What was really fun, when I got older, was making baseball cards of myself and all my teammates. Using the APBA method that uses 2 dice and gives 36 equally likely outcomes, I went through our game score books, compiled detailed stat sheets, and then correlated every plate appearance to an outcome. When a player had fewer than 36 PA, I'd put a dash next to that dice roll, meaning the dice needed to be re-rolled. So in my game, if a kid had 4 singles, a double, and 6 strikeouts in 25 plate appearances, I c ould replicate those odds exactly.
When players had more than 36 plate appearances, I divided the stats by 2, and when there was an odd number, I put 2 outcomes with a "/" which meant another roll was required - a single die and 1,2,3 meant the first result, 4,5,6 the second.
So it looked like:
21: fly out
22: strikeout
23: single
24: double/ground out
25: walk
26: strikeout
31: ---
And so on. It was neat because it was a precise statistical representation of the batting stats. However, pitcher strength wasn't factored in, or defense, and I don't think stolen bases had a very sophisticated mechanism.
I contacted a couple coaches of other teams to see if they had well-maintained scorebooks, but none did. That would've been cool, to have cards of all the kids in the whole league.
This concept stayed in the back of my mind and later, in graduate school (software engineering) I played with the idea of developing such a program to basically determine optimal batting lineups. In other words, run 1,000 simulated nine-inning "games", rotate the lineup to the next configuration, run a thousand games with that lineup, and so on. Kick the program off before going to bed and in the morning see which lineup produced the highest number of runs.
I pitched (that's a pun) the idea to the baseball coach at the school, and right away he said: "Some of my guys get hot, some can't hit left-handed pitching, some can't hit the good fastball, some can't touch a curve... so I have to weigh factors like that into setting my lineup on any given day."
And... I realized he was right. A manager has to set his lineup based on the people on his team, not just their stat lines.
Also, with college teams in Minnesota, the dataset is just too small to be reliable. At the major league level, when guys get 500 at bats, it's a different story.
Anyway, statistics-based simulated baseball is a fascinating topic, and really the only way that Rod Carew can take some cuts against Grover Alexander...
|