Management by Baseball

In the last entry, I discussed allegations of Randy Johnson's current immaturity and I cited his 1998 behavior. The Bugged Unit's contract was in its final year, and by the time the season started, he knew his current team, the Seattle Mariners, weren't going to make a market offer. He had an extraordinarily poor overall half-season with the Ms, got traded at the end of July to Houston, and had as good a two-month run as any pitcher in recent history has achieved.

There has been some debate in sabermetric circles about Johnson's 1998 pre-trade record, and fortunately Repoz over at Baseball Primer reminded me, whacked me upside the head actually, with the specifics of Johnson's game-by-game data.

THE AVERAGE IS NOT THE GUY

There's a key lesson about both employee evaluation & the application of metrics here. If you just look at Randy's record and split it between the four months as an M and the two months as an Astro, it looks unarguable that The Unit could have been sandbagging:

Team	IP	W	L	ERA
Seattle	160	9	10	4.33
Houston	84	10	1	1.28

The reality is a lot more complicated. For one thing, Seattle's home park was pretty neutral that year, favoring neither batters nor pitchers, while Houston's depressed offense by about 4%. And items like wins and losses are very team-specific (someone who's performing at B+ pitching for the Expos this year is likely to have a far worse won-lost record than someone pitching D+ for the Yankees). But that ERA difference is something one can't argue away, right? The argument he was sandbagging to guarantee a trade and then showed his real talent in Houston is supported by the numbers.

Not for sure.

Because Johnson's game-by game performances indicate he had three different 1998 seasons for Seattle. Here are his game lines (excuse the imperfect table alignment...bad tool I gotta replace) courtesy of Retrosheet.

..#. Opponent .IP BFP H BB K HR R W L ERA
3-31- VS CLE A 5.2 29 11 2. 7 1 6 0 0 7.94
4- 5- VS BOS A 6 ..28. 6 4 10 1 7 0 1 9.26
4-10- AT BOS A 8 ..32. 2 3 15 1 2 0 0 6.41
4-15- AT CLE A 2.1 .9 .1 1 .4 1 1 0 0 6.14
4-20- VS KC A .3.1 20. 8 3. 3 1 6 0 0 7.46
4-28- AT KC A .7 ..30. 5 4. 9 0 1 1 0 6.12
5- 3- VS DET A 7 ..34 .8 5 11 0 6 1 0 6.41

5- 8- VS TOR A 6 ..26 .7 3 .7 1 3 1 0 6.15
5-14- AT CHI A 7 ..30 .7 2 .7 2 5 0 1 6.19
5-19- AT TEX A 3 ..16 .5 1 .2 1 6 0 1 6.83
5-24- VS TB .A 9.. 36. 7 2 15 0 1 1 0 6.02
5-29- AT TB A .8 ..29. 3 2 10 0 2 1 0 5.47
6- 3- VS ANA A 6.2 32. 9 3 .5 1 6 0 1 5.58
6- 8- AT SF N .7 ..30. 6 3 10 1 4 0 1 5.34
6-13- AT OAK A 7 ..33. 8 6 .7 1 7 0 1 5.61

6-19- VS OAK A 8 ..35 .8 2 12 0 1 1 0 5.17
6-24- AT SD N .9 ..35. 6 0 12 1 1 1 0 4.83
6-30- VS COL N 8 ..35 10 1 12 1 6 0 1 4.81
7- 5- AT TEX A 8 ..35 .9 2 12 3 8 0 1 5.07
7-11- VS ANA A 9 ..34 .5 2 15 0 0 1 0 4.73
7-16- VS MIN A 9 ..31. 1 3 11 0 0 1 0 4.44
7-22- AT TB A .7 ..30 10 0 .5 0 7 0 1 4.35

We can divide his performance in three sections. A clearly awful first seven starts, an erratic and inconsistent eight game slalom, an eight game conclusion leading up to the tradewhere he was his dominating Mr. Snappy-on-Speed self.

The average of Johnson's 1998 Seattle line, what the summary says, is he lost roughly as much as he won, and w/a roughly league average ERA -- that after some years of transcendant brilliance, he became league average. So the argument, if you use the average, is that he was dogging it. The individual numbers look much different and give you fodder to examine it with an alternate viewpoint. If you graphed the quality of his starts into the three chunks, and made the Houston run a fourth chunk, you could argue the trajectory makes sense. It'd be the same if you created a four-game rolling average. You could argue he started cold and just picked up steam all season long.

The counter-argument is, The Bugged Unit went into the season knowing his agent hadn't received a market-level offer from his team, he knew he would be gone next year, a situation that hits many players with small or no effect. But for him, the sting of not being wanted or perhaps not getting what he wanted made him act out or undermined his ability to perform during the first seven starts. After a while, he internalized his situation, and while it upset him, he was resolving his situation in his mind and was able on some days to show his Randy-ness. Finally he realized the team would trade him and his odds of going to a good team with a chance to make the playoffs would improve if his performance did, so he locked in and had that fine eight game exit run. That counter-argument is a long way to suggest he allowed his immaturity to interfere with his work performance.

Either way, the average is not the guy. The summary doesn't fully describe the reality, it's merely a frequently-useful artifact of it.

BEYOND BASEBALL

In your own personnel evaluation, you may or may not use hard measures. It's better when you can, but even when you do, it's important to remember that numbers have context that affect what they mean (like the home stadium differences tending to help Johnson's numbers after he moved to Houston). And summaries of those numbers may mask internal trends and streaks that, if you paid attention to them, would affect the amount of reality you were able to observe.

And when people have low self-awareness (Third Base in the MBB Model), it can cause them to tank at work, intentionally or unintentionally. There are lots of baseball examples of this (I'll save those for another time). Managing someone in that zone is a major challenge.

But there is a valuable & easy thing to do: Remember to pick apart performance measures and don't let the summary or average or highest or lowest point dictate what you can understand about the numbers. Keep looking for the reality in the artifacts. And thanks, Repoz, for that great reminder.