Wednesday, February 21, 2007

2006 Reds Hitting Review

Before we all get seduced by whoever outperforms expectations in spring training when those games begin, I wanted to take a few final looks at the 2006 season. Today, we'll look at the hitting.

2006 Reds Hitting Recap
The 2006 Reds were a very different team from those that we knew the year before. While the offense was amazing in '05 (5.03 r/g, 1st in league), the 2006 Reds--following a brilliant April--were much weaker offensively, scoring "only" 4.62 r/g, which was good for just 9th in the 16-team National League in runs scored (749). They were second in the league in home runs (217) and walks (614), third in stolen bases (124), but were seventh in OBP, SLG, and OPS (0.336, 0.432, 0.768, respectively), and were near the bottom in batting average (0.257-15th), total hits (1419-13th), and doubles (291-13th).

The causes for this drop-off between years were many, though I would identify two primary factors that led to this change. The first, of course, was the trade of Austin Kearns, & Felipe Lopez for bullpen help on July 13th. And second, was a substantial drop-off in production from the two most important offensive threats on the '05 team, Adam Dunn and Ken Griffey, Jr. These problems would eventually come to a head in September/October, when the Reds would score only 3.3 runs per game as they eventually finished 3.5 games behind the would-be World Champion St. Louis Cardinals.

This graph shows the Reds' runs scored (red) and allowed (blue) over course of the year. The black line indicates their cumulative record each month. Runs scored fell below runs allowed the last three months of the season, which is reflected in the downward trajectory of their record during that time.

The Graphical Reds

I'm a big advocate of using a graphical approach in data analysis because I find that graphs often yield insights that are difficult to see when looking only at the numbers. Here are a few graphs that I found interesting when looking at the 2006 Reds' individual hitting stats:

On-Base Percentage versus Isolated Power:
Vertical and horizontal lines indicate league averages. Here, as with all subsequent graphs, I only included the 15 players Reds with 100 at-bats or more with the team. Kearns and Aurilia overlap in the upper-right, while LaRue and Phillips overlap in the lower-left. Isolated power = SLG - AVG.

This is a graph that has been used over at Hardball Times for quite a while (though usually it's used to analyze teams, not players), and I think it gives a nice overview of the strengths and weaknesses of the Reds' players. A few observations:
  • Those in the upper-right are the players that everyone wants: above-average ability to get on base (OBP), and above-average ability to get people around the bases (isolated power). The Reds had five of these players last season, though they traded away one mid-season.
  • During his career, Griffey has almost always been in the upper-right of these graphs. Last year, however, he had a very difficult time getting on base, posting the lowest OBP of his career. I know there were injuries involved, but he may have been starting to show his age.
  • Dave Ross was superhuman last year, eh?
  • Royce Clayton's lack of offensive value smacks you in the face in this graph, being the only Reds player who was deep in the bottom-left corner. What a supposedly "sabermetric" G.M. like J.P. Riccardi wants with him, I have no idea.
Strikeout vs. Walk Rates:
Horizonal and vertical lines are league averages. Encarnacion and Denorfia overlap near the center.

This one is more of an interesting diagnostic about hitter type than anything else:
  • There is a fairly tight relationship between walk rate and strikeout rate for most players (at least on the Reds, though this is a commonly-seen pattern throughout baseball):
    • Players who take more pitches and work the count are more likely to draw a walk, but are also more likely to strike out. Adam Dunn, of course, is the poster boy for this approach, though Dave Ross was just behind him last season.
    • Players who are less patient, like Brandon Phillips and Rich Aurilia, tend to have both fewer walks and fewer strikeouts.
    • Both approaches (Dunn v. Aurilia) can result in valuable offensive contributions.
  • Of course, then there's Scott Hatteberg. I always knew he was a unique hitter, but wow.
Update: Please see my 2006 Reds Hitting Review Addendum for an additional analysis comparing strikeouts and walks.

Value Over Replacement Player v. Win Probability Added:

Update: Please note that I'm no longer comfortable with the following analysis, and have revised it in the 2006 Reds hitting Review Addendum, posted 22 Feb. I'm going to leave it as it is, however, to avoid causing additional confusion.Regression line is the relationship between VORP and WPA on the Reds last season (I couldn't easily get the data to calculate the league-wide regression line). Freel and Hatteberg overlap near the center.

VORP is a widely-respected indication of overall offensive contribution, adjusted for position (shortstops score better than first basemen with the same offensive stats). It uses classical scorebook statistics like hits, doubles, home runs, walks, and stolen bases when it is calculated, and gives its values as runs produced above what you'd expect from a "replacement player" in that position (replacement players are players that would be easily attainable either through a mid-season free agent signing, minor league promotion, etc).

In contrast, WPA is a "new" statistic (ok, not really new, but it became very popular last season) that reflects the situation-specific player contribution to winning ballgames (a tie-breaking hit in the 9th inning has greater impact than a tie-breaking hit in the first inning). Both try to measure overall contribution to winning, but do so in very different ways and allow some interesting observations to be made:
  • Two Reds made contributions to winning ballgames that were far in excess of what you'd expect from their "scorebook" offensive stats that were used to calculate their VORP:
    • Griffey's contributions tended to come with drama, such as his 3-run home run to win the last home game of the season.
    • Edwin Encarnacion had a few highlights, but overall was just consistently good all year in high-leverage situations. I wouldn't expect to see much debate about whether he plays this year, regardless of his error rate at 3B (and I'll wager that he'll be better in that regard).
  • Kearns and Phillips seemed to do worse than you'd expect in "clutchy" situations. Phillips was very streaky. He was excellent in April (WPA=0.91), June (WPA=0.46), and August (WPA=0.14, but VORP = 14). But when he slumped, he did so badly and seemed to be an automatic out in important situations during those times (May WPA = -0.36, July WPA = -0.73, September WPA = -0.61). It'd be nice to see him avoid some of those disastrous months this year--he can be awesome when he's "on."
  • Aurilia sure was good last year, wasn't he?
Surprises and Disappointments:
Diagonal line indicates a perfect match between PECOTA's projections and what actually happened. Valentin and Lopez overlap slightly near the center.

Despite the disappointments of last season, the Reds also had some wonderful surprises. I wanted to try to illustrate this graphically, and this was the best I could come up with. I didn't always agree with PECOTA's projections, but they were generally close enough to my expectations that they seemed to make a good comparison. Observations:
  • The amazing seasons by Ross, Aurilia, and Hatteberg really pop out at you. Aurilia and Hatteberg had both had one of those rare, late-career surges that came close to matching anything they'd done in previous seasons, while Ross almost certainly had his career season last year after starting the season as the #3 catcher.
  • Phillips' season doesn't look quite as good, but his streakiness as well as his AVG-heavy batting line resulted in a season that wasn't quite as good as I think most of us remember it being.
  • Freel was right in line with his previous seasons. He shows up as a mild surprise here because PECOTA didn't like him last year (it's a bit better this year).
  • The enormously disappointing seasons by Griffey, Dunn, and LaRue are glaring in this graph.
Bad luck, or just bad performance?
Line indicates a perfect match between actual OPS and OPS predicted by batted-ball data from last season.

I'll finish up with this graph, which I'll use to try to explain some of the variation we saw last season.
  • PrOPS has been a stat that I've grown increasingly infatuated with after I spent part of my wife's child labor reading about it while we were waiting in the delivery room last May (no joke--she wanted me out of her hair, and I happened to have the Hardball Times 2006 Annual with me).
  • It predicts OPS based on batted ball data, and seems to be a good analog to defense-independent predictors of ERA for pitchers. Players that outperform their PrOPS generally do worse the next season, while players who underperform their PrOPS tend to "improve." Therefore, divergences between PrOPS and OPS might be chalked up to "luck."
  • Most Reds group pretty close to the line, which indicates that their performances are explainable by their batted ball data--that's a good thing. It means that there's some reason to think that Ross, for example, really was that good and wasn't just lucky last season (doesn't mean he'll repeat, of course).
  • Two Reds, however, Adam Dunn and Jason LaRue, had PrOPS's that were much higher than their actual OPS. This indicates that they were hitting baseballs in a fashion that would predict much better outcomes than we saw in their statistics. That's not due to peculiarities of these players either--in previous seasons, PrOPS has done a much better job predicting their OPS. Barring an injury, a decline in skills, or some radical change in approach at the plate, this graph hints that Dunn and LaRue might hit much better in this coming season. LaRue won't help us, of course, but Dunn is probably the key to the Reds offense in 2007.
  • Unfortunately, Griffey's disappointing season seems to be right in line with his batted-ball data.
Now that we've seen the graphs, I'll close with a selection of informative batting stats on the '06 season, sorted by value above replacement player (VORP):
Player PA %K %BB %HR SB SB% AVG OBP SLG OPS PrOPS ISO VORP WPA
Rich Aurilia 474 11% 7% 5% 3 100% 0.300 0.349 0.518 0.867 0.877 0.218 27.4 1.27
Adam Dunn 673 29% 17% 6% 7 100% 0.234 0.365 0.490 0.855 1.001 0.256 23.5 1.19
Brandon Phillips 571 15% 6% 3% 25 93% 0.276 0.324 0.427 0.751 0.772 0.151 22.6 -0.14
David Ross 284 26% 13% 7% 0 --- 0.255 0.353 0.579 0.932 0.967 0.324 22.6 0.89
Edwin Encarnacion 447 17% 9% 3% 6 67% 0.276 0.359 0.473 0.832 0.839 0.197 17.0 1.95
Scott Hatteberg 530 8% 14% 2% 2 50% 0.289 0.389 0.436 0.825 0.870 0.147 16.9 0.05
Ryan Freel 511 19% 11% 2% 37 77% 0.271 0.363 0.399 0.762 0.758 0.128 16.4 0.09
Ken Griffey Jr. 467 17% 8% 6% 0 --- 0.252 0.316 0.486 0.802 0.870 0.234 16.0 1.82
Austin Kearns 360 24% 10% 4% 7 88% 0.274 0.351 0.492 0.843 0.851 0.218 15.7 -0.30
Felipe Lopez 390 17% 12% 2% 23 79% 0.268 0.355 0.394 0.749 0.775 0.126 14.9 0.25
Javier Valentin 199 15% 7% 4% 0 --- 0.269 0.313 0.441 0.754 0.837 0.172 4.5 -0.18
Norris Hopper 45 9% 13% 2% 2 50% 0.359 0.435 0.462 0.897 0.841 0.103 3.4 0.10
Juan Castro 100 13% 5% 2% 0 0% 0.284 0.320 0.421 0.741 0.745 0.137 1.7 -0.31
Chris Denorfia 117 18% 9% 1% 1 50% 0.283 0.356 0.368 0.724 0.692 0.085 1.1 0.08
Tony Womack 22 14% 18% 0% 0 --- 0.222 0.364 0.333 0.697 0.756 0.111 0.2 -0.01
Brendan Harris 11 36% 9% 9% 0 --- 0.200 0.273 0.500 0.773 0.820 0.300 0.2 0.04
Brandon Watson 0 --- --- --- 1 100% 0.000 0.000 0.000 0.000 0.000 0.000 0.0 0.03
Andy Abad 5 0% 40% 0% 0 --- 0.000 0.400 0.000 0.400 0.773 0.000 -0.3 -0.03
Todd Hollandsworth 74 26% 8% 1% 0 0% 0.265 0.324 0.397 0.721 0.653 0.132 -0.4 -0.53
Cody Ross 5 40% 0% 0% 0 --- 0.200 0.200 0.200 0.400 0.366 0.000 -0.6 0.02
Ray Olmedo 48 8% 8% 2% 1 100% 0.205 0.271 0.318 0.589 0.718 0.113 -1.6 0.25
Quinton McCracken 57 16% 7% 2% 2 100% 0.208 0.263 0.321 0.584 0.716 0.113 -2.1 -0.52
Royce Clayton 160 20% 7% 1% 6 67% 0.235 0.290 0.329 0.619 0.689 0.094 -3.8 -0.32
Jason LaRue 218 23% 12% 4% 1 100% 0.194 0.317 0.346 0.663 0.861 0.152 -4.3 -0.85
Dewayne Wise 38 16% 0% 0% 0 --- 0.184 0.184 0.237 0.421 0.534 0.053 -4.6 -0.36

Still to come...2006 Reds Pitching Review, 2006 Reds Fielding Review, and my own awards for the Reds' Best Hitter, Best Fielder, and Best Pitcher of last year.

11 comments:

  1. Effing good stuff, J.

    The Phillips thing has been bugging me for a few weeks. I used to think he had a bad second half, but it's something else - up and down from month to month. Maybe that's typical of a guy whose production is so batting-average-dependent, but it worries the heck out of me.

    ReplyDelete
  2. First of all, excellent post. There are times when I think about getting back into the game and then you come along with stuff like this and show me that I am unnecessary. Great work!

    I'm not as sold on PrOPS, especially for someone like Dunn. If you look at his numbers, he consistently underperforms his PrOPS every year. The thing that sticks out for me is is PrAVE and it makes me think that PrOPS breaks for someone with such a high percentage of strikeouts. Honestly, I can't imagine a world where Dunn would bat .293 in a season. His BABIP would have to be around .500 to accomplish that.

    All that being said, I think Dunn was unlucky last year. I did a diary at Red Reporter back in January on Dunn's batted balls and I found that Dunn had about 10 more singles on line drives than expected and about 14 fewer extra base hits than expected based on his previous levels. The big thing that caught my eye, and that I think you've mentioned before, was that Dunn was terrible at getting hits on ground balls. It seemed like teams were playing the shift against him much more last season, and even though he hit considerably fewer GB than previous seasons, he still managed to have about 8 hits fewer than his previous rates on GB.

    So, much like PrOPS predicts, it's entirely possible that Dunn could put up a very good season by doing nothing different except getting hits to sneak through more frequently (though he will need to up his HR per line drive rate as well). I doubt he'll ever put up a 1.000+ OPS with all of those strikeouts, but I don't think .930 is out of the question. Hopefully he can manage it, because like you said, he's probably the key to the 2007 offense.

    ReplyDelete
  3. Hi Joel,

    It is true that Dunn consistently has underperformed PrOPS in his career, but not to the degree that he underperformed last year. Here's the difference between OPS and PrOPS for him the last three seasons:
    2004 -0.035
    2005 -0.054
    2006 -0.145

    It may very well be that an extreme flyball/strikeout hitter like Dunn might not be well-predicted by PrOPS, but the difference is probably more on the order of 0.05 OPS points, not three times that--which is what we saw from this last past season. While he might not have hit 1.000, I think that, with "normal" luck, he could certainly have hit 0.950.
    -j

    ReplyDelete
  4. Just an additional small comment--I was thinking about this in bed last night, and I think the WPA vs. VORP comparison is probably inappropriate. The reason is that VORP adjusts for player position, whereas WPA does not. While that's normally a wonderful feature for VORP, it makes for some invalid comparisons.

    I'll try to re-do that part of the analysis with Runs Created or Win Shares tonight, as those aren't position-adjusted. It should make some guys, especially Phillips, look much better--his VORP is much higher than his relative RC because he's at a low-offense position. -j

    ReplyDelete
  5. You may be right J (about PrOPS), though I still think .930 is a more realistic number for his OPS, it's not too far off of the .950 that you suggest.

    I need to correct something in my post. I said he would need a BABIP around .500 to bat .293, but that was a silly mistake on my part. I forgot to take out the HR as hits in my BABIP calc and it's really more like .380, not .500. I think .260 is about the best we can expect from Dunn for a batting average if his strike outs stay at the level they've been at the last 3 seasons. Personally I don't mind if he's still OPSing at a .900+ rate. I think he'll get back there in 2007.

    ReplyDelete
  6. Wow. Great stuff. A picture really is worth a thousand words.

    ReplyDelete
  7. I know many have said it already, but it deserves it many times over: awesome work. Awesome awesome awesome. I can't wait for the pitching review.

    ReplyDelete
  8. Thanks folks. I was pretty excited to publish this, and it's great to receive such an enthusiastic response. I'm working on the pitching review right now, though it might take a few more nights to put together. -j

    ReplyDelete
  9. Absolutely, J. This is fantastic and incredibly fun to read.

    Keep up the good work!

    ReplyDelete