The March of WAR: Batting

by Doghouse on Feb 7, 2009 5:48 PM EST

The first big (and fun) chunk of projecting 2009 is offense. How well do we think our boys will be smacking it this season? As a starting point, I've got excerpts from two projections of player offense, plus 2008's performance. Take a look and give me your judgement on whether the numbers look high, low, or about right. Stats and comments after the jump!

Two of the big systems for projecting player perfomance are Marcels (which is a simple projection of historical performance that weights recent seasons more heavily) and CHONE (which is, uh, more complicated). Here are the projections for the people we've identified as our main players, along with what they did last year: # of plate appearances, batting average, on-base percentage, and slugging.

This is CHONE:

			PA	AVG	OBP	SLG
Alberto Gonzales	241	0.263	0.315	0.397
Anderson Hernandez	568	0.251	0.301	0.343
Austin Kearns		571	0.252	0.350	0.413
Cristian Guzman		508	0.287	0.331	0.407
Dmitri Young		441	0.273	0.347	0.429
Elijah Dukes		466	0.263	0.373	0.462
Jesus Flores		387	0.235	0.292	0.377
Josh Willingham		514	0.263	0.362	0.458
Kory Casto		498	0.243	0.331	0.386
Lastings Milledge	505	0.279	0.350	0.435
Luke Montz		459	0.219	0.292	0.358
Nick Johnson		519	0.275	0.420	0.477
Ron Belliard		504	0.272	0.331	0.419
Ryan Langerhans		351	0.243	0.350	0.379
Ryan Zimmerman		569	0.296	0.364	0.488
Wil Nieves		344	0.251	0.297	0.350
Willie Harris		380	0.254	0.345	0.386

CHONE is the most sophisticated model I've looked at. Its projections assume a neutral ballpark and league (neither hitter-friendly nor pitcher-friendly). This will skew the data a little bit, but the single season of data on Nats park says that it's close to neutral. I like CHONE because it say NJ will last almost the whole season! Yay! It's forecasting everyone will be durable except Dukes and Flores. CHONE says Hammer, Zimmy, Dukes and NJ will be tearing it up in 2009. Lasto and AK won't be bad, but Flores might disappoint.

This is Marcels:

			rely?	PA	AVG	OBP	SLG
Alberto Gonzales	0.34	258	0.264	0.329	0.398
Anderson Hernandez	0.36	246	0.270	0.333	0.396
Austin Kearns		0.84	446	0.253	0.341	0.402
Cristian Guzman		0.76	525	0.304	0.345	0.439
Dmitri Young		0.74	341	0.282	0.352	0.436
Elijah Dukes		0.68	389	0.255	0.362	0.453
Jesus Flores		0.67	382	0.262	0.319	0.412
Josh Willingham		0.84	468	0.262	0.355	0.453
Kory Casto		0.49	297	0.233	0.303	0.353
Lastings Milledge	0.78	514	0.274	0.339	0.418
Luke Montz		0.1	213	0.261	0.338	0.420
Nick Johnson		0.69	274	0.266	0.394	0.464
Ron Belliard		0.83	424	0.274	0.333	0.420
Ryan Langerhans		0.7	294	0.225	0.327	0.372
Ryan Zimmerman		0.86	505	0.282	0.345	0.456
Wil Nieves		0.51	305	0.247	0.298	0.364
Willie Harris		0.76	451	0.256	0.337	0.397

Marcels (named after Marcel the Monkey from "Friends" as dig/boast about its simplicity) is more-or-less a straight historical projection. That first column labeled "rely?" is a rough measure of how good the projection is (closer to 1.0 = more reliable). Notice that most of the younger players have low numbers here, because they have less data to look at. Marcels says that NJ is going to break by the AS star game--boo! It also has Zimmy, Dukes, and Hammer doing well and figures Jesus will do a good bit better than CHONE says. AK and Lasto are looking at "meh" seasons.

And this is the actual batting data from 2008:

Player			PA	BA	OBP	SLG	OPS+
Alberto Gonzalez	54	0.347	0.407	0.531	144
Anderson Hernandez	91	0.333	0.407	0.383	109
Austin Kearns		357	0.217	0.311	0.316	65
Cristian Guzman		611	0.316	0.345	0.440	104
Dmitri Young		180	0.280	0.394	0.400	109
Elijah Dukes		334	0.264	0.386	0.478	125
Jesus Flores		324	0.256	0.296	0.402	81
Josh Willingham		415	0.254	0.364	0.470	118
Kory Casto		182	0.215	0.297	0.313	60
Lastings Milledge	582	0.268	0.330	0.402	91
Luke Montz		26	0.143	0.308	0.286	57
Nick Johnson		147	0.220	0.415	0.431	123
Ron Belliard		336	0.287	0.372	0.473	120
Ryan Langerhans		137	0.234	0.380	0.396	104
Ryan Zimmerman		466	0.283	0.333	0.442	101
Wil Nieves		191	0.261	0.309	0.341	70
Willie Harris		421	0.251	0.344	0.417	98

There are a lot of players there with few PA (remember all the injuries?) and some crappy hitting lines (I'm looking at you, Austin Kearns!). Dukes and Willingham showed some flashes with the bat (between injuries) though CHONE and Marcels have them penciled in to fall back a bit. Marcels figures Guz will repeat last year's mad batting tear, although CHONE says he'll regress (Never! LASIK FTW!). On the bright side, both systems project Zimmy, Lasto, and AK to improve their 2008 numbers (although in AK's case, he could hardly have been worse, and he and Lasto will only improve up to roughly league-average).

Anyhow, faced with this wall o' numbers, what do you think? Where do you think the numbers are high or low? Who'll have a good season, who'll have a bad one? Who'll get hurt?

Tweet Comment 12 comments | 0 recs |

Do you like this story?

Comments

As you may have expected

I’ve got a boat load of questions.

1. Define neutral league. Is a league neutral if the majority of the ballparks in which games are played in THAT league are neutral?
2. How long have these CHONE / MARCEL predictions been around?
2a. Have they been shown to be reliable projections? The reason I ask this is because if not, then, why would I waste my time comparing whether or not I think the numbers are high or low?
3. What data should I be using to determine whether or not the projection numbers are too high / too low / just about right? Or am I just using a gut feeling?

I’m actually interested in all of this because on AA (Amazin Avenue), people are spewing this data out as if it were there own personal information such as height, weight and birth date. Sometimes, if you don’t know what they’re talking about, you really can’t be part of the conversation. So, I’m actually interested in what’s going on here.

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 17, 2009 4:42 PM EST reply actions

Answers!

1. This is a little voodoo-y to me. I think the idea is that the NL is easier for hitters and the AL is harder, just like the some ballparks are friendly to pitchers or catchers. In theory, that means we should “correct” the values to reflect how hitting-friendly the player’s league and home ball park are. I’m not a serious enough stat-head to try to do this, plus the early estimate for Nats Park “ball park factor” is pretty close to neutral. I think we’ll end up adding in this “correction” without even thinking about it as we decide how reasonable we think these projections are.

2. Marcels has been around since at least 2001, and I think CHONE and ones like PECOTA have all been around at least a few years, too.
2a. “That depends.” The interesting part is that, on average, they’re all pretty close to each other. They all start with a player’s historical performance, then stir in various blends of secret herbs and spices to account for aging, changing ballparks, the phase of the moon, and who knows what-all (a lot of these are subscription products—I think Marcels is the only one where the guy who runs it shows his work publicly). They generally predict league averages pretty well (say, within 10-20 points of OPS), and individual players a bit less well (say, within 60 points of OPS). This leaves a good bit of slop to think a system’s projection might be high or low. Why consider them at all? Well, smart people came up with them, and it gives us another point of comparison besides what happened last year.

3. You should use whatever data you think are appropriate to make your judgement. What do we do as fans? We look at last season, or maybe the career average. We think about whether a player faded at the end of the season or seemed to be getting stronger. Is he coming off an injury, or going through stuff in the offseason—will he be distracted? Have we heard he’s working out every day? How old is he? Do we think he’ll fall back toward “average” after a hot/cold season last year? The guys doing the projections have come up with mechanical ways to add up a bunch of that stuff and make a prediction. But the casual fan notices plenty over the course of watching games and following a team. Your “gut” picks up a lot of stuff subconsciously.

People tend to throw around numbers out of mathematical models like TRVTH! carved into a granite block—just because what you get out is a number (I deal with this fallacy all the time in my day job, “The model said so! It must be true!”). But any of these systems is just a structured way of making guesses (in other words, something that adds up a pile of assumptions). Of course, the assumptions behind the guesses might be good ones made by people who know way more about baseball than we do. However, anything that has to work on all of MLB is not necessarily going to catch every special case or rookie without much history. The guys doing the systems know more about baseball and statistics than we do, but we (as a group) probably know more about the Nats. That’s what I want to get from you guys. Take a look at previous seasons’ performance, take a look at the systems (realizing that a player’s OPS will be an average of 60 points different from the projection), think about what you’ve heard in the news and seen at games, check your gut and make a pick (do as many of these steps as you feel comfortable doing, as long as you do the last one!).

And please keep asking questions. I am not some sabremetrics savant—I’m learning it as we go along, too!

"Next year we'll make it better." -- Mannyger Acta

by Doghouse on Jan 17, 2009 6:26 PM EST up reply actions

Wow! By far one of the best responses I've ever read on SB Nation as a whole.

Thanks for taking the time to do that. Time for me to start doing some homework then.

Follow up question. Is there any difference between PA (Plate Appearance) and AB (At Bats)? If not, why are people using PA instead of AB and who was the jerk that invented another term for the same thing?

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 17, 2009 6:51 PM EST up reply actions

Wait a minute. Maybe I can answer my own stupid question.

Scenario:

Two outs, Milledge on first, Guzman at the plate. Guzman has a 3-1 count, Milledge attempts steal and is thrown out. Milledge is the 3rd out so Guzman gets another plate appearance in the next inning.

So, that’s one example of a PA vs. an AB. But now I have another question. If the PA resulted in nothing other than a few more pitches thrown by the pitcher (perhaps), why is PA important?

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 17, 2009 7:06 PM EST up reply actions

PA vs AB

“”http://en.wikipedia.org/wiki/Plate_appearance" target="new">Plate appearances" are essentially every time a player goes to the plate. “At-bats” are “Plate Appearances” minus walks, hit-by-pitch, and sacrifices. The reason for the difference is that batting average is calculated by dividing AB into the number of hits. If you included walks in the number of AB, you’d penalize the batting average of people who are good at getting walks—like our own Nick Johnson. The same logic applies to HBP and sacrifices—these are good outcomes (you either get on base or get an RBI), so you don’t want to punish a player’s batting average by adding them to the number of AB.

Note that batting average and slugging percentage are both calculated with AB, but on-base percentage is calculated with PA. This can lead to funny situations where a player with more sacs than walks might have an OBP lower than his batting average (as our own Cristian Guzman did for a while early in 08)!

Also, in the example you give above, Guzman would get a single PA (and a single AB—unless he walked, sacrificed, or was HBP). I believe AB/PA only count as one from the time you come to the plate until you either reach or make an out. This brings up the question of what happens is someone is CS for the last out of the game—does the guy who’s up get an AB/PA? I would imagine not, since he didn’t make an out or reach base; it’s probably just dropped from the stats. Ah, baseball, a game of nuances! (Which reminds me, Wikipedia points out that reaching due to defensive interference does not count as an AB, but it is included in PA.)

"Next year we'll make it better." -- Mannyger Acta

by Doghouse on Jan 17, 2009 8:04 PM EST up reply actions

Brilliant! I will have to read this more than a couple of times to have it SQUARELY under my belt.

O.K., now for what you called me in for. My batting projection methodology / comments. Here is what I did.

I used your charts and compared:

AVG / OBP / SLG from both CHONE & MARCEL against CAREER AVG’s. (I left out PA because I was standing by for the answer and because I wanted to get to work on this. So, PA was not a factor in my analysis. I hope that’s o.k.).
I then compared CHONE & MARCEL against each other.

This is what I’ve learned:

1. Using the lineup that “I” think will start (see below at end of post), the Nats have a pretty young (but not too young) group of batters. The average age for “my” lineup is 26.4 yrs.

2. On just about every occasion both CHONE and MARCEL project the batters to perform much better than their career averages. I have to agree with this projection mostly due to the young age of the players. It’s my assessment that players get better with age up to about 31/ 32. So, based on improvement for players in this age group, I don’t believe the projections are TOO high.

3. On a 5:3 ratio, MARCEL projected my lineup to perform lower than CHONE. What this tells me is that, whatever data MARCEL is using, it favors players projected performance less than CHONE. Long story short, 5 out of 8 times MARCEL projected lower performance numbers than CHONE.
For Example: Josh Willingham
CHONE:…..AVG…..OBP…..SLG
…………….263…….362…..458
MARCEL….AVG…..OBP…..SLG
…………….262…….355…..453

“My” starting lineup:
1. Milledge (CF)…………….Dukes
2. Dukes (RF)……………….Kearns, Milledge
3. Willingham (LF)…………Dukes, Mo Pena
4. Zimmerman (3B)……….Gonzalez
5. Johnson (1B)…………….Belliard
6. Guzman (SS)……………Gonzalez
7. Hernandez (2B)…………Harris
8. Flores ( C )……………….Nieves
9. Pitcher

If the CHONE / MARCEL projections hold up, the Nats are definitely due to improve. Difficult to say by how many games. I’m just going to throw a number out there – 15 game improvement max. I would project a higher improvement, but I just think that the NL East is a little too competitive and that will hinder the Nats. I really believe the NL East is like the AL East. They (NL EAST / AL EAST) have a higher number of quality teams in their respective divisions. Also, I just can’t go higher than 15 better because of the Nats rotation. Sorry “E”.

Well, this was my first attempt to use my limited analytical ability with CHONE/MARCEL data and actual numbers. I’m ready for my grade professor!

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 17, 2009 8:53 PM EST up reply actions

Fine job, Lou, you pass

…not that this goes on your SBN transcript, or anything. See, you looked at the projections and the records, used your fan “instincts” and figured the projections looked about right. Don’t worry too much about converting that into actual wins—that’s my job (actually it’s my job to plug the numbers into a spreadsheet someone else made and then it spits out the wins… anyhow…).

"Next year we'll make it better." -- Mannyger Acta

by Doghouse on Jan 18, 2009 10:13 AM EST up reply actions

(Whiping sweat from brow) Whew!

Splendid. I look forward to seeing how others fared and I also look forward to seeing how it compares to the real thing at the end of the ’09 season. BUT!…I will still savor every sweet day of the regular season.
Cheers!

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 18, 2009 10:18 AM EST up reply actions

Take a look... your lineup sez 75 wins...

I’m not going to do this for everyone’s lineup individually, but here’s an example of what happens when you plug it into the spreadsheet. That’s with the Marcels numbers. As you noticed, CHONE would give a little more production. There are some adjustments in there that we haven’t talked about yet (baserunning and fielding), but we will! Also, it has my totally half-assed attempt at a pitching rotation with CHONE projections for ERA (we’ll get to that later this week, too).

The “win talent” box on the right is the projected season total of wins, and there’s a little table of the probability the team will get to a certain number of wins, too. According to Lou, we’ve got a 14% chance of making .500! Woo-hoo! (No, this doesn’t take into account how many times we have to play the Mets and Fillies—I imagine the guys at BtB will do something with that when they put all the team spreadsheets together to project standings.)

Also for your amusement, check out these lineups. This was more an attempt to get numbers up so we can argue about them than a go at serious lineups, so don’t concentrate too much on who I’ve got playing for how long. The three sheets are Marcels batting #s, CHONE #s, and a straight carryover of 2008 #s. The pitching is the same in all three (CHONE #s, half-assed rotation).

"Next year we'll make it better." -- Mannyger Acta

by Doghouse on Jan 18, 2009 10:39 AM EST up reply actions

Wow!

Based on batting alone I was awfully close! 59 wins last season plus my projection (for ’09) of 15 max = 74 (projected). Pretty neat! Maybe I need to head on up north and apply for a job on Wall St.?

When I get to work today I’m going to print out both sheets and bounce them off one another. This is actually more interesting than I thought it would be. (Don’t worry Buddha. I’m not going to turn into a stat geek, but I will remain a blog geek.)

" PLEASE! CHANGE THE PATCH! "

by LOUtheMETandNATSfan on Jan 18, 2009 12:16 PM EST up reply actions

Snap judgment time!

In order:

- Don’t care about Gonzo; differences aren’t worth discussing.
- Maybe a .260 / .315 / .370 line for Hernandez?
- I like CHONE for Kearns.
- Guzmania for Marcel; I think the improvements in hitting are real and will last through this season as well.
- Don’t care about D’Meathook, although if he sees that many PA in either projection that’s not a good sign (unless he picks up 100 PA as the first bat off the bench).
- No opinion on Dukes.
- Flores …amplify CHONE a bit; maybe a .250 / .305 / .400?
- No opinion on Willingham; that’s about what I’d expect from him.
- God, I hope that Kasto doesn’t get that many PA.
- I like Milledge; go with CHONE and add .010 / .015 / .025.
- Luke Montz? Um …. CHONE? Man, that looks nasty.
- CHONE for NJ, but 350 PA. Split the difference.
- The rest of the guys look about right.

Generally speaking, if there’s not a lot of difference between CHONE and Marcels, and they both have a high confidence level, that’s a good sign that I’ll have to be talked out of believing them. Anyone want to convince me otherwise?

by Chris Pendley on Feb 9, 2009 7:08 PM EST reply actions

Thanks, Gray! I want to revisit the batting stats, and I appreciate the input.

Take a look at the playing time post, too. It has our current more-or-less-consensus on how many PA people will end up with and where. As always, the latest info is in the on-line W.A.R. spreadsheet (click the “WAS” tab on the bottom… no, I don’t know how to link directly to the Nationals sub-sheet, and I’m not sure it’s possible).

"We’re all neighbors in NatsTown™!" --NNN

by Doghouse on Feb 9, 2009 7:22 PM EST up reply actions