it only hurts when i gaffe

by lboros on Dec 19, 2007 12:43 AM EST in St. Louis Cardinals News

here's an early xmas present for ev'ybody: an extra 25 runs on tuesday's projected team run total. i inadvertently changed a plus sign to a minus in the base runs formula, and that threw all my calculations off --- they came in too low. here are the corrected figures:

	AB	H	2B	3B	HR	W	AVG	OBP	SLG	BR
ZIPS	5555	1414	288	19	160	560	.255	.323	.400	725
CHONE	5555	1439	280	23	171	589	.259	.330	.410	759
2007 TOTAL	5529	1513	279	13	141	560	.274	.340	.405	761

the cards' actual run total last year was 725, so ZIPS --- bearish though it is on nearly every cardinal hitter --- projects the 2008 offense to be no worse than 2007's. and CHONE projects the offense to be about 35 runs better than last year's. . . . . well, sort of. the issue is confused by the fact that the cards underperformed their base runs estimate last year by 36 runs; their basic inputs (baserunners and extra-base hits) should have produced about 761 runs, but only 725 crossed the plate. that's a significant discrepancy and worthy of its own post; i'll take the matter up on thursday.

i owe a big thank you to Sleepy CA for checking my math and helping me find the error in my calculations.

Tweet Comment 30 comments | 0 recs |

Do you like this story?

Comments

Damn the integers anyway!!!

Damn the positive and negative integers anyway! I'm just relieved that the numbers don't work out as bad as they looked orginally.

"The Cardinals have won a World Series in THEIR new stadium!" --my Uncle Jim to a heckling Cubs fan

by fourstick on Dec 19, 2007 12:36 AM EST reply actions

I am glad

the mistake was on the low side. Thank goodness for small blessings.

The hot stove is burning...

by cardzfan24 on Dec 19, 2007 1:01 AM EST reply actions

so, does this mean that it's not completely

out of the question for the Cards to score enough runs to win the World Series?

all this math makes my head hurt...

When cheese gets its picture taken, what does it say?

by RosevilleRedbird on Dec 19, 2007 1:19 AM EST reply actions

We are all such nerds

So we've made up 35 runs on out pythag by doing nothing basically? Still well under .500...

Cheeseburger in paradise.

by joker24 on Dec 19, 2007 1:28 AM EST reply actions

70% accurate

lb:
going back to the "70% accurate" idea: For which measurables are zips and chone projections so accurate?

The St. Louis Cardinals- 11 time World Champions!

by Zubin on Dec 19, 2007 1:29 AM EST reply actions

That's just

what projections are seeing as r coefficients. Marcel (basic three years of stats and a regression) gets you to like .65 and the more complex systems like PECOTA and ZiPS only really make incremental improvements on that, near or at .7 or so. I forget.

There are estimates that .73 to .77 are the limit, even if God handed you each player's true talent level, just because of expected variance within a small sample (say 550 plate appearances).

Pitcher projections are much worse. Something like .45 or .5 or something.

by plh903 on Dec 19, 2007 5:21 AM EST up reply actions

I know that...

But what statistic(s) or variable(s) do the models project? Also what is the total variation in each variable so we can figure the amount of variation that the model can't account for (essentially our error bar).

The St. Louis Cardinals- 11 time World Champions!

by Zubin on Dec 19, 2007 7:54 AM EST up reply actions

Z, i'm not sure

which stat columns the .70 R coefficient refers to. i'll send out a couple of e-mails to see if i can find out.

by lboros on Dec 19, 2007 10:26 AM EST up reply actions

OPS is what i've seen the correlations

Cheeseburger in paradise.

by joker24 on Dec 19, 2007 1:46 PM EST up reply actions

What do

you guys think about Clemens - do you believe his denial statement?

It seems like most guys are admitting they used the stuff.

by Toddius396 on Dec 19, 2007 1:33 AM EST reply actions

"I did not take steroids"

Of course not, someone ELSE administered them in your a$$.
Of course you didn't "take" them. By mouth. You injected them.

by sdrone on Dec 19, 2007 10:04 AM EST up reply actions

Is he guilty,

or is he innocent, well it depends on what the definition of IS, IS. Looks like he's gonna use the Bill Clinton defense; deny, wiggle and squirm and deny, deny. It'll work, the pressure will be off soon enough and everybody'll move on.

by ridgesee on Dec 19, 2007 10:49 AM EST up reply actions

He didn't take them,

he received them.

The St. Louis Cardinals- 11 time World Champions!

by Zubin on Dec 19, 2007 10:53 AM EST up reply actions

Glad you caught that part

the issue is confused by the fact that the cards underperformed their base runs estimate last year by 36 runs;

When you use the formulas that adjust for GIDP's, CS's, SB's, and IBB's (I can't figure out why IBB's are different than BB's and HBP's, but the "complicated" formula separates them), the discrepancy falls significantly (to ~13 IIRC, spreadsheet is on comp at work), and when you factor in SH's (which none of the "base runs" formulas account for, but since they are outs that are given away, it seems that they would fit fine in the "c" part AFAICT) the discrepancy drops to "underperforming by ~7 runs".

Yeah, what an awful run-on sentence.

Anyway I hate being such a sabermetrics n00b, but one thing i can't figure out is how or if base runs accounts for runs scored on errors. It seems that should almost be a scalar factor multiplied across the rest or the runs scored (except for maybe HR's)? Does anyone know? We had 64 ROE's last year, but of course we would have scored some of those even if an error hadn't occurred.

Christmas lights are an extremely efficient method for converting Christmas Spirit into heat. -anon

by SleepyCA on Dec 19, 2007 2:07 AM EST reply actions

They can be

largely ignored and you'll still get good results, but here you go.

by plh903 on Dec 19, 2007 5:26 AM EST up reply actions

Oh, to answer another

question, IBB are separated because they have a different linear weights value than regular BB. Since it's a strategic decision, they lead to less run scoring, despite the stupidity of managers and the prevalence of the intentional unintentional base on balls.

by plh903 on Dec 19, 2007 6:04 AM EST up reply actions

Wait, explain this



the cards' actual run total last year was 725, so ZIPS --- bearish though it is on nearly every cardinal hitter --- projects the 2008 offense to be no worse than 2007's

Then your table says:

Projection BR ZIPs 725 CHONE 759 2007 Total 761

So, wouldn't the 2007 be 761, not 725?

I will be boxer briefs

by Hardcore Legend on Dec 19, 2007 2:14 AM EST reply actions

it's complicated...

There are 2 formulas for "base runs" The formulas are shown here, at wikipedia.

The 2007 actual runs scored (by the real-world MLB team) was 725.

The 2007 "predicted runs scored" using the limited version of dan agonistes' "base runs" model that only uses total bases, HR's, BB's, HBP's and AB's by the actual 2007 MLB team -- ie does not consider sacrifice hits, GIDP's, or IBB's -- and then tries to determine how many runs they should have scored given the offense performance above was ~761.

In the statement "the issue is confused by the fact that the cards underperformed their base runs estimate last year by 36 runs", 36 = 761 - 725. Last year we scored 36 runs less than dan agonistes "basic" formula for base runs predicted we should have.

Now, using that same formula and the ZIPS player projections, we'll score 724.8 runs next year. Make of that what you will.

The other form of the "base runs" model uses tb, hr, hits, stolen bases, caught stealing, GIDP, etc. Calculating 2007 "predicted runs scored" value using this model gives ~738. Adding "sacrifice hits" into the "c" formula on the wikipedia entry reduces the number of predicted runs to ~731.

That isn't really useful unless you have a way of predicting how many sacrifices, SB's, GIDP's, and CS's we'll have next year, but if you have that then you can probably use the improved "base runs" model to provide a better estimate of total performance. I'll try to put up some sensitivity analyses tomorrow if I get time.

Christmas lights are an extremely efficient method for converting Christmas Spirit into heat. -anon

by SleepyCA on Dec 19, 2007 4:14 AM EST up reply actions

If only all of our gaffes felt as good as this one

Here's a simpler analysis, done last weekend, which is aligned with your projections, Larry. that the Cardinals should score more runs next year than they did in 2007:

OPS+ for Cardinals 2008/career

Projection - Player - Record:

162 Pujols 157/167
Regression to the mean should raise Albert's performance slightly

120 Duncan 115/125 - Ludwick 110/98
Duncan's second half performance was evidently hindered by his hernia, so we should expect improvement from him in 2008. But Ludwick, the presumed platoon in LF, may perform less well than this year, if he regresses to his mean. So production from LF should be about the same.

120 Ankiel 120/NA
Rick seemed to be very badly affected by the news reports of his HGH use. Given his vindication by MLB officials and the far worse allegations against big ML names, it would not be surprising to see Rick match his performance in 2007. Even if he falls off 10%, he would have a higher OPS+ than Encarnacion et. al. Production from RF should improve.

100 Schumaker 111/89 - Barton [105 Edmonds 88/133 (110 in 2006)
Assuming Barton and Schumaker platoon in CF until Rasmus arrives, that combination of players should be able to match the 2007 performance from CF. Call it a wash.

108 Rolen 89/126 (126 in 2006)
Scott will reportedly be much improved after this latests surgery. If he can make up even half of his drop-off from his 2006 season, the production from 3B will be significantly better (assuming he isn't traded at some point during the season).

80 Kennedy 50/88 (86 in 2006)
Regression to the mean for Kennedy, plus being backed up by Ryan rather than Miles, should lead to a significant improvement from 2B.

85 Molina 85/69
Yadier may regress to the mean, but it wouldn't be surprising to see him match his 2007 performance, now that he has adopted a consistent stance. He's also in his prime years.

88 Izturis 69/67 (88 in 2004) - Ryan 95/NA
years.
Much has been written about the supposed drop-off expected in production from the SS position. I won't reiterate here the case for the Izturis/Ryan platoon being able to match the Eckstein/Miles combo of this year. I'll just say that the L/R splits for Cesar and Brendan fit together very well. Nevertheless, let's assume that performance will fall off at the SS position.

In this analysis, offensive performance in 2008 should improve significantly at 3B, be better in RF, LF, 2B, be about the same from C, 1B, CF, and decline only at SS. This broader stroke analysis supports what Larry reports. The Cards should see a modest increase in runs scored in 2008.

by CardsWin on Dec 19, 2007 2:38 AM EST reply actions

About those 2B projections

I think those projections for significant improvement out of 2B might be a bit optimistic. While we had 280 craptastic AB from Kennedy, Miles' surprisingly good season (for his standards) picked up a fair amount of slack there.

I guess I am just not very sold on Kennedy. His bat looked "dead" last year. Maybe his knee was an issue all along--I hope so, because the alternative explanation--that he simply is done--is much worse.

by Hungry Jack on Dec 19, 2007 10:42 AM EST up reply actions

Kennedy's bat did look

dead to me also and I'll be surprised to see him come back with any kind of offense. I have just seen too many players drop off at his age and not that many recover. He's got an odd swing anyway

by ridgesee on Dec 19, 2007 11:03 AM EST up reply actions

If Kennedy fails again, then what?!

I certainly wouldn't disagree with you about Kennedy's prospects. His outlook is a big uncertainty. At least Ryan could fill in vs. LHP, if Kennedy continues to falters, so they'd form a platoon. Kennedy's hitting vs. RHP is better.

by CardsWin on Dec 19, 2007 2:02 PM EST up reply actions

Kennedy

If he is really bad again the Cardinals should cut him. At that point they can give Hoffpauir a chance at 2B. That is the spot Hoffpauir will get a shot to play. If Kennedy is bad there is no point in keeping him around just because he is owed money. He is wasting a roster spot if he is terrible. I think Hoffpauir's chance of playing this year is dependent on Kennedy mainly. See Ryan can play 2B as shown last year so I only see Hoffpauir getting a shot if Kennedy is flat out awful.

by ICbirdfan on Dec 19, 2007 4:54 PM EST up reply actions

what is ZIPS/CHONE

Hi, I'm new to the site. I really enjoy reading everyone's comments. I was curious, what is ZIPS and CHONE? Also, I am pleasantly surprised to see that these projections don't cut down on the Cards run total...although I hope that they cut down on the number of runs our opponents score.

by idahobirds on Dec 19, 2007 4:02 AM EST reply actions

They are

projection systems. One can be found at http://www.baseballthinkfactory.org/files/oracle/ and http://lanaheimangelfan.blogspot.com/

Dan Szymborski does the ZiPS at BTF and Chone Smith (as he calls himself on his blog, "Rally" or something at BTF) does the CHONEs.

They take various details about a player like comparables, batted ball information, aging curves, mainly past performance, whatever else I'm forgetting and come up with computer generated projections for the MLB population (and plenty of minor leaguers).

They will cut down on the runs prevented total, just because you really can't "predict" a pitching staff to be that bad again. How much, I don't know offhand but regression to the mean will be a good thing in 2008.

by plh903 on Dec 19, 2007 5:35 AM EST up reply actions

Thanks

Very interesting. I've got those pages bookmarked.

by idahobirds on Dec 19, 2007 5:47 AM EST up reply actions

ZiPS

Dan also wrote up a brief ZiPS FAQ that explains his method in a bit of detail. I'm don't know if there's a description published of CHONE's method, but they're essentially all refinements of Tango's basic weighted regression model, Marcel.

Surprisingly, no. I'm free, I'm unfettered.

by liam on Dec 19, 2007 2:29 PM EST up reply actions

I will

take that early present, lb. Yesterday's projections were harshing my Christmas joy.

by cardsgirl95 on Dec 19, 2007 10:54 AM EST reply actions

Congratulations, Larry, you're now an accountant..

FWIW, any accountant worth his salt has swapped a debit for a credit (or vice versa) at some point in his or her career; it's an occupational hazard that strikes everyone now and then.

Just make a few more and you'll qualify to be a CPA..... :)

Dave

by Sydney dave on Dec 19, 2007 3:49 PM EST reply actions

Old Joke

From here, sort of:

There was a mad scientist ( a mad ...social... scientist ) who
kidnapped three colleagues, an engineer, a physicist, and a
topologist, and locked each of them in seperate cells with plenty
of canned food and water but no can opener to see how they would cope with their imprisonment.
A month later, returning, the mad scientist went to the engineer's
cell and found it long empty. The engineer had constructed a can
opener from pocket trash, used aluminum shavings and dried sugar to
make an explosive, and escaped.
The physicist had worked out the angle necessary to knock the lids off
the tin cans by throwing them against the wall. She was developing a
good pitching arm and a new quantum theory.
When the mad social scientist came to unlock the topologist's cell, he found it empty aside from the can of food, still closed and seemingly untouched. Suddenly, a knocking comes from inside the can. The can is opened, the topologist comes out blinking and says: "Oh shit, sign error."

Surprisingly, no. I'm free, I'm unfettered.

by liam on Dec 19, 2007 4:05 PM EST up reply actions