Wow. With that kind of closeness, I think it’s fair to say the big factors are going to be the things you can’t really predict in the offseason: major injuries, unforseen skills implosions, or mid-season pickups.
I’ve used a few different projection systems to approximate these things. Each system I’ve looked at has put the Mets about 4-5 games behind the Phillies except for CHONE seems to have the Mets about 8 games behind. That seems like an outlier. This is all before Lowe joins the Mets. Here are the numbers that I’m seeing. I didn’t include defense, because I assumed that the projections for pitcher ERAs and RAs were supposed to reflect the defense that was behind them last year. Regardless, with the Phillies better than the Mets at defense by your numbers, that should only help my argument that the Phillies project better than the Mets as currently formed. I’m not weighting the bullpen by leverage, so that could be another source of mistake.
I’m factoring in SB/CS but not other measures of baserunning. I’m not sure if baserunning is factored in your analysis or your dollar value of players, but it should be. There seems to be nearly as much variance among player contributions in baserunning as there is in defense if you believe Baseball Prospectus’ EqBRR.
When I plug in the marginal lineup value compared to a replacement level player at their respective positions (I reverse engineered replacement level their previous years’ VORP to do this), I get the following numbers for a few projection systems:
The average is a little inflated from when I did this with runs scored. I adjusted the Phillies and Mets replacement level RS so that their Pythagorean record would be 45-117 if they had 6.00 and 5.50 RA respectively, and got that the Phillies beat the Mets by 4-8 games depending on the projection system. The real difference in between how we each did it was that you seem to give the Mets starting pitching a whole lot of credit. Did you adjust your wins above replacement for the fact the projections used reflected different parks? I assumed the Mets pitching projections were for a stadium with similar park factors as Shea. Otherwise, I don’t see how Maine, Pelfrey, and Niese look as good as they do in your numbers.
Matt, cannot speak for how you arrived at your numbers, but the way we tend to calculate WAR may differ from yours.
First, we don’t use RA or ERA. We use FIP. Starting Pitchers are compared to an FIP of 5.50 up to 150 IP, and after 150 IP, to a 4.60 FIP for a reliever. And then we throw in a little bonus, a few runs for pitchers who log over 180 or so innings because they are keeping the bullpen fresh, preventing replacement relievers from entering.
So, a 3.50 FIP pitcher in 200 innings would give up 78 runs. The replacement level would give up ((150*5.5)/9) + ((50*4.6)/9) = 118 runs. Therefore, this pitcher is +30 runs above replacement. With 200 innings, I would give him another +3 runs. This makes the pitcher +3.3 wins.
As far as the Mets new park, we really have no idea how everything will work there, so I wouldn’t want to adjust something that we are unsure about yet. I could see it depressing their pitching by 1-1.5 wins if it isn’t as favorable of a pitcher’s park, but at the same time, that difference could be made up for in the offense.
You also have Knight in there and Niese without a lot of innings. I put in Niese at a 4.45 FIP in 150 innings and Derek Lowe, who adds over 3 wins himself.
I also don’t use Bill James Projections because they tend to be too offensively-happy. I do use EqBRR but there is only 2 yrs of data at BP and so I do not want to use something with such a low level of reliability at this point. With just 2 yrs for that, why not estimate it based on what we see, you know? That could be a source for our differentials. Here are my player by player projections:
Howard: +30 wRAA, 0 UZR, +10 adj, +3.9 WAR
Utley (with injury): +20 wRAA, +12 UZR, +18 adj, +5.0 WAR
Rollins: +15 wRAA, +7 UZR, +28 adj, +4.9 WAR
Feliz: -11 wRAA, +12 UZR, +20 adj, +2.0 WAR
Ibanez: +9 wRAA, -13 UZR, +17 adj, +1.2 WAR
Victorino: +5 wRAA, +5 UZR, +22 adj, +3.1 WAR
Werth: +16 wRAA, +12 UZR, +12 adj, +4.0 WAR
Ruiz: -8 wRAA, +4 def, +24 adj, +2.0 WAR
The bench is highly variable but I have them at a combined +1.5 WAR.
Hamels: +4.0 WAR
Myers: +2.1 WAR
Moyer: +1.4 WAR
Blanton: +2.7 WAR
Happ: +1.5 WAR
Delgado: +18 wRAA, -6 UZR, +9.6 adj, +2.1 WAR
Castillo: -6 wRAA, 0 UZR, +18 adj, +1.2 WAR
Reyes: +14 wRAA, +5 UZR, +30 adj, +4.9 WAR
Wright: +40 wRAA, +3 UZR, +25 adj, +6.7 WAR
Murphy: +11 wRAA, 0 UZR, +10 adj, +2.0 WAR
Beltran: +28 wRAA, +6 UZR, +23 adj, +5.6 WAR
Church: +4 wRAA, +3 UZR, +10.4 adj, +1.8 WAR
Schneider: -12 wRAA, +4 UZR, +25.5 adj, +1.8 WAR
Santana: +4.5 WAR
Maine: +2.1 WAR
Pelfrey: +2.8 WAR
Lowe: +3.7 WAR
Niese: +1.6 WAR (150 IP at 4.45 FIP)
Also, your bullpen numbers seem incorrect. There is no way Clay Condrey is worth +1.2 wins, or that Romero and Madson are worth the same as Lidge. This is likely due to not including leverage. Again, small differences, but very interesting how it sways one set from a +2-3 advantage for the Phillies to +2 wins for the Mets.
Why does the value of the replacement player matter when compaing teams head to head? I gies the question I’m asking is why the Phillies lineup which you state with offensive and defense accounted for is +120 only worth 26 WAR when the mets lineup contributes +112 runs is worth 26.1 WAR. Seems like the phillies would be ~.7 wins ahead if 10 runs approximates a win. Thanks
I wasn’t using my own projections. . Those also have the Phillies about 4-5 wins ahead before adding in Lowe which would close the gap to 1 win or so. But I figured I would only use reliable sources.
You used FIP and then made adjustments for defense. I used projected ERAs by each of those four sources, and did not include adjustments for defense. If those four sources accounted for defense effects on RA/FIP differential as well UZR. Tom Tango, Bill James, Sean Smith, and Ron Shandler are all aware of FIP and I’m sure they did their RA/ERA projections trying to proerply account for defense.
I have a methodological disagreement with how you do your WAR for starters. It works in isolation to determine value of a pitcher, but when you then also count the relievers’ WAR, you are double counting the effect of innings being pitched by relievers.
The way that I do it is mathematically equivalent to assuming that the difference in between 1450 IP and the total innings of the 12 pitches listed would be 6.00 RA or 5.50 RA depending on which park.
Keep in mind that FIP includes park effects– QERA does not, but FIP includes HR projections which are based on park effects. I guessed that each of those projections that I named did their projections as though the Mets were staying in Shea. If they are using weighted averages with age adjustments (which I think is all Marcel does), then they pretty much are using projections of the Mets in Shea.
Bill James does tend to be offensively biased when it comes to offense, but he tends to be pitcher biased when it comes to pitchers. His numbers don’t add up. But this is irrelevant when comparing two teams. Unless he persistently biases two kinds of players differently– say, by not reverting superstars towards the mean enough– then his projections are an additional piece of data.
Your Hamels projection seems low, but so does your Santana. Certainly a source of bias is that Bill James and Ron Shandler both have Hamels markedly better than Santana.
My innings projections of Niese and Knight don’t really matter because they are replacement level and I have their non-innings being replaced by similar quality pitchers. If you’re using FIP, I’m confused how you get Pelfrey higher than Moyer, Myers, and Blanton. Niese seems wrong too. 1.6 WAR? that must be because of park effects.
I really think the main difference is that you are using FIP as though it is the same for both teams, when the projected FIPs are based on different HR effects. Even if Citi is neutral, CBP is not.
Condrey seems high at 1.2 wins for Marcel, but Marcel has him 65 IP at 4.43. That seems wrong, but that’s why I used 3 systems to check– all of which had him lower but others higher.
I agree that there is a problem that the bullpens are not weighted by leverage, but if the two teams are equally affected by this– as they are similarly talented and similarly distributed– that shouldn’t affect things. The real difference really seems to be rotations, and I think it’s largely HR where the difference lies. I think all four projection systems I’m using in my methodology are projecting the Phillies and Mets as having comparable lineups, benches, and bullpens (with the exception of Chone giving more credit to the Phillies), but the way that you are doing it gives the Mets rotation a boost and the Phillies less of a boost.
It’s just a better baseline. The Phillies lineup would log less PA than the Mets in these projections, likely because Utley is only being counted at 130 GP and 520 PA. Before adjustments, the Phillies look ahead close to 0.7 wins, but with the adjustments for playing time and value above replacement, they are virtually identical. The difference is that Utley’s time would be made up for by the likes of Bruntlett, Donald, etc, whereas someone like Reyes or Wright would get the whole porton of playing time. That’s all.
Yeah, it’s a methodological disagreement that is causing the differences. But, as you said, with Lowe, your projections call for a 1-win difference between the teams. That could easily tilt either way, so we’re very close including Lowe.
As far as Pelfrey vs. Myers/Moyer/Blanton, it shouldn’t be confusing since you know the methodology Dave and I use here.
Pelfrey projects to an FIP between 4.05-4.10. Myers and Moyer are markedly higher than that, and Blanton is virtually equal to that.
Niese at 1.6 comes from a 4.45 FIP in 150 IP. That equates to 75 runs allowed. A 5.50 FIP in 150 IP, or replacement level, gives up 92 runs.