A Midseason Look at Projections vs Reality

by Adam Peterson on Jul 13, 2010 8:38 AM EDT in By the Numbers

During the offseason, I posted a number of articles focused on "Wins Above Replacement" (WAR) projections and the Twins roster as free agents were considered and signed. The latest article, with projections, was posted after the Jim Thome signing. At that time, I projected the Twins as an 87 win team, bumped up to 90 wins after signing Orlando Hudson. This was considered this to be a relatively pessimistic projection, compared the roughly 93-95 win consensus across the Twinkie Town community. With a 46-42 record at the All Star Break, the Twins currently project to win 84.7 games, let's round this up to 85 wins. Simply compared to my projections before the season, it appears the Twins have underperformed by around five wins. How do we explain the difference? Who has underperformed and over performed relative to my projections? (Which by the way I don't claim to be any better or worse than other projections, simply a baseline to which we can compare.) In this article, I will examine, position by position, player by player, my WAR projections to actual performance as posted on fangraphs.

	C	1B	2B	3B	SS	LF	CF	RF	DH	SP	RP	Hit	Pitch	WAR	Wins
To Date (All Star Break)	2.2	5.0	2.0	1.1	0.6	1.3	2.1	0.5	1.9	8.0	1.9	16.5	10.4	26.9	-------
Projected to end of season	4.1	9.2	3.7	2.0	1.1	2.4	3.9	0.9	3.5	14.7	3.5	30.4	19.1	49.5
Projection prior to season	6.1	3.5	2.7	0.8	3.1	0.4	3.1	1.8	2.7	13.5	5.4	24.2	18.9	43.1	90.1

I must admit, I'm comparing some apples to oranges here, as the fangraphs WAR calculations don't exactly match my projections, which use Sky Kalkman's spreadsheet from Beyond the Box Score. That said, there's a large 6 WAR gap between how the Twins players have actually performed relative to my baseline projections, and even worse, this gap is a full 11 wins in the wrong direction from the Twins actual 85 win projection based on their actual win-loss record. What gives? I'll attempt to answer this after the jump.

Before I get into a position by position, player by player analysis, let me attempt to explain the 11 win difference. First, remember that WAR projections rely on the Pythagorean method of projecting number of wins based on total runs above replacement (RAR). To date, the Twins have scored 408 runs and allowed 372 runs, which projects to a Pythagorean winning percentage of 0.546, or 48.1 wins. So at the All Star Break, the Twins have underperformed their Pythagorean by two wins, or 3.5 wins prorated over the entire season. This explains 3.5 of the 11 win difference, leaving 7.5 to account.

Second, consider that WAR, which is based on wOBA, does not consider double plays when determining value. Effectively, WAR assumes that a team grounds into a league average number of double plays. To date, as we all know, the Twins have grounded into MANY more double plays (a league leading 102 GIDP) than the average MLB team (69). Using a rough approximation of one run per GIDP (based on my rough analysis of expected runs in various double play situations, I'll refine this later in the season), this means the Twins have lost around 3.3 wins to date, 6.1 wins prorated over the full season, simply due to double plays.

We've explained nearly the entire 11 win difference above. Considering general errors in these types of projections, as well as the fact it appears (though I have not confirmed) that the Twins have grounded into so many double plays in a bases loaded situation to end an inning (-1.5 expected runs), we may find that more than 6.1 wins would be lost due to GIDP.

Looking at the projections, as a whole, it appears that the pitching staff has performed about where I projected them to be, +19.1 WAR versus +18.9, with the starters outperforming by around a run and the relievers underperforming by about the same amount. Considering the way the Twins starters have pitched over the last month, this is more than a little surprising. Looking a bit deeper, remember that WAR projections are based on Fielding Independent Pitching (FIP). The Twins starters have an overall FIP of 4.03 (same as xFIP), good for 11th in MLB, 4th in the AL. Interestingly, the overall starters ERA is much higher at 4.60, the second largest spread in the Majors (Detroit is highest at 4.81 ERA versus 4.21 FIP). One obvious explanation for this type of spread is defense. In 2009, the Twins had an overall UZR of -23 runs. This season, the Twins have much improved their defense, rating at +20.8 runs. So defense doesn't explain the gap, what does? My suspicion is that the Twins starters 20.2% line drive rate allowed (worst in AL, 4th worst in MLB) has meant many more base hits and runs, and this is not considered in either FIP or xFIP.

Offensively, first base is the largest positive, due to Justin Morneau's MVP start +5.0 WAR (+9.2 prorated over a full season). Left field (+2.0 prorated compared to baseline) has also outperformed projections, due to Delmon Young's solid start, which is a relief to most everyone here. Surprisingly, considering I projected second base at a solid +2.7 WAR, it is projecting to +3.7 at the end of the season. Orlando Hudson has really solidified what was a black hole last season. And finally, third base hasn't completely sucked (I projected +0.8 WAR of suckitude prior to the season. On the negative side, catcher and shortstop (-2.0 WAR each) stand out. At catcher, Joe Mauer's slow start has been well documented, and hopefully he'll heat up in the second half. Though I have a nagging suspicion that a hesitancy to trade Wilson Ramos may be an indication that Mauer's not 100% healthy. And at shortstop, losing J.J. Hardy for much of the first half has hurt, as has his .270 wOBA when he has played.

To sum up, WAR values to date indicate that the Twins should be winning many more games than they have actually won. This is due to under performing our Pythagorean record, which may very well be a matter of luck, and grounding into a ridiculous number of double plays, which may simply be a "skill" for this team. And our starting pitchers need to stop giving up so many line drives. What do you think?

Tweet Comment 30 comments | 3 recs |

Do you like this story?

Comments

Great analysis. 102 GIDP compared to the average of 69 is really quite atrocious.

by funnytrain on Jul 13, 2010 8:53 AM EDT reply actions

Yup

And 35 of these GIDP have been with RISP, 13 with the bases loaded. Enormous impact so far.

by Adam Peterson on Jul 13, 2010 9:52 AM EDT up reply actions

Outstanding.

Interesting note about the line drives. It might be worth looking at consequence numbers, like strand rates or percentage of base runners who score and comparing those to league average.

by Jesse on Jul 13, 2010 9:23 AM EDT reply actions

LOB%

Is right about average, 70.2% for the starters. So probably not a factor.

by Adam Peterson on Jul 13, 2010 9:32 AM EDT up reply actions

Bunt down the 3rd base line

Can’t get doubled up that way, and it advances the runners. These guys do know how to bunt…right??

by jimbo55403 on Jul 13, 2010 10:29 AM EDT reply actions

Mauer

Should be bunting down the third base line more often. They’re basically giving him a bunt single with that ridiculous shift.

by Dave MN on Jul 13, 2010 12:23 PM EDT up reply actions

Stolen Bases

Twins 26th in MLB in stolen bases 35 vs. MLB average 55. If Twins had MLB average of SB’s that would be 20 more times a guy’s on second instead of first.

by rancher33 on Jul 13, 2010 10:56 AM EDT reply actions

Depends

Adam’s crunched a lot of numbers regarding this, and he’s found that for nearly every team, the success rate is such that stealing is a losing proposition overall. The Twins have “only” lost something like one or two runs from stolen bases, which I believe is actually above league average.

In other words, twenty more steals would have given the Twins twenty more guys on second instead of first, but unless they’re able to put up a terrific success rate, it would likely also mean that they’d have had five or ten more guys erased after reaching base without any benefit at all.

I’d note that it’s possible that the Twins’ propensity for hitting into double plays mitigates the downside from stolen base attempts – if the guy’s likely to be erased on a DP anyway, then it’s not as big a loss if he’s thrown out stealing.

"There are only two things that are infinite, the universe and human stupidity, and I'm not sure about the former." - Albert Einstein

by BeefMaster on Jul 13, 2010 11:25 AM EDT up reply actions

+1

From a first order expected runs standpoint, only a small number of teams increase runs by stealing bases. This is because given the large number of expected runs lost when a runner is caught stealing, success rate needs to be over 80% as a team (varies depending on situations, CS with first and third and two outs is a larger ER loss) in order to break even.

However, I acknowledge there may be a second order effect, base stealers on first disrupt the pitcher, etc. But I don’t know if anyone has quantified this.

And good point about attempting steals mitigating the GIDP. There is a combination of GIDP Rate to SB Success % where it makes sense to attempt a steal every time since the runner would be erased anyway. I suspect it’s a higher GIDP rate than we currently have, but I’d need to crunch the numbers per situation.

by Adam Peterson on Jul 13, 2010 1:13 PM EDT up reply actions

Could they at least not get picked off so often?

Especially when they are losing 7-1.

by Old Twins Cap on Jul 13, 2010 3:01 PM EDT up reply actions

Maybe.

The Twins also don’t have a lot of guys who can steal bases. And more important than stolen bases, is success rate. Remember SB’s aren’t just a flat number. People get thrown out, too, and that needs to be taken into account just as much as personnel.

by Jesse on Jul 13, 2010 11:48 AM EDT up reply actions

GIDP

I’m just looking at things that jarringly stand out on why there are so many GIDP’s versus the major league average. Percentage of ground balls vs line drives? Percentage of extra base hits? SB’s attempted?
The top 3 teams that have grounded into the fewest double plays…SD, TB, NYM, are also in the top 4 in stolen base attempts.
I realize there are a lot more variables (harder to attempt a steal or start the runner with a left-handed pitcher vs right-handed pitcher) but Knowing that you have less than average team speed might mean that you should try more hit and run’s….and stolen base’s.

by rancher33 on Jul 13, 2010 12:39 PM EDT up reply actions

That's really interesting, actually...

…the top 3 teams that have the fewest GIDPs are top 4 in SB. I wonder if this stands up consistently historically, or if it’s a one-year thing.

I also agree with the hit & run ideas…especially since Adam has shown the Twins actually do directional hitting pretty well.

by Jesse on Jul 13, 2010 1:03 PM EDT up reply actions

It would make sense

This seems logical for two reasons:

1) Every stolen base attempt (assuming most steals are of second base) eliminates a double play situation by advancing or eliminating the runner. I would assume that, if you correct for OBP, teams who are among the leaders in CS+SB would have lower total GIDP simply because they run themselves out of GIDP chances. It’s also possible that their GIDP% could be reduced thanks to runners in motion when the ball is hit, depending on whether they give their batters green lights to swing through the steal sign, although I would guess that is not nearly as important as reducing chances.

2) Teams that steal a lot of bases generally have fast players, who are more likely to beat out the relays on potential double play balls. One way to check this would be to test for a negative correlation between SB+CS and GIDB% (which would also go along with my hit-and-run theory).

"There are only two things that are infinite, the universe and human stupidity, and I'm not sure about the former." - Albert Einstein

by BeefMaster on Jul 13, 2010 5:45 PM EDT up reply actions

I wrote an article in early June

that focused on our double plays and examined number of double play situations, ground balls in those situations, and where the ground balls were hit. Didn’t look at stolen bases though.

by Adam Peterson on Jul 13, 2010 1:15 PM EDT up reply actions

True...assuming we stole 20 more bases without a single additional runner caught stealing

by Adam Peterson on Jul 13, 2010 1:07 PM EDT up reply actions

How about this one Adam…number of GIDP when other team is in the Mauer, Morneau, Thome shift? That seems to be something that is employed more against the Twins this year compared to the Twins in previous years or other teams as a whole.
Those balls that went up the middle on Mauer are now DP’s

by rancher33 on Jul 13, 2010 1:16 PM EDT up reply actions

That was pointed out in June

and makes a lot of sense. But when I looked at the positions fielding double play balls from Mauer, Morneau and Thome, I didn’t find an inordinate number fielded by the shortstop, who would be playing up the middle. Just way too many roll over the ball hit to the second baseman.

by Adam Peterson on Jul 13, 2010 1:18 PM EDT up reply actions

I agree on the “way too many roll over’s to second base”. Hard to quantify that one, other than opposite field hit’s? Is there a stat for that somewhere?

by rancher33 on Jul 13, 2010 1:23 PM EDT up reply actions

yeah, what happened to at least moving the runner over?

by bl4ckduck on Jul 13, 2010 1:24 PM EDT up reply actions

That's really something you do more from second to third

than from first to second. With a runner on first, a base hit obviously advances to second, often to third. But a ground ball ends up ripe for the double play. From a strategic standpoint, the only way to attempt to move the runner over is with a sac bunt, and that’s a loser from an overall expected runs standpoint and only makes sense in late inning situations where it increases your overall chances of scoring a run, even though it minimizes chances of a bigger inning.

by Adam Peterson on Jul 13, 2010 1:32 PM EDT up reply actions

When I did the GIDP analysis

I used my software to capture who fielded each ground ball. For example, in early June, Mauer had hit 18 ground balls in double play situations. 17 of these were fielded by infielders (only one base hit), 4 by 1B (2 GIDP), 6 by 2B (4 GIDP), 5 by SS (4 GIDP), none by 3B and 2 by the pitcher (2 GIDP). That was a really high double play rate when he hit ground balls. Probably due to speed as well as a tendency to hit ground balls more sharply than most.

by Adam Peterson on Jul 13, 2010 1:37 PM EDT up reply actions