keyboard_arrow_uptop

Sorry for not having an article Monday, readers. Hurricane Irene saw to that, but everything’s fine and I’m back in action today!

This past Friday, Baseball HQ’s weekly newsletter reached my inbox. At the bottom of each newsletter, HQ founder Ron Shandler usually writes a few paragraphs, riffing on a fantasy topic that’s caught his eye that week. This week, something he said caught my attention, and I thought it warranted a deeper look:

…chasing Wins is a fool's quest and strikeouts can be amazingly resilient because everyone contributes. There is no guarantee that a roster stocked with starters will win any more games than one stocked with middle relievers who have vulture potential. If I'm searching for Wins, I'd rather have high-skilled arms on my team like Sean Marshall, Sergio Romo, Glen Perkins, Matt Belisle, Al Albuquerque or Antonio Bastardo than inning-eaters like Wade Davis, Nick Blackburn, Mike Pelfrey. Jeremy Guthrie, James McDonald or Carl Pavano.

I thought this was an interesting point of view and something that was testable, so today I’ve set out to do just that. In the final month of the season, are high-skilled relievers more likely to win games than low-skilled starters?

The Study
To study this, I’m going to create two groups of players: good relievers and bad starters. To be included in the “good relievers” bucket, a pitcher must have pitched in at least 40 games with no starts over the entire season, pitched in 10 games in September (or non-postseason October), posted a FIP below 3.50 for the season, and accumulated no more than two saves in September (a crude way of ruling out closers, who likely aren’t available to be picked up in fantasy leagues).

To be included in the “bad starters” bucket, a pitcher must have made at least 10 starts over the entire season, made at least five starts in September (or non-postseason October), and posted a FIP above 4.50 for the season. In this study, I’ve included all seasons since 1993 excluding the strike-shortened 1994 and 1995 seasons.

The Results
When I run the numbers, here are the average win and strikeout totals for each of the two groups over the last month of the season:

Group

Sample

W

K

Good Relievers

263

0.77

12

Bad Starters

704

1.72

20

As it turns out, bad starters are still the way to go if you’re chasing wins, especially if you pick ones on good teams and use them for favorable matchups. While “chasing Wins is a fool's quest” is a popular notion among fantasy players, it really isn’t that foolish. As I’ve discussed before, there are definite ways to go about it. Sure, wins have a lot of volatility and give you a wide range of possible outcomes, but if we’re playing the percentages, bad starters are still good bets to accumulate some wins—at least relative to relievers. There is a direct relationship between starts made and wins, so if you want wins and don’t care about anything else, trot out as many pitchers as possible.

As far as strikeouts go, yes, everyone contributes, but a bad starter will still accumulate more than a top-notch middle reliever based on innings. Of course, Ron’s larger point was “protecting categories with downside”—namely, ERA and WHIP.

Two of the categories that most fantasy leaguers don't realize have incredible downside and need to be protected — even in the final week of the season — are ERA and WHIP. With volatile mixes of established and rookie pitchers taking the mound this coming month, there will be a good many pitchers "taking one for the team." Blowout games can make even an apparently safe 0.20 ERA lead disappear in a matter of days.

In this respect, top-flight relievers are naturally a better bet than poor starters. Using our same groups as before, here’s how they stack up:

Group

Sample

IP

ERA

WHIP

Good Relievers

263

12

2.77

1.17

Bad Starters

704

31.7

4.9

1.46

The difference is huge: a full two-point difference in ERA and a 30-point difference in WHIP. And because the starters throw nearly three times as many innings as the relievers, they’re going to have a much larger relative impact on the ratio stats. Still, at this point in the season, most fantasy teams have accumulated well over 1,000 innings, meaning a starter who posts a 4.90 ERA over 30 September innings will probably only budge it somewhere between 0.02 and 0.04 points. Having multiple starters like this can add up, but don’t go crazy if you need a few extra wins and need to employ one. There is always the chance of a huge blowup, but the odds are against it, especially if you play the matchups.

 The flip side of this is that a good reliever will be unlikely to budge your ERA or WHIP. They’ll accomplish Ron’s stated goal—protecting downside—which can be very important, especially if you’re in first place or even just in the championship race. Just don’t look at that shiny ERA and expect it to move your team’s ERA much. Those 12 innings aren’t going to make a dent in the 1,300 innings or so a fantasy team will ultimately muster. We’re talking 0.01 points, 0.02 max. These relievers will protect the ERA and WHIP to an extent, but for each reliever you choose over a starter, you’re likely to lose one win and eight or nine strikeouts (on average). How your league’s standings break down will be what determines the proper move for your team.  

You need to be logged in to comment. Login or Subscribe
abskippers
9/01
But if you're league has an innings cap (but no distinction between starter or reliever innings), wouldn't wins and K's per inning be the most important measure? I'm struggling with that question now because I need wins but also have tight margins on ratios and I'm trying to decide whether to keep Johnny Venters on for September. He's been good for about 12 innings a month w/ 1 win, 12 K's and of course sterling ratios. That's the equivalent of 2 great starts and there aren't a lot of starters who can match his season rate of 1 win for about every 12.5 innings. He's used in the right spots (close games in the late innings). The problem is I need several wins to make up ground.
derekcarty
9/01
Yes, since every league is different, it's impossible to give advice that will apply to everyone -- so comments like these are encouraged :) As I stressed in my first column at BP, context is incredibly important, and that includes league context. If you have an innings cap and you know that you're going to hit it, then W/IP and K/IP are all that matter. In this case, a top middle reliever will be much better than an innings-eating starter like Davis, Blackburn, etc. They will be slightly worse in terms of wins than your top notch starters, but, of course, they come at a fraction of the cost. I wrote an article at THT earlier this year that I'd highly recommend reading because it addresses this exact kind of situation: http://www.hardballtimes.com/main/fantasy/article/predicting-reliever-wins/ It finds which relievers are the best bets for wins and how many you can expect from them. The answer is that a top notch reliever with a good win expectancy is about as valuable on a per-inning basis (in terms of wins) as a starter who wins 15 games over a full season. And these relievers will likely have better ratios and a potentially high K rate. So if you have enough starters/innings to know that you'll hit the cap, Jonny Venters is still an excellent play. Even if you need several wins to make up ground, all that matters is the rate of Wins per IP in a league with a cap. 1 Win from Venters in 12 innings will be more valuable than 2 Wins from Paul Maholm in 30 innings. Also, be sure to take into account that some of the teams ahead of you might be above the innings cap and will stop accumulating them at some point.
jfribley
9/01
In roto leagues this makes sense where the impact of bad starts is muted somewhat, but in H2H leagues that can be catastrophic (like Jake Peavy last night grr..). In this situation might relievers make more sense? Minimize your maximum possible loss while trying to add wins.
derekcarty
9/01
It really depends on your team, your situation, and what kind of H2H league it is (points, roto, etc). It can really become a balancing act in a league where all five categories matter and reset every week. You're not likely to get as many wins from a reliever, but on the week you get lucky and your reliever does get the win, you're going to be in good shape. It's whether you think you can afford to take that risk of not getting a win and getting fewer Ks. And if one of your other starters blows up, those reliever innings are going to be less effective in containing the damage than if you chose an additional starter over a reliever and he did well. It's all about context and balancing risk. For me, in a situation like that, I take the stronger percentage play and let variance fall where it may.
bubba3m
9/01
In your previous article, you have what appears to be a solid methodology to predicting wins for pitchers. Surely it's better than relying on extrapolating a previous year's win total, which is what most prediction algorithms probably use. For this reason, I generally ignore predicted win totals because I assume they're simplistic, or else I make my own blend of combining PECOTA-predicted ERA, IP, and team win total to generate a win estimate. Does PECOTA use a sophisticated algorithm to predict wins for the pitchers, or is it simplistic? Perhaps I'm spending too much time reinventing something that's already there.
derekcarty
9/01
From Colin Wyers: "We look at a pitcher's projected RA (not ERA - remember, unearned runs count toward W-L record) and a pitcher's expected run support, based on the PECOTA-projected runs per game of his teammates. We feed that into a linear model based upon historic pitcher W-L rates and I get expected pitcher wins and losses."
lloydecole
9/01
Great stuff. FWIW, this seems to me like your best column since joining BP, by a fairly wide margin. Thanks.
abskippers
9/01
Derek, thanks for the link to your prior article - very helpful.