Why do the number of outcomes recorded for a pitcher in a given year
(strikeouts, groundouts, flyouts, hits, walks, and hit batsmen) often
exceed the number of batters faced in that year?  For example, in 2004
Pedro Martinez recorded 227 strikeouts, 186 groundouts, 228 flyouts,
193 hits, 61 walks, and 16 hit batsmen, for a total of 911 outcomes. 
Yet he only faced 903 batters (plate appearances).  Why is that?

The only situations that I can think of that might make the "batters
faced" number differ from the total of the "outcomes" you mention
would result in a "batters faced" number larger that is larger that
the "outcomes" total, not smaller.  However, since neither involves a
"completed" plate appearance, they probably aren't counted as "batters
faced."

One of these is "catcher's interference," a situation where a batter
is awarded first base, the catcher is charged with an error and the
batter is not charged with an "at bat."

The other one is the situation where a batter is left stranded at the
plate at the end of a game when the final out is made by "picking off"
a base runner, for example.

Namer --

In the case of a strikeout where the ball is dropped or in the dirt,
the catcher has to tag the runner out (2 unassisted) or throw them out
(2-3) at first base.

I believe that this may account for the groundouts and strikeouts
being double-counted.  It would easily account for 8 or more outs for
Pedro.

Best regards,

Omnivorous-GA

Reached on error?  Sacrifice flies?  Caught stealing for third out of inning?

These plus the possibilities mentioned in prior comments, when added
and subtracted, could explain the difference between outcomes and
plate appearances.

Any sabermetricians out there?

I have emailed my favorite baseball columnist asking for his thoughts.
 I will let you know if and when I get a response.

Markj
Thank you for your thoughts.
I agree that catcher's interference would count as a plate appearance
but would not generate an outcome.  Therefore, we would expect plate
appearances to exceed outcomes.  I beleive the same is true when a
batter reaches first on an error.  It is a plate appearance, but not a
hit, out, or walk.  Therefore, I would expect plate appearances to
exceed outcomes by a significant number, to account for errors and
catcher's intereference.  But it is always the other way around.
When a batter is left stranded at the plate at the end of an inning,
this is not counted as a plate appearance or an outcome, so this is
probably not the source of the discrepanc

Justaskscott,
Thank you for your thoughts.  However:
Reached on error would cause plate appearances to exceed outcomes, not
the other way around (it would count as a plate appearance, but no
outcome would be recorded--no out, not hit, no walk)
The sac flies are included in flyouts.
Caught stealing would end a batter's tenure mid-appearance.  This
would not count as either a plate appearance or an outcome and would
not creat the discrepancy.

If a pitcher is removed during an at-bat, I believe that the removed
pitcher is charged with the outcome of that at-bat.  However, it would
be consistent with other rules about partial at-bats *not* to charge
the removed pitched with a "batter faced" in those circumstances, I
believe.  Do you think that this could explain the discrepancy?

Namer could be right about the effect of the occurrences I suggested,
though I can't say for sure.  I figured it was worth mentioning these
possibilities as food for thought.

As for markj's suggestion, I believe a pitcher is almost never removed
during an at-bat.  I'd be surprised if it happened during 2004 to
Pedro Martinez, unless he was injured in the middle of an at-bat.

I believe double plays are a major reason for this discrepacy as
mentioned in an earlier post already.
when a player grounds into a double play this counts twice in the
groundout statistic and in this way increases the "outcomes" of a
pitcher. other things might play a role as well and were mentioned
before, too (for instance, dropped foul ball errors).

in the referenced 2004 season stats of Pedro his 8 GIDP can account
for the mentioned difference between  911 outcomes and 903 batters
faced.

To illustrate this I suggest to check the 2005 stats of Pedro and then
analyse each of the 6 games he started so far (this is easier than
checking the whole 2004 season).

http://sports.espn.go.com/mlb/players/gamelog?statsId=4875

For some strange reason the ground out and fly out numbers are wrong
on this page, correct ones can be obtained at mlb.com (they are also
correct in each box score):

http://mlb.mlb.com/NASApp/mlb/mlb/stats/mlb_individual_player_gamebygamelog.jsp?playerID=118377&statType=2

The respective numbers for 2005 are:

52 strikeouts, 41 groundouts, 37 flyouts, 22 hits, 8 walks, and 0 hit
batsmen, for a total of 160 outcomes.
Yet he only faced 158 batters (plate appearances).

Checking each of the six games shows that outcomes and faced batters
are equal 4 times. They are not in the season opener (23 to 22) and in
the game on April 26 (30 to 29). You can count the outcomes in the
play-by-play version of the game:

http://sports.espn.go.com/mlb/gameLog?gameId=250404117

counting is easy here, mostly K's .. ;-)
2 players grounded out, one of them into a double play. The stats
count 3 groundouts.

The same is true for the April 26 game.

Two more outcomes than batters faced because of two double plays induced.

You guys ought to get out to the ballground.  Help that pallid teint
of computer jockeys.

There is a difference of 8, and 8 double plays:  8 men got on base
with hits and then were counted again as groundouts.

Omni was right with his now deleted answer, and should be back at bat by now. :-)

I only went to one pro game in my youth, but it was a biggy:  Larsen
of the Yanks pitching a no-hitter for nine innings against the Red Sox
in Boston, Ted Williams struck out and gesturing at the press box, for
which he got a walloping fine of $ 500 (?, just a raised fist, no
extended digit). After Larsen came off, the Yanks won in the 10th or
11th inning.

Just name the year ...

macchiato,
Counting each ground out into a double play as two groundouts for a
pitcher is a reasonable theory as to why outcomes exceed events. 
Let's see if it holds up beyond Pedro:

Consider Tim Wakefied's 2004 numbers: 116 strikeouts, 237groundouts,
205 flyouts,197 hits, 63 walks, and 16 hit batsmen, for a total of 834
outcomes. He faced 831 batters--two more outcomes than events.  And
there were 11 groundouts into double plays.

Or Curt Schilling's 2004 numbers:  203 strikeouts, 219groundouts, 251
flyouts, 206 hits, 35 walks, and 5 hit batsmen, for a total of 919
outcomes. He faced  910 batters--9 more outcomes than events.  And
there were 12 groundouts into double plays.

Does it still make sense

Also, if GDPs result in two outcomes for the doubled-up player,
shouldn't outcomes be double counted as well for the doubled-up player
in line-drive/pick-off (and other) double plays?

Namer --

You're assuming that a GIDP is counted as two groundballs.  I'm
reasonably certain that it DOES NOT.  Pitching coaches and others
track ratios of groundball outs and flyball outs discretely -- it's an
important measure in setting defenses.  If you have a sinkerball
pitcher like Jake Westbrook, you set infield and outfield defenses
differently than for those pitchers who elevate the ball and get more
flyball outs.

I stick by yesterday's answer -- now removed -- that the difference is
GIDP.  But now we need to find the evidence to support the argument.

Best regards,

Omnivorous-GA

markj --

it wouldn't make much sense to count other double plays like
line-drive/pick-off from the perspective that "outcomes" and
"groundouts" are pitcher statistics. while it is not really due to the
pitcher that a batter lines out and subsequently a runner gets doubled
up, it is quite a lot due to the pitcher when a batter grounds into a
DP.
therefore, it would make sense to count the times a pitcher induced a
GIDP (partly his effort), but not lined into DP or others (random or
not related to pitcher's effort).

macchiato -- 

You may well be right, but I'm not sure I agree with yhour logic.  It
seems to me that a pitcher has precisely as much of a role in
generating a line drive as he does in generating a ground ball.  And,
either way, the pitcher usually has no direct role in generating the
second out in the double play.

Namer --

sure, you are right, "outcomes" minus GIDP not necessarily equals "batters faced".
this is because there are other factors that make "outcomes" and
"batters faced" diverge as mentioned previously in the post.
in the aggregate data of a whole season or part of it this is pretty
difficult to control for.
if you analyse the data of one single game which you can "score" based
on play-by-play data and then compare to the statistics given in the
box score you can pin point this reasons better.
I have given an example of that with Pedro's 2005 games until now in
the earlier post. the data of one single game with one outcome more
than batters faced and one GIDP occurring is quite evident, I would
say.

macchiato--
Thank you very much for the espn link with the pitch-by-pitch records
and the suggestion to score those games and compare the results to the
official record.  That has been extremely useful.
I have now scored 14 recent games in this way and compared the result
to the official MLB records for the pitchers involved.  In all 14
cases, the official results are consistent with the following
methodology:
(1) GIDPs are scored as two groundouts.  They are the reason outcomes
exceed events (batters).
(2)  Batters who reach first on an error and catcher's interference
are not counted at all.  Apart fronm the GIDPs, this would cause fewer
outcomes than events.
(3)  A fielder's choice is scored as a groundout.
I will do another dozen games to be sure, but so far this method
creates a 100% match.
Having said that, I agree with the other comments that this is dumb. 
By this method, a pitcher can be credited with either one or two
groundouts for the exact same pitch, depending on whether or not there
is already someone on base and how the fielder's handle the ball
situation to get the double play.  This does not seem to be a good way
to measure pure pitching performance.  But then again, we had to
expect that any statistical method which regularly produces more
outcomes than events was going to have a pretty big flaw in it
somewhere.

You'd think someone really knew ...

Could it result from the pitcher coming in as relief in mid-inning
with men on base who were put out?

With all those numbers, could there just be some miscounting?

Myaorin,
I don't think so.  There were many instances of pitchers coming in
mid-inning, but thsi always squared with the batters faced.  And I
think there are far too many instances for this to be miscounting. 
Outcomes exceed batters for the majority of games--because there are
double plays in the majority of games.