I knew from experience that Ken Warren's frequently hawked modus operandi of using relievers to soak up the innings of your 6^th, 7^th, and 8^th starters had a large degree of merit. Theoretically, you can draft 8 or 10 relievers with ERA below the mid 4.00s, whereas it is very difficult to find an extra three starters with ERAs as low as 4.50. Long ago, I had a run of injuries to my starters and filled in nicely using my fortunately deep bullpen. My team fared better using all relievers than they did with my weaker starters. Still, I've been reluctant to use that strategy as it takes up more picks and you are less likely to find a really good starter worth protecting. Furthermore, if a couple of your core starters are hurt early in the season before you have a chance to beef up your staff with in-season picks, your season could be over very quickly.

This year I had only four starters worth protecting. In the second round I took Gio Gonzalez, but by the third round the only available choices looked quite uninspiring. I hate over-drafting out of need, so I semi-committed myself to the Warren way this year and drafted only one other starter: the emergency-only innings of Brian Bannister.

Life has its ironies. One of my "relievers" has become one of my best starters: C.J. Wilson. Gio is doing well and even Bannister has been a help with Cliff Lee missing the first month. Meanwhile, my bullpen hasn't helped much at all, well, except for Mariano, their collective ERA after the first two weeks is 6.35.

This inspired a little research to see how often relief pitcher picks flop as compared to starter picks flopping. Warren's strategy counts on the all-star effect caused by their being 10 teams in a typical A. L. Scoresheet League holding the American League's 14 teams worth of players plus a few outstanding crossovers. The all-star effect brings a Scoresheet League's replacement level closer to real life league average than that league's actual replacement levels. So, for a pitcher to help your Scoresheet team, he should have an ERA above the A. L. league average, because that should be about the level of your worse pitchers. For Warren's strategy to work then, the ability is needed to draft about three times the number of relief pitchers with ERAs under the league average as finding one starter who is. That is assuming we have enough picks to spend on that many relievers.

To test this, I checked both of my teams' drafts over the last five years. My methods for selecting pitchers have changed enough to make going back any further probably irrelevant. If my pitcher had a real life ERA higher than the league average then he counts as a bad pick. If he didn't pitch the equivalent of 54 relief games or 18 starts, that is if ((G - GS) / 3) + GS < 18, then he was also considered a bad pick - unless he was a prospect who kept his prospect status, was protected at the end of the season, then subsequently went on to have a successful season. All picks who beat the league average ERA and met the appearances requirement were good picks. The tally came to 23% good starter picks and 47% good reliever picks. It was twice as easy to find a good reliever as it was to find a good starter, not three times as easy. You could point out that it was assuredly easier to find a reliever with an exceptionally low ERA - in the 2.00s or 1.00s, but that is probably offset by the lack of relievers compared to starters worth protecting or trading for a draft pick at the end of the year. So, it is a reasonable route to take if you just hate the starters available in your draft, but I don't recommend it as a rule. It is just too difficult to load up on enough good relievers to beat one good starter.