Alternating Heat-And-Junk
Adam Coutts
to give runs away this World Series. As we've discussed, DHing Dunston and
leaving Minor off the roster is one such move. As we've discussed, letting
tight-hamstring Lofton play center and miss balls while wide-ranging Shinjo
DHs or sits on the bench is another.
But the construction of his pitching rotation has to be the most pointless
waste of resources this World Series. As we've discussed, by pitching Ortiz
two games on the road, the Giants miss the advantage his fine hitting. As
we've discussed, by starting Woody at home, the Giants miss the precedent
that he is 16-16 at home, and 23-13 on the road. As we've discussed, it
means that Woody is the guy who only starts once, rather than loss-leader
Livan.
Here is something new, however:
Worst of all, it turns out that there is a big cost to not alternating heat
and junk starters. Jamal pointed out that "Anaheim's bats were sped up just
a bit by seeing Schmidt the night before spelled curtains for good ole'
Russ.... ". When you've just seen Schmidt's 96-98 MPH fastballs, Ortiz'
93-94 MPH fastballs are batting practice. Woody's junk might have cooled
'em down in Game 2, however.
Turns out that, looking at the numbers, the last two years Ortiz HAS A
EXTENSIVE AND PROVEN RECORD OF GETTING SHELLED WHEN STARTING AGAINST THE
SAME TEAM RIGHT AFTER SCHMIDT'S HEAT. Baker could have looked it the fuck
up. The team has won two-thirds of the time, on the other hand, when Ortiz
has blazed in against the same team right after Woody's junk.
In fact, according to this analysis, the difference between the rotation as
Dusty set it up (Schmidt, Ortiz, Livan, Woody, repeat) and as some of us
thought it should be set up (Schmidt, Woody, Ortiz, Livan, repeat) is
IMMENSE.
Team
ERA W-L
------------------------------------------
4.38 9-5 0.643% Schmidt being first pitcher to face a new team
4.81 5-10 0.333% Ortiz following Schmidt against same team
4.47 6-9 0.400% Livan following Ortiz against same team
3.91 10-11 0.476% Woody following Livan against same team
4.25 6-4 0.600% Schmidt following Woody against same team
4.81 5-10 0.333% Ortiz following Schmidt against same team
4.47 6-9 0.400% Livan following Ortiz against same team
------------------------------------------
4.44 0.455%
------------------------------------------
4.38 9-5 0.643% Schmidt being first pitcher to face a new team
2.71 10-2 0.833% Woody following Schmidt against same team
4.01 16-8 0.667% Ortiz following Woody against same team
4.47 6-9 0.400% Livan following Ortiz against same team
2.54 13-3 0.813% Schmidt following Livan against same team
2.71 10-2 0.833% Woody following Schmidt against same team
4.01 16-8 0.667% Ortiz following Woody against same team
----------------------------------------
3.55 0.694%
The difference between 0.694 and 0.455 is 0.239. Multiply 0.239 by seven
games, and you get 1.7 games. Put another way, if this analysis is fully
predictive, Baker and Righetti gave away two games (and thus almost
definitely the series) by setting up their rotation wrong. Even if this
analysis is only partially predictive, there will probably continue to be a
cost - as there already was in game two, with Ortiz getting shelled.
The only reasoning that I have heard for setting up the rotation the crazy
way that it is is that it "felt right", and that they didn't want to be "too
scientific" about their decisions. Arrrgggghhh. It's at times like this
that I am ready for someone else to be making these tactical decisions for
the Giants, whatever Baker's leadership skills are.
Adam Coutts
simy1
> Dusty has made a few decisions that sometimes make it seem like he is ready
> to give runs away this World Series. As we've discussed, DHing Dunston and
> leaving Minor off the roster is one such move. As we've discussed, letting
> tight-hamstring Lofton play center and miss balls while wide-ranging Shinjo
> DHs or sits on the bench is another.
>
>
Fine post. I had posted about the rotation as proposed here, earlier,
but I had no idea the effect was so large. Incidentally, the error is
binomial and approximately equal to 0.045, that is, it is more than
five standard deviations - with virtual certainty, a true effect. To
me it is clear that we are going to lose this series, and that Game 2
was the decisive turn. He may have cost us the series.
michael bakunin
same table with runs against, or avg/obp/slg against?
-mb
j...@socrates.berkeley.edu
: Fine post. I had posted about the rotation as proposed here, earlier,
: but I had no idea the effect was so large. Incidentally, the error is
: binomial and approximately equal to 0.045, that is, it is more than
: five standard deviations - with virtual certainty, a true effect. To
: me it is clear that we are going to lose this series, and that Game 2
: was the decisive turn. He may have cost us the series.
No, it's not a true effect. Well, it might be, but there's nothing like
virtual certainty going on here.
There are an infinite number of ways to slice baseball statistics. By
chance, lots of them are going come up odd. If you're on the lookout for
odd splits, you'll find them, and some of them will be really, really out
there -- but that doesn't prove anything. That's not how statistical
testing works.
JHB
Adam Coutts
:> virtual certainty going on here.
:>
:> There are an infinite number of ways to slice baseball statistics. By
:> chance, lots of them are going come up odd. If you're on the lookout for
:> odd splits, you'll find them, and some of them will be really, really out
:> there -- but that doesn't prove anything. That's not how statistical
:> testing works.
:>
:> JHB
If I had run a whole bunch of random tests, and had come up with this one
result out of all of my results, then your point would be valid. But what
happened is that I started with one singular hypothesis - that alternating
the Giants heat and junk would work out better than the way Baker did it -
and then tested it. And I found an extreme result. So I think that its
signifigance is valid.
Adam Coutts
Adam Coutts
Thanks.
:>but not the use of W-L. Can you give us the
:> same table with runs against, or avg/obp/slg against?
Did you see that I had starting ERA in the left hand column?
I could also do WHIP or [total team runs allowed/game], but it would take
some time.
simy1
Adam, this is the kind of work that has given the Baseball Prospectus
a good reputation. Guys like Davenport or Wolverton will recognize at
once that this is almost certainly a true effect- five standard
deviations just do not happen due to random statistics - and it can be
tested against other teams in a variety of ways, eg, teams with two
lefties and teams with two fireballers and two junkballers. If you
manage to post it on that site before the end of the WS, it will
attract a lot of attention. A guy like Gammons, always looking for
tidbits, would be all over it.
Rubio, can you help? No need to worry about the Giants, the WS are
over.
michael bakunin
Thanks -- I'm blind sometimes.
> I could also do WHIP or [total team runs allowed/game], but it would take
> some time.
Earned runs are fine with a large-enough sample. My worry is the park
effect. How many of each sample were home games? Studies I'd seen in
the past found no support for the common wisdom of breaking up fire-
and junkballers, so this was intriguing.
-mb
Jamal Bernhard
"Adam Coutts" <pre...@sbcglobal.net> wrote:
> Team
> ERA W-L
> ------------------------------------------
> 4.38 9-5 0.643% Schmidt being first pitcher to face a new team
> 4.81 5-10 0.333% Ortiz following Schmidt against same team
> 4.47 6-9 0.400% Livan following Ortiz against same team
> 3.91 10-11 0.476% Woody following Livan against same team
> 4.25 6-4 0.600% Schmidt following Woody against same team
> 4.81 5-10 0.333% Ortiz following Schmidt against same team
> 4.47 6-9 0.400% Livan following Ortiz against same team
> ------------------------------------------
> 4.44 0.455%
>
>
> ------------------------------------------
> 4.38 9-5 0.643% Schmidt being first pitcher to face a new team
> 2.71 10-2 0.833% Woody following Schmidt against same team
> 4.01 16-8 0.667% Ortiz following Woody against same team
> 4.47 6-9 0.400% Livan following Ortiz against same team
> 2.54 13-3 0.813% Schmidt following Livan against same team
> 2.71 10-2 0.833% Woody following Schmidt against same team
> 4.01 16-8 0.667% Ortiz following Woody against same team
> ----------------------------------------
> 3.55 0.694%
Thanks for the great stats, Adam. Several people have questioned the
validity of these findings based on sample size. There is certainly not
enough data here to take all of these differentials as "true"
differences. For example, Schmidt's numbers are different when following
Woody versus Livan, even though both pitchers are on the "junk" side of
things. In fact, you might think the lefty/righty thing would come into
play and Schmidt would be more successful after Woody, when the stats
show otherwise. Also, while Ortiz is 5-10 when following Schmidt and
16-8 when following Woody, his ERA difference is not as dramatic (4.81
vs 4.01).
Still, I do believe there is *some* effect shown here. Plus, this
doesn't factor in (1) Woody's road/home records, or (2) The DH and Ortiz
not being able to hit at home. Anaheim is a great fastball-hitting team
and a well-disciplined team at the plate. This combination matches up
poorly with Russ. No one expected him to have as bad an outing as he
had, and who knows what would have happened with Livan in game 2 and
Ortiz in game 3 -- the Angels are hot at the plate and it might not have
made any different in what order we threw our pitchers to the lions --
but since I was against this rotation before the series started I'm
allowed to bitch about it without having to claim that it's all in
hindsight... :-)
Tommy
>Dusty has made a few decisions that
>sometimes make it seem like he is ready
>to give runs away this World Series. As
>we've discussed, DHing Dunston and
>leaving Minor off the roster is one such
>move. As we've discussed, letting
>tight-hamstring Lofton play center and
>miss balls while wide-ranging Shinjo
>DHs or sits on the bench is another.
Etc. (about the pitching). As has been discussed by so many people in so
many venues for so many years, the sloop is a "players' manager." Which
is why so many players love to play for him, as opposed to an Earl
Weaver or a Bobby Valentine or someone of their ilk, managers that
players hate playing for. That is sloop's strong suit.
His weak suit, sadly, is managing the game technically. Crunching the
numbers. Whom to play and start and where and when. Strategizing. Trying
to outfox his opposite number in the other team's dugout. Sloop's
managing style? Sit back and wait for Kent or Bonds to hit a three-run
homer.
This is why Gigantic and I -- and others, perhaps -- are so hopeful that
he might leave after this season. I, for one, am ready for an asshole
who knows what he is doing in the dugout to manage the Giants. Screw the
players' feelings!