Without adding to Brad’s terrific Castro-Warren analysis, I’d already been mulling over a question about Adam Warren that also applies to several free-agent starters. Assume the Yankees aren’t landing a Johnny Cueto-level ace: how much are they hurt by pitching a 5ish-ERA fifth starter like CC Sabathia (or 2015-quality Ivan Nova), rather than an average-level starter, like Warren or a second-tier free agent like Yovani Gallardo or Wei-Yin Chen? Sabathia and Nova were an identical 1 WAR/yr in 2015, so a 2-3 WAR starter is worth 1-2 more wins over a whole season. I’m not sure it’s worth the roughly $20 million a year (when you include luxury tax), and the 4+ years, it would take to lock in Gallardo’s or Chen’s age 30-34 seasons. But here’s another way to look at it: How much did it hurt the Yankees to have stuck with CC in the rotation, rather than replace him with the clearly superior Warren? I looked at CC’s nine starts after Warren was demoted to the pen (late June to late August), and asked this question: If they’d kept Warren rather than CC in the rotation, and Warren pitched in those nine starts like he did in April-June, how many more games would the Yankees have won?
Answer: Zero. Warren’s April-June was terrific, averaging 2.6 runs and 5.9 innings per start. CC’s nine late June-August starts (the period when Warren was bumped to the pen) weren’t so terrific, averaging 2.9 runs and 5.4 innings per start. So I looked at each of those nine games’ boxscores to see whether, if the Yankees had a typical Warren start (2.6R/5.9IP) rather than CC’s, that would have flipped any game outcomes. It would be silly to assume Warren would give up “2.6” runs in each game, both because you can’t give up 3/5 of a run and because it’s unrealistic to assume no variance, so I assumed Warren would alternate between 5 2/3 and 6 IP, and between 2 and 4 runs in those nine starts. Here’s the outcome:
Basically, if you substitute Warren’s April-June performance into CC’s June-August starts, you don’t get the team any extra wins. Warren’s on-average better performance would have turned one loss into a win, and one loss into a tie score (so basically a coin toss that'd be left to the pen) – a net of 1.5 extra wins. But CC had two really strong starts that won close games; a typical Warren start would've turned one of those into a loss and one into a tie score (another coin toss left to the pen) – a net of 1.5 extra losses. So Warren would have (a) turned one W into an L, (b) turned one L into a W, and (c) turned one W and one L into tie games left to the pen.
That’s a little odd, but not a lot odd. Warren was great as an RP but just a tick above-average as an SP -- maybe worth 2.5 WAR/yr as an SP. One-third of a year of Warren’s 2.5 WAR/yr versus CC’s 1.0 WAR/yr amounts to a difference of 0.5 wins – meaning only a 50% shot that he wins the team even one additional game.
I don't say all this as President of the Rocky Mountain District of the CC Fan Club. I was pounding the drum pretty hard last season to declare CC a lost cause and to demote CC to the pen. But in retrospect, it wouldn't have mattered much.
This is a silly little simulation, but maybe it hints, for those who don’t love stats like WAR, that WAR has it about right: the difference between a below-average starter and a modestly above-average starter is only about one win a year. That’s less of a difference than I would’ve guessed between a decaying husk like CC (or damaged goods like Nova) and a shiny new #3 starter. Which is to say that if Cashman’s winter doesn’t include a new starter like Gallardo or Chen, let’s not break out the pitchforks about a decision not to ink a 4-year, $60 million contract to net one extra win next year.