in 68 IP
Tully Sparks C B ERA was 13th out of 34 qualifying pitchers
Snake Wiltse C(Z) B(Z) ERA just a touch better than league average
Joe Yeager B(Z) A(Z) 3rd best ERA in the league makes him deserving of one of the 2 A grades
Cy Young A&C(Z) A&C(Y)(Z) Led league in ERA and was a very close 2nd in strikeout rate

So I hope some folks find that interesting.

Honestly, I’m this far into my replay already that I will probably just stick with things as they are.

But as I get into this a little further, I’ll be interested to see a few things in particular from the top 4 pitchers on each staff:

  • Tom Hughes and Rube Waddell should easily simulate to be the top 2 strikeout rates in the National League. But with B(Y) readings as-carded, will they even make the top 5?
  • How badly will Jack Harper and Jack Powell over-perform? Because they sure don’t seem to warrant B grades.
  • How badly will Sam Leever under-perform?
  • Will the top 3 strikeout leaders in the American League resemble real life or will the 3 A pitchers dominate it despite two of them ranking 14th and 16th that season?

The Trouble With Game Winning Drive

A couple of months ago I saw a YouTube video on Downey Games’ Game Winning Drive.

GWD is what you might call a quick simulation game.  In about 10 minutes, you can roll up an entire football game, ending up with a final score and certain basic team statistics (rushing touchdowns, passing touchdowns, field goals, interceptions and fumbles lost).

It doesn’t do individual statistics, although it’s not too difficult to rig something up if you want to know who, for example, scored a particular touchdown.

That’s not really the point of the game engine, of course.  The point is to allow you to roll up an entire week’s worth of NFL games in 3-4 hours and therefore make it plausible to simulate an entire season over the course of a few weeks or month.

I was taken with the idea and found it pretty interesting, so I went ahead and picked up a copy.  An e-book version of the game is only $10 for a season, which isn’t shabby.  For $15 you can get a printed copy delivered to your door front along with the four 6-sided dice required to play the game. I have enough dice lying around, so I didn’t really need that.

I got to rolling games from the season I purchased (1985) and about 8 games in started noticing something peculiar. My games were, in general, running pretty high in the scoring department.

The historical season averaged 21.5 points per team-game and I was just a hair over 24.

That’s not colossally larger, but it’s noticeable.

As I do with pretty much every card & dice game I’ve ever purchased, I started to reverse engineer the game and try to figure out what the hell might be going on.

At this point, I’ll need to offer a quick breakdown of the game.

A game is broken down into 20 possessions.  So, generally speaking, each team gets 10 per game.

For each possession, you roll four dice.  Two of the dice are used to determine whether a team records a Score or a Turnover.  The other two dice are then used to break down either that Score (Run TD, Pass TD or Field Goal) or Turnover (Fumble Lost, Interception, Punt or Missed Field Goal).

Pretty simple.

I started taking a guess at how they might come up with the range of rolls required for the Score rating for each team and went through things.

Example #1: Atlanta

In 1985, they scored 14 rushing touchdowns, 13 passing touchdowns and 24 field goals.  So my math figured the following: 16 games multiplied by 10 possessions per game equals 160 total possessions for the season.  14 rushing touchdowns plus 13 passing touchdowns plus 24 field goals equals 51 scores in those 160 possessions. 51 divided by 160 is 0.31875.  Multiply that by the 36 combinations you get from rolling a pair of 6-sided dice and you get 11.475, so you might guess that their range for Score is from 11 to 25. And, in fact, that’s what the official season book reads. Eureka!

Just guessing a little more, I’m looking at 14 rushing touchdowns divided by the 51 total scores for a total of (roughly) 0.2745, multiplying that again by 36 to get to 9.88 and guessing that the “TD Run” listing within score will list 11-24. What do you know? It does! And, similarly, 13 passing touchdowns divided by 51 total scores, multiplying by 36 to get to 9.18 and I’m guessing there will be 9 total “TD Pass” listings on their card. Again, there was.

I repeated this process with 3 other teams just to take a guess at how they were doing things and every time came up correct. So I think we’ve got that.

So why are game scores running high?  That seems correct.

Here’s the problem.  It’s something I haven’t pointed out about the rules yet.

If a team recovers a fumble or intercepts the opponent, they get a +6 bonus towards their Score range for the ensuing roll.  In other words, instead of Atlanta needing a roll of 11-25 to score, they instead need a roll of 11-35.

Atlanta’s defense had 34 turnovers in 1985 – more than 2 per game.

So, in an average game of GWD, Atlanta will have 8 possessions where they score on an 11-25 and 2 where they score on an 11-35.  Instead of averaging 11/36 on their chance to Score per possession (as they should), they instead average 12.2/36, an increase of 11%.

(Not coincidentally, I’m also running about 12% over right now…)

So while the game engine itself is pretty solid for what it’s trying to accomplish, there is a flaw in the way the charts are put together.  They don’t factor in the “+6” bonus when coming off of a turnover.

If, in Atlanta’s example, we change their Score range from 11-25 to 11-24, they now have 8 possessions per game with 10/36 chance of scoring and 2 with a 16/36 chance of scoring, that comes out to an average of 11.2/36, which is more what we want.

I went ahead and plugged everything into a spreadsheet and verified that my guess at the Score reading was correct in 100% of the cases.

For most teams, it turned out where you’d have to adjust their Score range down just 1 chance.  For example, instead of 11-25 they should be 11-24. Instead of 11-31 they should be 11-26.

Some teams that had an extraordinary number of turnovers, however, like that vaunted Bears’ defense, should be adjusted from an 11-35 to an 11-33.  They’re going to get over 3 scoring chances off of turnovers per game.

All told from the 1985 season, 20 teams were adjusted down 1 chance and 8 were adjusted down 2 chances.

If folks are interested in this kind of work, I’m more than happy to post the spreadsheet so you can do the same with re-calculating other seasons.

It’s a neat game engine and I rather like the game itself for what it is.

It just has a few things that it didn’t consider.

Time to zero out my scoreboard and standings and start all over again.