Hey, We’re the Replacements

No, not these Replacements.

Nor these Replacements (though I love them):

And definitely not these Replacements.

On our SiriusXM show today, Chris Liss and I discussed Clayton Kershaw’s projection, his rank our cheatsheets using our algorithm and the role of replacement value during his projected absences in determining his rank. I’ve spent plenty of bandwidth discussing Kershaw’s decline, most notably last summer in this space. I’m worried about the decline in his strikeout rate, his fastball velocity, his fastball usage and effectiveness. Most of all, I’m worried about his health, and his lack of innings. My original projection for Kershaw 163 innings, 2.82 ERA, 1.03 WHIP, 12 wins and 173 strikeouts in that span. Such a projection would place him 64th overall in the current rankings, good for 13th among starting pitchers.

I’m satisfied with the ranking – I don’t want to take him as my “ace” in mixed leagues – roughly defined as a top 13-15 starting pitcher, taken in the first three rounds of the NFBC 15-team Main Event. That’s not going to get him, by the way – the current NFBC ADP for him is 33.84, with a range from 15-to-44. So yeah, I’m saying don’t take Kershaw at his current price. There’s too much injury downside, and worse, risk of continued performance decline. Chris raised a couple of objections about the projection and tie-in to the ranking. The one I found to be most persuasive is that the ratios and especially the strikeout rate aren’t downgraded enough to account for the risk. He’s right – the ERA is still well below average at 2.82, as is the WHIP at 1.03 (and slightly lower at 1.04), and the K/9 of 9.55 is actually better than what he did last year (8.65). So an adjustment is forthcoming on those merits.

RotoWire has the best daily fantasy football tools on the web.
Try Our NFL Lineup Optimizer Now

In fact, let’s go ahead and adjust those ratios right now, as it will lead to the next part of the argument. I won’t bore you with the details on how I adjusted him, but to say that his new projection remains at 163 innings, but gives him a 3.15 ERA, 1.12 WHIP, 12 wins and 149 strikeouts. I think that better reflects what a “decline year” looks like for Kershaw. Still well above the average starter, but also worse than before and better representative of what he did over the second half of 2018. For what it’s worth, Steamer’s projection for Kershaw is more bullish on health but similar in performance – 185 innings, 3.23 ERA, 1.12 WHIP, 13 wins and 180 strikeouts. Steamer’s projection would push him closer to the NFBC ADP.

Before we discuss Part II of my conversation with Liss, my favorite underrated Replacements song:

The other part of Liss’s objection was that the ranking using those projections didn’t adequately reflect “Replacement Value” for the periods when Kershaw was out. When a player is on the DL, we’re not left with an open space. We’re allowed to replace him from our bench and/or the waiver wire. That has value, and our ability to select that pitcher should be strong enough to provide positive value.

But how much value is that? What level of pitcher are we choosing from? Who are those replacement level pitchers? I think that the easiest way to determine them is to go by a good set of ADP – which, for me, always is the NFBC ADP results. As we get closer to draft season, it’s probably best to even limit that set to recent ADP, while controlling for the particular type of draft (15-team, 12-team, Cutline, etc…). But with such a limited sample now (currently at 42 drafts), we’ll use the full results. If you do your own projections, or have another set of good projections, you can use those instead, though that will not correspond with the available pool as well – in fact, you’ll have a “better” or at least higher-ranked set of pitchers to choose from.

But let’s say you’re in a 15-team mixed league, with seven-man benches like you’d have at the NFBC. That would mean 135 pitchers would be active (15×9). Let’s also assume each team (including yourself) has at least two more pitchers on reserve. That would raise our total of “taken” pitchers to 165, minus your two, for a total of 163. So to find our replacement pitchers, let’s look at the next 10 – pitchers 164-to-173.

15-team Mixed Replacement Pitchers:

164. Derek Holland
165. Lance Lynn
166. Matt Barnes
167. Matt Strahm
168. Jeff Samardzija
169. Keone Kela
170. Seth Lugo
171. Jace Fry
172. Trevor Rosenthal
173. Lucas Giolito

If you want to filter out the relievers from this group, the next five starters are Clay Buchholz, Wade LeBlanc, Matt Shoemaker, Daniel Poncedeleon and Mike Leake (I’m also excluding Mitch Keller and Jon Duplantier, though both could easily make their debut in 2019).

I don’t know about you, but I can’t hardly wait to use these guys when Kershaw goes down.

Here are the projections for three of the starters in this group.

Holland:

  • RotoWire: 148 innings, 4.56 ERA, 1.43 WHIP, 7 wins, 132 K’s – obviously regressing pretty hard after last year’s nice year with the Giants.
  • Steamer: 89 innings, 4.58 ERA, 1.40 WHIP, 5 wins, 78 K’s.

Lynn:

  • RotoWire: 170 innings, 3.97 ERA, 1.39 WHIP, 11 wins, 159 K’s.
  • Steamer: 138 innings, 4.46 ERA, 1.43 WHIP, 8 wins, 126 K’s. Steamer is more pessimistic than me, which was surprising – I didn’t think my projection was all that optimistic.

Strahm auto-corrects to “Stream” on the Mac, which is appropriate, as we don’t know if the Padres will stick to their intention of using him as a starter. My (RotoWire’s) projection treats him as more of the same as last year, with 48 games and six starts. Steamer doesn’t give him any starts.

Samardzija:

  • RotoWire: 125 innings, 4.61 ERA, 1.25 WHIP, 6 wins, 111 K’s.
  • Steamer: 165 innings, 4.26 ERA, 1.30 WHIP, 9 wins, 134 K’s. My WHIP is probably too low, Steamer’s ERA is probably too low. Perhaps that’s a reflection of a pitcher that’s often defied traditional analysis.

What about 12 team mixed leagues, you ask? Using the same methodology, we’re looking at picks 130-139 (12×9 =108, plus 24 = 132, minus two).

12-team Mixed Replacement Pitchers:

130. Brandon Woodruff
131. Marcus Stroman
132. Pedro Strop
133. Corbin Burnes
134. Zach Eflin
135. Ryan Yarbrough
136. Anibal Sanchez
137. Wily Peralta
138. Anthony DeSclafani
139. Justus Sheffield

Now this list is a lot more interesting. You’ve got a number of young pitchers with breakout potential, an actual closer, and a 16-game winner from last season among your options. Stroman has a classic “Last Year’s Bum” profile, too – one that could recover quicker with a trade out of the AL East.

One can argue that listing these replacements isn’t a fair assessment of replacement value. There’s some validity to that argument. For starters, your list is better than the ADP. Of course it is! Why else would you be playing unless you didn’t think you could beat ADP? Moreover, you’re not picking up these replacements for the full season – ideally you’re selecting them at the optimal time – home starts, favorable starts, two starts, etc…. You’re not going to pick up said replacement level starter in Coors Field or against the Red Sox and Yankees.

Yes, of course that’s true. But that bump is countered by a few other factors. Your opponents also have superior replacement options to ADP, and they’re going to be pursuing the same pitchers in that pool, maybe even using the same reasons to pick up that pitcher. Even if you’ve properly identified the right pitcher, you might get beat out on him, either by getting outbid or on waiver priority. And of course we’re not just looking at replacing your strong-performing but injured ace. We’re also looking at replacing that closer that’s lost his job, or your fifth starter that turned into a pumpkin (I see you, mid-summer Gio Gonzalez) or even a starter that has a Coors Field start. I tend to believe the two sets of factors cancel each other out.

Let’s go back to our 15-team replacement SP’s. Their combined projected ERA (as per RotoWire’s projections) is 4.35, with a 1.36 WHIP, and an average of 134 strikeouts. Let’s assume that you need a replacement starter for eight starts. Adding our replacement stats to Kershaw’s projection now gives us a pitcher with a 3.47 ERA over 210 innings, with a new WHIP of 1.19. The extra K’s and wins that they might add while Kershaw (or any other injured pitcher) help, but enough to move him back up in the rankings?

The good news is that we can test this out. First, our adjustments of Kershaw have moved him farther down the rankings – that’s to be expected when you project a worse ERA, WHIP and strikeout total. But I was surprised just how much he dropped. With the new projection, he drops down to 114 overall. I’m actually uncomfortable with that ranking … but I don’t want to change his projection just to satisfy a ranking preference.*

* Or do I? After all, are we in the business of getting a perfect projection, or are we trying to tell you who to draft and give you a set a competitive bidding values? Fodder for another column and debate, I suppose.

The easiest way to test this is to plug in the numbers of the Kershaw + replacement pitchers and see where the rankings place him. So let’s do that. It turns out, adding those replacement stats help Kershaw’s ranking out quite a bit! He now jumps up from 114 to 88!

So … it kills me to say this, but Liss is right. I do need to find a better way to account for replacement value, at least if I’m going to try to project playing time. In the past I’ve done a set of playing-time neutral rankings, and I think that’s a process I’m going to need to undergo again to get a better feel for just how much playing time matters. In the process of doing this I’ve tweaked a few starting pitchers’ projections, including Kershaw’s yet again, and you’d be surprised the difference just one or two wins make.