Bill James on starting pitchers

[brett]

Bill James has claimed that it has been statistically proved that teams would be better off today if they let their starters go deeper into games-and that possibly if they pitched on a 4 man rotation.

My understanding is that he has argued this based on the fact that relievers tend to be substandard pitchers who get better than average ERA+'s because they face batters few times, and enter innings with 1-2 outs already on the board but no responsibility for inherited runners.

It has been shown that a starter moving to a relief role will tend to have a relative ERA 25% lower just because of these factors (times facing a batter, and entering innings with favorable conditions toward ERA). This would mean that a 130 ERA+ starter would typically become a 170 ERA+ reliever. This is based on actual pitchers moving from starting to relief roles.

What I don't understand is, why does this data mean that teams would be better if they let their starters go deeper into games? It looks like just changing your pitcher helps to hold down the offense of the other team even if the relievers are not as good as the starters.

[SABR Matt]

Agreed brett.

Relievers have higher K rates, lower walk rates, lower HR rates, lower BABIP, and lower LD% historically. Why should we use the SP MORE? I think we shuold use the SP less per game and go to a 4-man rotation. 90-100 pitches instead of 100-110 pitches per start saving enough work so that the pitcher can get an extra five-six starts.

[Honus Wagner Rules]

Agreed brett.

Relievers have higher K rates, lower walk rates, lower HR rates, lower BABIP, and lower LD% historically. Why should we use the SP MORE? I think we shuold use the SP less per game and go to a 4-man rotation. 90-100 pitches instead of 100-110 pitches per start saving enough work so that the pitcher can get an extra five-six starts.

Taking this idea a bit further how about not having traditional starters at all? Have 3-4 "relievers" pitch every game. I seem to remember Tony LaRussa tried this once with the A's.

[SABR Matt]

That ruins the advantage relievers have...that being that they can mathc-up against the favorable side of their platoon split, throw at max effort and be used in high leverage situations multiple days in a row. I think a starter going 5-6 innings is still required to make the bullpen edge work...but going with 4 starters and 8 relievers might be better.

[digglahhh]

I think a part of this is that relievers' rate stats are artificially lowered by the times in which the come in an surrender runs that are charged to the SP.

Aaron Heilman, for example, has not only been terrible to the tune of god-knows what ERA, but he has allowed something like 3/4 of his inherited runner to score.

In turn, that increases the SP's ERA. I presume part of the reason SPs would have better numbers, regardless of their quality, as relievers is the number of partial innings they'd get as a function of the role.

If you come in with an out, or two, you still don't get charged a run until a guy you let on scores (you don't even get charged for guys who reach on FCs as they take the place of the original Ps runners in terms of liability). So, teams have fewer chances to score a run off YOU. Those partial innings add up over the course of the year to amount to free innings, basically.

Here's an exercise: look at relievers' ERAs in innings they start as opposed to innings they enter partially completed

[SABR Matt]

The reason SPs get lifted for relievers though is that managers can see them losing effectiveness and getting into trouble.

BTW the relievers' strand rate is better than the average strand rate for starters, so no...the relievers aren't hurting the starters any more than they'd hurt themselves.

[Tango Tiger]

digg: the component numbers are much better for relievers than starters. This has nothing to do with the runners inherited charged to starters.

In The Book, I advocated that the #4/#5 starters be paired so that they each relieve each other (obviously not back-to-back games). You get the relief advantage, plus you get an extra pinch hitter. You can easily extend this to a: 3 man, two group rotation, where the group would be the #4,5,6 starters, each pitching 2 innings relieving each other.

[StillFlash]

So far this year, teams seem to be getting some extra starts to their #1 and #2, some guys are on pace for 36 starts.

If a pitcher goes every five days (not games), he can make 36 starts, and those are fewer starts that have to be made by the #5, and there's usually a huge gap in quality from 1 or 2 down to 5 or 6. My 5th starter would be a swing guy who gets 18-20 starts and relieves in between.

[Ubiquitous]

Generally speaking the first month of the season those first couple of starters get some "extra" starts due to scheduling.

[brett]

Taking this idea a bit further how about not having traditional starters at all? Have 3-4 "relievers" pitch every game. I seem to remember Tony LaRussa tried this once with the A's.

My question about this has to do with the variance in runs allowed per game.

For the stat guys, do pitchers runs allowed per game follow a normal distribution, or is it bimodal with basically a set of "good" games and a set of "poor" games?

If you plan to use 4 guys, and the distribution is bimodal then you have 4 chances to catch just one guy on a poor night and have one big inning.

It seems to me that using a 4 man rotation where the starters get to work through the lineup 2 times would be optimal because a lot of pitchers drop off the third time through a lineup. The problem is that this would put guys just below the 5 inning mark most of the time.

Anyway, then if the game is close, you bring in your best reliever right then-a guy like Gossage who may go through the lineup once or even twice.

I think that managers really have a fear of blowing a 4-5 run lead though and are less likely to put a lower level pitcher in in those games than they should.

[digglahhh]

The reason SPs get lifted for relievers though is that managers can see them losing effectiveness and getting into trouble.

BTW the relievers' strand rate is better than the average strand rate for starters, so no...the relievers aren't hurting the starters any more than they'd hurt themselves.

Fine. I stand corrected. So, it's not artificially inflating the starters' ERAs, generally speaking.

But, it does put relievers in positions where they have to get fewer than three outs before allowing one of their runners to score. Therefore their ERAs may look better than they otherwise would.

[leecemark]

--Regardless of whether the inherited runners or more or less likely to score with a starter or reliever in the game the situation does work to the advantage of reliever vs starter ERA. ALL those shared runs scored will be charged to the starter who put them on and none to the reliever who allowed them to come home.

[brett]

--Regardless of whether the inherited runners or more or less likely to score with a starter or reliever in the game the situation does work to the advantage of reliever vs starter ERA. ALL those shared runs scored will be charged to the starter who put them on and none to the reliever who allowed them to come home.

So it may lower the starters ERA, but it will raise his relative ERA. Agreed.

[SABR Matt]

--Regardless of whether the inherited runners or more or less likely to score with a starter or reliever in the game the situation does work to the advantage of reliever vs starter ERA. ALL those shared runs scored will be charged to the starter who put them on and none to the reliever who allowed them to come home.

That's not the point. The point is...would using the starter longer improve tam run prevention...and the answer is...NO! Leaving the starter in longer would just mean the starters would be giving up those runners, not the relievers.

[StillFlash]

The starter is tiring, in each subsequent inning, at least after 4 or 5, the starter will be declining in quality.

However, if the bullpen is given more average innings per game, you have to find enough warm bodies who are talented enough to perform at a level better than the starter would have.