I was surpised to find that it's hard to argue against many of the projections Marcel has made (at least for Mariners) as I rarely agree with the projection systems' views. But of course there's always Ichiro, there's only been one year in his career where he didn't best at least one of the .365 OBP or the .418 Slugging projected for him. There's also the fairly dismal projection of Rob Johnson (where his BABiP only improves to .280) and the slightly optimistic projection of Jack Wilson (an improvement on both his career OBP and SLG). If you make an adjustment where you drop Wilson down to his average numbers, improve Ichiro up to his, and move Rob Johnson half-way to Adam Moore's (also rather optimistic) projection, then the run expectancy inches upwards a bit more to 4.99 RPG/808 RPY for an extra win by the pythag formula. But of course the run projection system is flawed in more ways than just not accounting for backups. In the tool's eyes, all players run at the same speed, hit in more or less the same manner no matter the runners on base (or their batted ball profile), and all play in the same run neutral stadium against the same league average defense.
.
=== Lucky? Or did he do it on purpose? ===
4. Go back to the Musings article and notice that Wok's lineup is whisper-close to HAL-9000's lineup.
That's not really miraculous; it's natural to have your two OBP guys at the top. I wonder if all teams whose only good hitters are table-setters have very good theoretical lineups...
One more argument for the HOF leadoff hitter. Though his OPS is relatively low, he gets lots of AB's. It's like a pitcher throwing 250 innings. Volume counts too.
Chone Figgins is not only here, but also, he will get a LOT of plate appearances. Ain't that wunnerful?
..............
It was very interesting that the M's have their best OPS guy #4 instead of #3... cutting-edge research shows that #4 is the Illinois Avenue of the lineup Monopoly board.
Maybe Wok intuited some of these little finesses. I suspect that maybe his oversized brain sensed that 1, 2, blank, 4 would work well. That, or I guess he actually ran a lineup optimizer on his PC. What if SABRMatt had a press pass and asked Wok that question? Wonder what would happen, exactly :- )
.
=== M's with the great bench this year ===
Even more good news. We proceeded on the assumption that the bench would cost any given team -40 runs.
But is that the case for the 2010 Mariners? Or are Sweeney, Byrnes, Tuiasosopo and perhaps Garko actually better hitters than the starting players?
If you were to conclude that the 2010 benchies would combine for a 100 OPS+, then add back in a +0.25 for the M's relative to the rest of the league and it looks like this:
- +0.08 ... KC ... 4.23 ... 4.31
- +0.12 ... CHW ... 4.47 ... 4.59
- -0.18 ... Det ... 4.56 ... 4.38
- +0.27 ... Bal ... 4.57 ... 4.84
- -0.19 ... Oak ... 4.69 ... 4.50 (gimme five, dude)
- +0.26 ... Cle ... 4.77 ... 5.23
- +0.26 ... Tex ... 4.84 ... 5.10 (yowch)
- -0.33 ... Tor ... 4.93 ... 4.60
- +0.20 ... TB ... 4.96 ... 5.14
- +0.21 ... Min ... 5.01 ... 5.22
- -0.10 ... Bos... 5.38 ... 5.28
- -0.51 ... LAA ... 5.45 ... 4.94
- -0.07 ... NYY ... 5.65 ... 5.58
- +0.97 ... Sea ... 3.97 ... 4.92
.
=== Dr's Diagnosis ===
It's true that the 2009 M's scored a measly 640 runs.
It's also true that per their OBP and SLG, they should have gotten more like 709 (their actual runs created).
It's also true that those 709 RC came in a season that included the following hitters:
- Betancourt - 250 PA
- Beltre - 475 PA
- Balentien + Saunders = 300 PA at 50 OPS+
- Chavez - 80 OPS+ in left field where it don' belong
- Ronny Cedeno = 200 PA at a 33 OPS+
- Bill Hall = 130 PA at a 50 OPS+
The grisly 2009 offense has largely been swapped out for mediocre offense. A 100 OPS+ isn't just feasible; it's what these 14 hitters have done in the last couple of years.
I really didn't know that the offense looked as good as it does.
Cheers,
Dr D
Comments
While I didn't do the full set - I was playing around with team RC versus real runs -- and what I was seeing was a pretty strong pattern of the RC calculation consistently coming out above the actual runs scored. I think the Angels ended up being the closest to matching their RC with their Runs, but I think if you're gonna run a RC vs. Runs comp - you'd want to run the entire league, just to set the landscape.
Depending on which RC tool you were looking at....but most RC tools are league normalized to start with. In other words, mathematically forced to come out with the league run total after all the math is done on the whole league. So no...there shouldn't be a high bias if you're talking about wRC.
...a few points of correction...
1) Marcel does not ignore the bench...PINTO, when he calculates his RC for a team...only uses marcel scores for the starters...but Marcel rates every player for which there is enogh data.
Also, adjusting for playing time is done in Marcel by claculating average PT figures for the player the same way it calculates BA or OPS. Go to fangraphs.com if you want to see an individual Marcel for any 2010 hitter...there are even Marcels off of MLEs for newbies. But you'll see, for example, that Marcel ALWAYS underrates Ichiro, ALWAYS underestimates PT for guys who've had one injury in the last three years, etc. They call it Marcel the Monkey because it's monkey stupid. No thought required. It works because most regular MLB players hit about the same year to year. Marcel is, essentially, a benchmark that all other predictive tools have to prove they can beat by enough that they show real skill...it's like comparing the latest numerical weather prediction model to an old baseline forecasting tool like the Nested Grid Model.
That the Mariners have gone from a deeply unbalanced offense with 4 black holes in the line-up to a balanced, if a bit unsexy, mediocre offense cannot be denied. I have been highlighting this very aggressively in the WAR Stack position player analyses...especially when it comes to the bench players and scrubs. Unless we gegt some unexpectedly horrible performances from bats 10-16, we're going to have one of the most useful and offensively productive benches in all of baseball this year...compared to last year's DISMAL reserve showing which dragged the club down by 62 runs (count the RAA for all the bench bats...-62 is the total). We swapped out almost all of the guys on the roster who produced large negative RAAs and negative WARs and replaced them with a bunch of guys who are above replacement level. That adds 5+ wins right there.
Marcel doesn't use MLEs - if a player hasn't played at all (or enough) in MLB, the projection is just league average. If you want Marcels + MLEs, then you're talking about Oliver.
The point is that the system uses only regression to the mean and an aging curve. As you say, there's nothing complicated about it, and it's not really intended to be a 'real' projection system - it's a useful benchmark to grade the 'real' projection systems like PECOTA or CHONE. Problematically, it keeps beating many of them. CHONE seems to do well v. Marcel, but the little monkey's done very well vs. PECOTA.
Per bbref - here are the RC totals and actual runs produced totals for AL in 2009:
TEAM - RC - Runs
NYY - 1003 - 915
BOS - 918 - 883
LAA - 880 - 872
TBR - 867 - 817
TOR - 852 - 803
MIN - 851 - 798
(avg) 820 - 784 ***
TEX - 812 - 781
CLE - 810 - 773
DET - 784 - 759
BAL - 783 - 743
CHW - 751 - 741
OAK - 747 - 724
KCR - 719 - 686
SEA - 709 - 640
I imagine other sites might use self-updating formula (re-setting based on current total league runs). But, bbref OBVIOUSLY doesn't - being that the average RC comes out 36 runs above the actual average runs scored. (But, even with that - Seattle was 69 runs off, which would be 33 runs short if you do the simple linear adjustment.
They're still using the completely debunked Bill James version of RC...though apparently, when Doc was commenting on the Mariner RC-RS differential...he was indeed referring to this version...so the differences aren't as great as I thought. wRC pegged the Mariners at 669 runs last year...they actually score 640.
I was under the impression that the marcels were using MLEs...but now I do remember reading about Oliver.
And yes...Marcel is statistically defeating most of the projections out there today..it's actually quite humorous.
I'd have thought that the RC roughly balanced out. If it's off by that much, that's unacceptable.
So, absolutely. Subtract -30, -40 runs from the figures given in the main article.
Still leaves you at what, +80, +90 runs. Lemme go see what Matty's post says below :- ) ...
Pinto arrived at the final values by (1) taking Marcel's numbers and (2) plugging 9 players into LAT.
That step 2 is what is under discussion.
Again, the Marcel method is pretty much what the first roto magazines did. And that's fine.
The point is merely that it's an objective set of numbers with no 'spring training optimism' attached.
I'm just fascinated by the amount of impact the worst players on a roster have.
A lot of things in all baseball history fall into line, once you understand that. Inferior managers and GM's get culled from the herd in large part because of this factor.
If you pick just the line-up and score it with a line-up calculator...you won't see the bench affect.
The first thing Martin always did when he got a new club was cut about 8 underperforming players and pick guys who he knew would be average solid from his farm teams. That is how you turn a 70 win club into a 90 win club.