Starters v Relievers

[Gastric ReFlux]

From the OOTP reports:

Relief, low 2.98, high 5.23, middle (8th) 4.40
Starters, low 3.14, high 5.71, middle (8th) 4.23

Not enough to draw any firm conclusions, although it appears that relief pitching ERA is still on the high side compared to the major leagues in 2005.

[andymac]

I averaged the totals from MLB 2005 & the test league.

		MLB 2005	Test League

Starters	4.39		4.26	

Relievers	4.08		4.42

[Crapshoot]

Originally Posted by RonCo

Sigh...

Yes. That's often the case, but I've never seen anything to show conclusively that pitch-by-pitch a pitcher is fundamentally better as a reliever than as a starter. Instead, I think ERA reduction in a relief role to be more reasonably explained by usage patterns (greater exposure to the platoon advantage, a need to only get 1 or 2 outs rather than 3 to extract themselves from a situation where they might give up earned runs, only facing a batter once in a day, etc), and a lack of more than one major league pitch (and the expectation that if you throw that same pitch to a major league hitter, he'll eventually figure it out), etc.

So, along your thought pattern, all pitchers could be modeled as starting with their good stuff. Great starters will hold onto their great stuff a couple times through the order. Relievers will lose their great stuff more quickly.

Now, the logic complaint to this is that you'll still see reliever who "aren't as good as they should be." I think the right solution to this is to improve the AI that decides how relievers will be used. Greater control over platoon advantage (as was discussed in another thread) is a key place to start. All you need to do to see what I'm talking about is to read about 30 game logs. OOTP usage of relief pitchers is better than it was, but it's still got a long way to go. Improve this, and I think you're a long way toward the goal you're trying to achieve.

"". I think I've been complaining about OOTP bullpen AI (alongside Draven and a few others) for 3 iterations - I think you're damning with faint praise when you suggest that it "is better than it was."

Fundementally, the handling of a bullpen (and relief pitchers overall) along with a player's logic when resigning (awareness of market conditions) are the two most important things to me in this iteration - I don't expect a perfection DIPS engine, but I would like to see these fixed.

[Crapshoot]

Originally Posted by tangotiger

Hi,

I'm the author of The Book, and I did the study on the starter/relief. If someone has any questions, I'd be happy to answer them.

Tom

Tom, I think our fervent hope is that Markus spends some time talking to you about this.

[tangotiger]

You can get split data from Retrosheet:

http://www.retrosheet.org/boxesetc/YS_2005.htm

Just page down to the relief section. It's fairly obvious that a team's best pitchers are starters, on average. If the average reliever has a better ERA (and better component stats too), something's up. This was addressed in my book, and I find that the relief advantage is on the order of 0.80 ERA points. That is, just moving the same guy from a relief role to a starter role will worsen that guy's component stats to the effect of 0.80 in ERA. That's a substantial change.

[QuestGAV]

Originally Posted by tangotiger

You can get split data from Retrosheet:

http://www.retrosheet.org/boxesetc/YS_2005.htm

Just page down to the relief section. It's fairly obvious that a team's best pitchers are starters, on average. If the average reliever has a better ERA (and better component stats too), something's up. This was addressed in my book, and I find that the relief advantage is on the order of 0.80 ERA points. That is, just moving the same guy from a relief role to a starter role will worsen that guy's component stats to the effect of 0.80 in ERA. That's a substantial change.

We appreciate having someone who has studied the subject for so long weigh in. If we're lucky maybe Markus will make use of your expertise as he tunes OOTP, I think the game would be better off for it.

[Joe Morgan]

I don't think anybody was saying that the game makes relievers worse than starters, RonCo. I think people were saying that it came out that way. In addition, as TangoTiger is pointing out (my Joe Morgan senses are tingling with AGONY here, people), all things being equal, relievers *should* outperform starters. This concept predates Tony LaRussa, so no, I don't think you can just put it down to the game not using LOOGYs effectively.

The quick fix here is IMO the best: give relievers a boost in ratings the first time they go through a lineup in a game. Simple and yet probably the most realistic way there is to do this.

[jbsnadb]

OOTP has always been very hit or miss with relievers. I've never had a reliever CONSISTANTLY pitch well, and have always has a relatively high ERA for my bullpen, no matter how well my team has performed during the season.

Also, I personally wish the AI handles the closer role better. Everytime I ever simmed a good chunk of the season, I always ended up with three or four different pitchers collecting saves, with no rhyme nor reason behind it.

[RonCo]

Originally Posted by Crapshoot

"". I think I've been complaining about OOTP bullpen AI (alongside Draven and a few others) for 3 iterations - I think you're damning with faint praise when you suggest that it "is better than it was."

Fundementally, the handling of a bullpen (and relief pitchers overall) along with a player's logic when resigning (awareness of market conditions) are the two most important things to me in this iteration - I don't expect a perfection DIPS engine, but I would like to see these fixed.

I am always complaining about it, too. My comment wasn't meant to be faint praise or anything else. It was merely a statement of fact based on data I've taken. It _has_ gotten better, but it's not good yet...and I think the game's statistical output tells that story, er, well.

[RonCo]

Originally Posted by Joe Morgan

I don't think anybody was saying that the game makes relievers worse than starters, RonCo. I think people were saying that it came out that way. In addition, as TangoTiger is pointing out (my Joe Morgan senses are tingling with AGONY here, people), all things being equal, relievers *should* outperform starters. This concept predates Tony LaRussa, so no, I don't think you can just put it down to the game not using LOOGYs effectively.

The quick fix here is IMO the best: give relievers a boost in ratings the first time they go through a lineup in a game. Simple and yet probably the most realistic way there is to do this.

The argument was based on the idea that relief pitchers should be given a ratings boost somehow in order to make them perform better than starters. This data suggests that ratings of relief pitchers are just fine, and that the best solution lies somewhere else. I point to the AI controlling bullpen use.

[Gastric ReFlux]

Originally Posted by tangotiger

You can get split data from Retrosheet:

http://www.retrosheet.org/boxesetc/YS_2005.htm

Just page down to the relief section. It's fairly obvious that a team's best pitchers are starters, on average. If the average reliever has a better ERA (and better component stats too), something's up. This was addressed in my book, and I find that the relief advantage is on the order of 0.80 ERA points. That is, just moving the same guy from a relief role to a starter role will worsen that guy's component stats to the effect of 0.80 in ERA. That's a substantial change.

That's an excellent resource, thank you for pointing out the link.

It appears to show that relief ERA is consistently lower than starting pitching ERA.

Relief vs Starting, Year
4.11 v 4.36, 2005
4.14 v 4.62, 2004
4.14 v 4.52, 2003
4.01 v 4.41, 2002
4.10 v 4.57, 2001
4.56 v 4.87, 2000

Originally Posted by RonCo

The argument was based on the idea that relief pitchers should be given a ratings boost somehow in order to make them perform better than starters. This data suggests that ratings of relief pitchers are just fine, and that the best solution lies somewhere else. I point to the AI controlling bullpen use.

But see, look at the data from the major leagues above. If we assume that starting pitcher talent and relief pitcher talent is equivalent to one another at that level, we'd expect the ERA to be roughly equal. The data shows that ERA in relief is consistenly, repeatably lower than that from starting pitching.

Now I'm not sure of this, but there might be good evidence that on average, the talent of starting pitching is superior to that of those who become relievers. If that's the case, then we have further reason to believe there is something going on, beyond just purely situational managing, to give a pitcher in relief an edge over pitching as a starter.

The most glaring difference is that starters have to face the same batters more than once in a game.

Another glaring difference is that starters have to pitch longer, and are more likely to pitch tired as the game goes on.

These are reasons why it might be a good idea to implement some sort of rating boost to relievers at least, and then later work on the whole model of pitching, where a pitcher might gradually lose from his ratings as a game progresses.

[RonCo]

Originally Posted by Gastric ReFlux

Now I'm not sure of this, but there might be good evidence that on average, the talent of starting pitching is superior to that of those who become relievers. If that's the case, then we have further reason to believe there is something going on, beyond just purely situational managing, to give a pitcher in relief an edge over pitching as a starter.

The most glaring difference is that starters have to face the same batters more than once in a game.

Another glaring difference is that starters have to pitch longer, and are more likely to pitch tired as the game goes on.

These are reasons why it might be a good idea to implement some sort of rating boost to relievers at least, and then later work on the whole model of pitching, where a pitcher might gradually lose from his ratings as a game progresses.

I agree with your basic discussion points, I just don't think giving ratings boosts to relief pitchers is the most palatable answer.

Yes, let's model pitchers to be more hittable after a hitter has seen more pitches from him. There is evidence that this is true. I've posted it elsewhere. Yes, let's model a starter with three excellent major league quality pitches as losing less over time than one with only one. All that is great because that's how baseball seems to work based on currently understood data. I don't think we have any points of dsagreement there.

Yes, let's work on the AI so that it can do a better job of optimizing a manager's use of the pen. I can't think of anyone who reads the game logs who woud be against that idea.

But I do disagree with the idea of putting a band-aid rating boost in for relievers because ratings represent the fundamental quality of the pitcher, and there is no evidence that I've seen to say that relief pitchers, as a whole throw better quality pitches than starters...in fact the "conventional wisdom" is the opposite. And what data there is, tends to support this.

All that said, I continue to have an open mind. If someone can point me to data anywhere that credibly shows that relief pitchers have been of higher quality than starters at any time during any era of the MLB, minors, Japanese leagues, Mexican Leagues, etc...I'll be happy to consider changing my position.

[RonCo]

Originally Posted by jbsnadb

OOTP has always been very hit or miss with relievers. I've never had a reliever CONSISTANTLY pitch well, and have always has a relatively high ERA for my bullpen, no matter how well my team has performed during the season.

Also, I personally wish the AI handles the closer role better. Everytime I ever simmed a good chunk of the season, I always ended up with three or four different pitchers collecting saves, with no rhyme nor reason behind it.

Never?

If so, then I think you either have a set-up issue, or you haven't actually looked at any data.

Attached is the last three seasons of a team I run in a v5 online league, sorted by starters and relievers. In 2 of the three years, my bullpen has out-performed the starters regarding ERA, and in the third year, they would have also done so were it not for a bomb of a year posted by Michael Bennett.

In addition, you'll see that most of my pen has been pretty dependable year-to-year, within some variance. Ray Dietrick leads the team with 30+ saves each season, and a few others pick up some here and there.

And this is with a moderately poor v5 AI.

I'm not saying all teams in OOTP are like this. The data says otherwise. But, if you've truely NEVER seen a team do well, then I worry about something local to your situation.

[RonCo]

I should note that this team won 90+ games in each of these seasons, won the division once, and the wildcard another time.

[Gastric ReFlux]

Originally Posted by RonCo

But I do disagree with the idea of putting a band-aid rating boost in for relievers because ratings represent the fundamental quality of the pitcher,

Try not to think of it as a band-aid, but an actual move towards moving to a better quality model of pitching in OOTP.

We have good data that reliever ERAs are better than starting pitching ERAs. OOTP functions through the use of ratings, so we need some way of representing the better performance seen by relief pitchers in real life. There may be a few ways to accomplish this:

1. Batters receive a ratings boost for successive at-bats against the same pitcher in the same game.
2. Pitchers receive a ratings boost upon entry into a game, this would be the "band-aid" fix for relievers.
3. A combination of those two.

I don't know how much time Markus has available to him to rework how pitcher's ratings are degraded in-game, but it's become more and more apparent over time that OOTP is failing to represent relief pitching accurately. It's not just that it's failing to represent, it even fails the appearance doing it, as seen by starter pitching ERA and relief pitching ERA in game.

That's a problem. It's a problem that if it persists through OOTP2006, I won't be happy about it. Others won't be happy either.

Perhaps it could even be considered that Markus puts a user-available setting in the game, where we can specify how much of a boost we would like relievers in relief appearances get to their ratings, much like the way we can adjust the talents of players entering the game through drafts. Then if a person doesn't want relievers to get a boost, just keep the modifier at 0. Those who want relief pitching to at least achieve appearance of accuracy can adjust the modifier percentage.

I believe this to be a reasonable proposal, and it merits attention in how it would address what has been a fundamental problem with OOTP.

[RonCo]

Originally Posted by Gastric ReFlux

Try not to think of it as a band-aid, but an actual move towards moving to a better quality model of pitching in OOTP.

...

I believe this to be a reasonable proposal, and it merits attention in how it would address what has been a fundamental problem with OOTP.

We have the same goals, GR, and I respect your proposal. If Markus puts it in, good on you.

We also have agree on the fact that bullpen usage is an issue within the OOTP universe, and always has been.

However, I will always see the proposal to give relievers an artificial boost as a band-aid because it is fundamentally flawed with respect to how I understand pitching actually works. Why put in an artificial rating that isn't supported by data when you can develop an approach that is?

I remain flexible in my stance, though. Show me something that "proves" relief pitchers in real life are made of better sutff than starters, and I'll consider changing my opinion that the proposal is anything but a band-aid.

[Gastric ReFlux]

Originally Posted by RonCo

However, I will always see the proposal to give relievers an artificial boost as a band-aid because it is fundamentally flawed with respect to how I understand pitching actually works. Why put in an artificial rating that isn't supported by data when you can develop an approach that is?

I remain flexible in my stance, though. Show me something that "proves" relief pitchers in real life are made of better sutff than starters, and I'll consider changing my opinion that the proposal is anything but a band-aid.

Ermm, you still need to, umm, reconceive how you're approaching the thoughts about this. How do I explain this?

We probably both agree that relief pitchers, in real life, don't have better talents than starting pitchers. We'd probably both be inclined to think that starting pitchers in real-life have better talents, and because of that, get the role that gives them the most innings.

Ahh, think of it like this. You know how there were stadium modifiers to the game? Functionally, these altered or modified the ratings of the players, but didn't alter the actual ratings we'd see.

Roughly, hitter had a power rating of 70, and played in a park which helped his homers some percentage. When in the game, we'd see 70 on our screen, but the actual calculation by OOTP in the game was using the modification, right?

That's what this band-aid would do to the ratings of pitchers when they appear in relief. If that same pitcher starts the game, he wouldn't get that modification.

Does that help explain the idea better?

[tysok]

Personally I don't see it... but I never looked at the relief ERA of anything but my team in OOTP so I'm missing some stats from the game of course.

Talent levels change from level to level (like minors to majors) but basically the 'rules' don't. I think there is a difference between relievers and starters, more than just relievers putting everything they have into every pitch... but that's unprovable and beside the point of this post.

I don't know how the results show up in-game, I don't care to look back at the posts in this thread to find them.... but here's the data from 2005.

The numbers are from the standings. So it would be Atlanta on top all the way to Colorado on the bottom... anyhow:
starter - relief ERA
3.65 4.74
4.2 4.24
3.71 3.86
3.81 4.85
4.03 3.55

3.61 3.17
3.46 3.63
4.02 3.87
4.17 4.24
5.38 4.75
4.63 4.06

4.49 3.49
4.58 5.4
4.52 3.98
4.36 4.42
5.3 4.8

Half the teams in the NL had a higher relief ERA.

4.59 4.37
4.56 5.15
4.2 3.81
4.82 4.1
5.62 4.99

3.75 3.23
3.96 2.8
3.93 3.2
4.85 3.78
6 4.7

3.75 3.52
3.82 3.39
5.04 4.83
4.9 3.6

1 team in the AL had a higher relief ERA.
ML total: 4.36 - 4.11
NL total: 4.23 - 4.20
AL total: 4.52 - 4.00

Whatever that's worth. I wanted to look it up myself, and I think it puts a better light on it than some of the piece meal data I saw above.
It's not that large a difference, but if someone can point out a pitcher that had a lot of starts and a lot of relief appearances last year I wouldn't mind looking him up... I don't know of any. This just shows ML relief verse ML starting staff... doesn't address whether a pitcher is "better" in relief than when he starts.
In my experience in the game when I put a crappy relief squad on the field I ended up with crappy numbers, when I put a moderate relief squad together I got moderate results, and when I put a stellar relief squad together I got great results.

[RonCo]

Originally Posted by Gastric ReFlux

Does that help explain the idea better?

I understand the idea, and I agree it could work to make the stats look more like MLB stats. I just think that it is not the only idea that will result in better stat conformity, and in the spectrum of all ideas, I think it is below some of the others with regards to its quality. I prefer a truer model to a band-aid. If anything, the band-aid I would implement would go the other way...reduce a SP's effectiveness each time through the order so that it increases the starter's ERAs. At least that kind of works in accordance with data and studies I've seen. In that band-aid, relievers would generally perform better because they would rarely receive the penalty that starters received.

But if you are looking for me to agree that your idea can work, I agree.

[Joe Morgan]

Originally Posted by RonCo

The argument was based on the idea that relief pitchers should be given a ratings boost somehow in order to make them perform better than starters. This data suggests that ratings of relief pitchers are just fine, and that the best solution lies somewhere else. I point to the AI controlling bullpen use.

No it doesn't. The data you put forth suggests that relievers are exactly the same as starters when they actually should be better to reflect reality. Although with Tangotiger's data it appears that what should really happen is that there ought to be a boost given to anybody who appears in relief.