clock menu more-arrow no yes

Filed under:

How accurate were last year’s ZIPS projections for the Arizona Diamondbacks?

New, comments

In some ways, remarkably accurate. But while they expected improvement from our pitching staff, they didn’t foresee how much.

National League Wild Card Game - Colorado Rockies v Arizona Diamondbacks Photo by Christian Petersen/Getty Images

The 2018 ZIPS projections for the Arizona Diamondbacks came out last week, and shoewizard beat me to the punch with a nice write-up on them and their Steamer siblings. Rather than re-hash that, I figured it might be interesting to take a look back at the 2017 ZIPS number for the D-backs, and how those stack up against actual performance. As shoe noted, “good projection systems can get hitters “right” about 70-75% of the time and pitching projections barely over 50%. For me the interesting part is the Macro look. This guy may end up over, and that guy may end up under, but at the end the TEAM should be close.” Of course, a good part of that will depend on your definition of ‘right’...

Position players

The chart below has three “blocks” for each player: the 2017 ZIPS projection, 2017 actual production, and 2018 ZIPS projection, with each block containing plate appearances, OPS+ and WAR. The players listed are the ones who were both part of the pre-season ZIPS in Arizona (so, no J.D. Martinez) and who saw significant playing time for the D-backs in 2017. So far, all of those will also be back on the team for 2018, which certainly makes my job easier. Save Martinez, the obvious omission is Chris Iannetta: just for completeness, he produced 2.2 fWAR for the team in 2017. His apparent replacement, Alex Avila, is projected to be worth 1.5 WAR in 2018 - rather more than Iannetta, incidentally, who is forecast for 0.7 WAR with the Rockies.

2017 ZIPS vs. 2017 reality vs. 2018 ZIPS

Player PPA POPS+ PWAR PA OPS fWAR Diff PPA POPS+ PWAR
Player PPA POPS+ PWAR PA OPS fWAR Diff PPA POPS+ PWAR
Paul Goldschmidt 630 143 4.8 665 140 5.3 0.2 638 140 4.1
A.J. Pollock 445 113 3.5 466 99 2.1 -1.6 510 116 3.4
Jake Lamb 517 104 2.4 635 110 2.5 -0.4 589 115 2.5
Ketel Marte 561 91 1.9 255 86 0.9 0.0 599 95 1.7
David Peralta 413 118 1.9 577 99 1.8 -0.9 530 104 1.4
Chris Owings 523 84 1.3 386 83 0.4 -0.6 472 87 0.7
Nick Ahmed 424 69 1.3 178 78 0.2 -0.3 335 77 0.8
Chris Herrmann 250 80 0.6 256 55 -0.6 -1.2 239 72 0.3
Yasmany Tomas 524 104 0.5 180 87 0.1 -0.1 426 109 0.4
Jeff Mathis 152 51 0.1 203 51 -0.2 -0.3 167 42 -0.2
Brandon Drury 581 87 0.0 480 89 1.2 1.2 546 86 0.3
5020 100 18.3 4281 97 13.7 -3.9 5051 102 15.4

This illustrates quite nicely what shoewizard meant by “the Macro look”. Some of the individual projections were some way wide of the mark. The “Diff” column in the middle shows how far - it scales actual 2017 production to the ZIPS-projected number of PA, to remove that from the equation. So a positive number means the player would have outperformed ZIPS, had he got the expected playing time; a negative number means he underperformed the projection. A.J. Pollock, Chris Herrmann and David Peralta were those who failed to deliver at the anticipated level, while Brandon Drury was the team’s top over-performer.

All told, the players were expected to be worth 18.3 WAR, and came in at 13.7. However, a good chunk of the difference is due to playing time going to other people, such as Iannetta and Martinez, with the eleven combining for about 750 few plate-appearances than expected. If you scale the total expected production down to allow for that reduced playing time, you get a projected value of 15.6 WAR, within two wins of the actual figure. Not too bad at the team level, I’d say.

Pitchers

Does the same hold true for pitchers? Again, fortunately, we have almost everyone of the major participants bar Fernando Rodney back in 2018. Although there is a bit of an issue, in that the 2017 ZIPS projections didn’t include anything for Patrick Corbin. He has thus been omitted from the table below. However, one thing which was noted at the time of the 2017 predictions is worth mentioning: “every starter [emphasis in original] in the D-backs’ rotation is projected to produce more wins in 2017 than 2016.” That turned out to be exactly what happened, and was largely the cause of the D-backs’ dramatic improvement last season. Here’s the chart for them (min 30 IP), again with the 2017 projections, actual results and 2018 projections.

Pitchers: 2017 ZIPS vs. 2017 reality vs. 2018 ZIPS

Player PIP PERA+ zWAR IP ERA+ fWAR Diff PIP PERA+ zWAR
Player PIP PERA+ zWAR IP ERA+ fWAR Diff PIP PERA+ zWAR
Zack Greinke 172.3 116 3.6 202.3 149 5.1 0.9 175.7 128 4.0
Robbie Ray 176.3 112 3.4 162.0 166 3.2 0.1 172.0 125 3.9
Taijuan Walker 155.3 97 1.9 157.3 137 2.5 0.6 156.3 103 2.2
Archie Bradley 141.3 99 1.9 73.0 278 2.1 1.1 75.7 134 1.7
Andrew Chafin 65.0 116 0.6 51.3 137 0.7 0.2 59.3 122 1.0
Randall Delgado 73.7 107 0.5 62.7 134 1.2 0.8 66.0 117 0.9
Zack Godley 74.7 99 0.2 155.0 142 3.5 3.1 155.0 111 2.7
858.6 107 12.1 863.6 158 18.3 860.0 119 16.4

Well, that’s a bit different, isn’t it? Overall, ZIPS did almost exactly nail the number of innings pitched by these seven men: it was off by less than 0.6% in total, which is quite impressive. But even with the note mentioned above, about all the pitchers performing better than in 2016, ZIPS still managed to underestimate the production of each and every one, after adjusting for innings actually pitched. Robbie Ray was the closest: his actual figure fell just short of ZIPS, but that was more than balanced out by the innings shortfall (and in Ray’s case, this was not the result of performance, but due entirely to the line-drive off the head which caused Robbie to miss four weeks).

Hopefully, Dan Szymborski might pop in and give us his thoughts on this. There surely can’t have been many times an entire pitching staff was projected to improve AND then outperformed even those expectations. All told, those seven ended up producing more than 50% over the ZIPS forecast, as noted, without any significant uptick in innings pitched. Zack Godley alone was responsible for about half of the difference, and his forecast for 2018 is, naturally, much improved. Most of the pitchers are expected to regress from the actual 2017 level: Ray and Andrew Chafin are exceptions, but that may be because I believe ZIPS WAR is closer to bWAR than fWAR. [If only I’d realized that before putting together the above tables...]

This regression is no surprise, and is part of the nature of most projection systems. When a player’s figures take a great leap forward in Year X, there’s inevitably going to be a lag before projections catch up, because they also factor in the lesser numbers from Year X-minus-1, etc. Is what we saw from Godley, Ray, Zack Greinke - indeed, virtually the entire D-backs’ pitching staff - going to be “the new normal” for them in 2018 and beyond? That’s why they play the games: fingers crossed, from a fan perspective, that it is indeed the case.