Solution to the Pitching Model?

[Mike Donlin]

Yet another entry in the ongoing saga, but this one has the benefit of simplicity, I think.

(1) At creation, all pitchers are capable of starting (i.e. throwing 100+ pitches per outing.)

(2) At some point in a pitcher's career, if he is repeatedly injured or if he does not have enough talent to start, he is switched to a reliever.

(3) Once a reliever, a pitcher will receive a boost, hidden or not, to his ratings, which will enable him to pitch better than as a starter; he will have a decreased chance for injury; his endurance will gradually drop; he may have a higher chance of a talent increase. (These to account for the fact that relievers consistently perform better than starters; OOTP seems to be hardwired for this anyway, so most of these provisions--excepting the decrease in endurance and injury risk--might be unnecessary.)

(4) It would be possible to make a reliever back into a starter, but only at considerable cost to his performance and increased injury risk, so much so that it would rarely be worth the effort.

That's it. Thoughts? If we come to some sort of consensus, I'll post this in the suggestions forum.

[injury log]

I think unless we answer the question: why do pitchers become relievers in real life? we're going to have a pitching model that doesn't feel like real life. I'd suggest the two most important factors here are:

-build
-repertoire

especially repertoire. Those pitchers who never master a third pitch normally end up as relievers.

Originally Posted by Mike Donlin

(1) At creation, all pitchers are capable of starting (i.e. throwing 100+ pitches per outing.)

This should likely be true of HS pitchers, but not of all college pitchers. Many relievers are drafted from college and no one ever considers converting them to starting, regardless of how good they are, Brett Cecil being an exception.

Originally Posted by Mike Donlin

(2) At some point in a pitcher's career, if he is repeatedly injured or if he does not have enough talent to start, he is switched to a reliever.

Injuries should definitely lower Endurance. I think it's clear that some relievers are just as, if not more 'talented' than starters, though their repertoires might not be as deep. I'd hate to see a game where the bad pitchers become relievers (and then get some kind of ratings boost) and the good ones stay starters; that doesn't seem to have anything to do with real life.

Originally Posted by Mike Donlin

(3) Once a reliever, a pitcher will receive a boost, hidden or not, to his ratings, which will enable him to pitch better than as a starter; he will have a decreased chance for injury; his endurance will gradually drop; he may have a higher chance of a talent increase. (These to account for the fact that relievers consistently perform better than starters; OOTP seems to be hardwired for this anyway, so most of these provisions--excepting the decrease in endurance and injury risk--might be unnecessary.)

Question about real life: do all pitchers get the same boost when they switch to relief? Those pitchers who excel at relief are, I'd think, typically those pitchers who have one outstanding pitch, or who can throw at high velocity in short spurts but tire quickly. But most don't get a Gagne/Papelbon jump. Relievers tend to have better ERAs than starters, but that's at least partly down to usage; relievers more often face batters of the same handedness, get credit for outs from inherited runners (they can come in with a man on first and get a free out because of CS or GIDP), etc. Finally, while I don't know that RPs are at less risk of injury than SPs in real life (I'll be looking at data on that soon), they certainly are in OOTP, since in OOTP, injuries happen on plays in game. The more a guy plays, the more likely he is to get hurt.

Originally Posted by Mike Donlin

(4) It would be possible to make a reliever back into a starter, but only at considerable cost to his performance and increased injury risk, so much so that it would rarely be worth the effort.

I'd add that these ideas - stretching relievers out to become starters, or reducing a pitcher's Endurance if he's used in relief - were discussed extensively by the Design team, and arriving at a simple implementation, one that could not be exploited by the user, is harder than it might seem.

Originally Posted by Mike Donlin

That's it. Thoughts? If we come to some sort of consensus, I'll post this in the suggestions forum.

Given how long these discussions have been going on, consensus seems unlikely. I made a proposal last year that I thought was simple, and grounded in real life, which I might dig up and post here. The basic idea was this (bear in mind, the goal was simplicity):

-for each pitch (Fastball, Sinker, Curve, etc), a pitcher would have a rating of 0, 1 or 2
-a rating of 0 would mean that the pitch is not in the pitcher's repertoire
-a rating of 1 would mean that the pitch is in the pitcher's repertoire, but he doesn't throw it well enough to get away with using it at the top league level
-a rating of 2 would mean that the pitch is in the pitcher's repertoire, and he commands it well enough to use it at the top league level

A pitch rated a '1' would be presented in normal type, and a pitch rated '2' in bold. So, a draft-eligible HS pitcher might have the following repertoire:

Fastball
Curveball
Changeup

Any pitcher without three pitches suitable for his level would have a severe penalty when pitching long outings. So the guy above would be capable of starting in the minors, but not in the Majors; he has three pitches of minor league quality, but only one of Major League quality. Of course, in the normal course of development, pitchers would have the opportunity to both upgrade their current repertoire and to learn new pitches, and I had worked out what pitch ratings pitchers should have on player creation, and what probability they should have of developing their repertoires, to produce an appropriate number of starters and relievers. So the guy above would have a decent but not certain chance of having a repertoire something like this after four years:

Fastball
Curveball
Changeup
Slider

In this way, some guys would naturally end up as relievers, despite having the Endurance to start in the minors. Drafting would be more interesting; do you pick the guy with the less developed repertoire but with better ratings? And we get one small step towards making pitch quality important in the game.

Using a 0-3 rating scale would be even better; guys with pitches rated a '3' could be those who get a greater boost when used as relievers. I had suggested that high-Velocity pitchers get a bigger boost than low-Velocity pitchers when used in relief, though that suggestion proved controversial.

[Mike Donlin]

Originally Posted by injury log

This should likely be true of HS pitchers, but not of all college pitchers. Many relievers are drafted from college and no one ever considers converting them to starting, regardless of how good they are, Brett Cecil being an exception.

But most college relievers were starters in high school.

Originally Posted by injury log

Injuries should definitely lower Endurance. I think it's clear that some relievers are just as, if not more 'talented' than starters, though their repertoires might not be as deep. I'd hate to see a game where the bad pitchers become relievers (and then get some kind of ratings boost) and the good ones stay starters; that doesn't seem to have anything to do with real life.

If their repertoires are not as deep, they would not be--by definition--as talented as starters. More on your second point below.

Originally Posted by injury log

Question about real life: do all pitchers get the same boost when they switch to relief? Those pitchers who excel at relief are, I'd think, typically those pitchers who have one outstanding pitch, or who can throw at high velocity in short spurts but tire quickly. But most don't get a Gagne/Papelbon jump. Relievers tend to have better ERAs than starters, but that's at least partly down to usage; relievers more often face batters of the same handedness, get credit for outs from inherited runners (they can come in with a man on first and get a free out because of CS or GIDP), etc. Finally, while I don't know that RPs are at less risk of injury than SPs in real life (I'll be looking at data on that soon), they certainly are in OOTP, since in OOTP, injuries happen on plays in game. The more a guy plays, the more likely he is to get hurt.

So the ratings boost would be minimal. And the data I have seen indicate that, yes, relievers spend considerably less (about half) time on the DL.

Originally Posted by injury log

Given how long these discussions have been going on, consensus seems unlikely. I made a proposal last year that I thought was simple, and grounded in real life, which I might dig up and post here. The basic idea was this (bear in mind, the goal was simplicity):

-for each pitch (Fastball, Sinker, Curve, etc), a pitcher would have a rating of 0, 1 or 2
-a rating of 0 would mean that the pitch is not in the pitcher's repertoire
-a rating of 1 would mean that the pitch is in the pitcher's repertoire, but he doesn't throw it well enough to get away with using it at the top league level
-a rating of 2 would mean that the pitch is in the pitcher's repertoire, and he commands it well enough to use it at the top league level

A pitch rated a '1' would be presented in normal type, and a pitch rated '2' in bold. So, a draft-eligible HS pitcher might have the following repertoire:

Fastball
Curveball
Changeup

Any pitcher without three pitches suitable for his level would have a severe penalty when pitching long outings. So the guy above would be capable of starting in the minors, but not in the Majors; he has three pitches of minor league quality, but only one of Major League quality. Of course, in the normal course of development, pitchers would have the opportunity to both upgrade their current repertoire and to learn new pitches, and I had worked out what pitch ratings pitchers should have on player creation, and what probability they should have of developing their repertoires, to produce an appropriate number of starters and relievers. So the guy above would have a decent but not certain chance of having a repertoire something like this after four years:

Fastball
Curveball
Changeup
Slider

In this way, some guys would naturally end up as relievers, despite having the Endurance to start in the minors. Drafting would be more interesting; do you pick the guy with the less developed repertoire but with better ratings? And we get one small step towards making pitch quality important in the game.

Using a 0-3 rating scale would be even better; guys with pitches rated a '3' could be those who get a greater boost when used as relievers. I had suggested that high-Velocity pitchers get a bigger boost than low-Velocity pitchers when used in relief, though that suggestion proved controversial.

The advantage of my proposal, I think, is that it does not involve rating individual pitches, which is, above all else, why I think Markus has been reluctant to fix the problem.

As far as I can tell, starters become relievers for two reasons. One, as you note, they do not have as many pitches as starters. Two, they have a history of repeated injuries (Kerry Wood, the Indians' prospect Adam Miller, et. al.) Number two we can reproduce now without much trouble. Number one is harder without rating pitches. But presumably that limited repertoire would ultimately be reflected in performance. Thus performance and talent (in the form of ratings and potential) would drive who remains a starter and who becomes a reliever--without having to rate pitches.

But you're probably right. Consensus is to much to aim for. I guess this approach just seemed simple and implementable now.

[CMH]

I like where you're going.

What if instead of a hidden talent jump, the player realizes a certain percentage of his talent?

Lets take for instance a guy who pitches in the 8th inning. He throws 98 mph with effort. He can do that because he's only going one inning. Throw him in the first and he might drop to perhaps 94-95 mph. He may use his other pitches more. He may pitch around. There are other factors that come into being a starter over a reliever.

Perhaps relievers seem to be more effective in the pitching model because they are overexherting themselves whereas a starter realizes he needs to turn it up only when necessary. This of course would then have to tie in with a hidden ability to recognize when you need to do this, but that's another discussion for another time.

The truth, and I think we all know this, is that almost every pitcher is capable of being a starter. They become conditioned to be relievers, or they have less pitches in their arsenal/less effective pitches in their arsenal so someone realizes down the line that they would be more effective in the pen for one inning.

Honestly, I think the entire endurance barometer needs to go. All starters, at least in my opinion, can throw 120 pitches. Some remain effective for 120 pitches, most do not. But those are the rare kind that can throw that much and be fine. A lot of the real reason most pitchers don't go for an entire game is because they either throw too many pitches early in the game or they walk too many batters. Roy Halladay is usually very good about keeping his pitch count low, and that has more to do with his complete games than this uncanny ability to throw 120+ pitches a game.

[RonCo]

Obviously, I think individual pitch ratings are required to go to the next level. But once you do that, you most likely break forever the linkage between the fictional game and the historical/modern.

[CMH]

Ah, I see that some of my points were already brought up.

[RonCo]

It's more complex than you might think. But ratings per pitch are the way to go.

[CMH]

Originally Posted by RonCo

It's more complex than you might think. But ratings per pitch are the way to go.

You mean on a scale of 20-80 (or whichever you choose) have every pitch rated for every pitcher?

I think that's closer to realistic than an endurance rating. Even if not a number rating, maybe a letter grade. I'm not sure what the scouts use to rate a pitch.

[RonCo]

Yes.

I developed a model to do just that--rate pitches on a 1-200 (250) scale in an OOTP environment. It was not well liked.

[RonCo]

It does, however, when combined with a good physical stamina model, resolve the issues inherent in the starter/reliever problem that has plagued OOTP since forever.

At least it does in my opinion...of course.

[RonCo]

But, when you do this, the historical and fictional game must almost certainly part for at least as long as it would take to manually create a massive pitcher database to replace Lahman. And there's the rub.

[nymets3000]

It wouldn't work for historical leagues for that reason. Correct me if I'm wrong but I think the Lahman Database lists pitches for a pitcher but have no ratings. I don't play historical so I don't know.

[RonCo]

That approach could work for historicals, too....but it would require a humongous amount of work to create a pitch-based database for every pitcher in history. It could be done. You could start with BIll James/Rob Neyer's pitching history book, but it would still be really, really hard.

[tysok]

Originally Posted by RonCo

That approach could work for historicals, too....but it would require a humongous amount of work to create a pitch-based database for every pitcher in history. It could be done. You could start with BIll James/Rob Neyer's pitching history book, but it would still be really, really hard.

As I pointed out repeatedly in those previous discussions, there's no reason that you have to have historically accurate pitch combinations. It doesn't exist now so why would it HAVE to if you went to this model? The game would know what creates a starter, and it should know what combination is better etc. If it sees Sandy Koufax it may give him a great fastball, a great change up and a super slider... as long as the end result works that's all that matters. That would be preferable, I would think, to the game seeing Sandy Koufax and giving him great "stuff", great "control", and great "movement", while showing he throws only a knuckleball and a slider.

[RonCo]

Originally Posted by tysok

As I pointed out repeatedly in those previous discussions, there's no reason that you have to have historically accurate pitch combinations. It doesn't exist now so why would it HAVE to if you went to this model? The game would know what creates a starter, and it should know what combination is better etc. If it sees Sandy Koufax it may give him a great fastball, a great change up and a super slider... as long as the end result works that's all that matters. That would be preferable, I would think, to the game seeing Sandy Koufax and giving him great "stuff", great "control", and great "movement", while showing he throws only a knuckleball and a slider.

Yes, you could also use trial and error to give historical pitchers a "fake" repertoire as long as it resulted in practically close numbers. It would still be hard, but maybe not as hard as I suggest. Dunno. Doesn't matter much what I think, though.

[Afino]

Originally Posted by tysok

As I pointed out repeatedly in those previous discussions, there's no reason that you have to have historically accurate pitch combinations. It doesn't exist now so why would it HAVE to if you went to this model? The game would know what creates a starter, and it should know what combination is better etc. If it sees Sandy Koufax it may give him a great fastball, a great change up and a super slider... as long as the end result works that's all that matters. That would be preferable, I would think, to the game seeing Sandy Koufax and giving him great "stuff", great "control", and great "movement", while showing he throws only a knuckleball and a slider.

Yeah - that was my thought as well. Especially considering the incomplete data we have (and the time and research it would take), why even bother if we can get the end result to work?

[tysok]

Originally Posted by RonCo

Yes, you could also use trial and error to give historical pitchers a "fake" repertoire as long as it resulted in practically close numbers. It would still be hard, but maybe not as hard as I suggest. Dunno. Doesn't matter much what I think, though.

Doesn't matter much what I think either... but we keep talking anyhow. I must like the look of my own words.

[Mike Donlin]

Originally Posted by tysok

As I pointed out repeatedly in those previous discussions, there's no reason that you have to have historically accurate pitch combinations. It doesn't exist now so why would it HAVE to if you went to this model? The game would know what creates a starter, and it should know what combination is better etc. If it sees Sandy Koufax it may give him a great fastball, a great change up and a super slider... as long as the end result works that's all that matters. That would be preferable, I would think, to the game seeing Sandy Koufax and giving him great "stuff", great "control", and great "movement", while showing he throws only a knuckleball and a slider.

Yeah, I should have pointed out that I was imagining this solution would apply only to fictional players.

And I think that it would be very difficult to produce historically accurate stats based on ratings that we ascribe to a pitcher's arsenal according to their DIPS rate stats. I suppose it's possible, but that would require an extensive rewriting of the pitcher ratings.

What I'm suggesting is that we fix fictional and leave historical based on the Lahman. After all, in historical, we know what roles pitchers eventually ended up in. Their role is not arbitrary in quite the same way that a fictional pitcher's is.

[Curtis]

Originally Posted by CMH

You mean on a scale of 20-80 (or whichever you choose) have every pitch rated for every pitcher?

I think that's closer to realistic than an endurance rating. Even if not a number rating, maybe a letter grade. I'm not sure what the scouts use to rate a pitch.

According to Mets Inside Pitch magazine, fastballs are rated purely on velocity, in 5 point increments from 20-80. Each increment coresponds to one of the velocity groupings in the game. Whether or not you can hit the broad side of a barn doesn't enter into it, and I can't remember if there's a seperate control rating.

All other pitches are rated from 20-80 in 10 point increments (effectively 2-8 scale), with 50 being an average major league quality pitch, 40 a good minor league pitch, 30 a bad minor league pitch and 20 meaning the pitcher can't throw that pitch. You'll have to guess what 60, 70 and 80 equate to. Unlike the fastball, other pitches include subjective evaluations of how much the ball moves, how wild the pitch is within the zone (ability to hit a spot) and how wild outside the zone (control).

Note that if a pitch has great movement, it's not neccessarily a bad thing to be wild outside the zone. It could mean that the ball is starting out aimed for the strike zone — thereby inducing the batter to swing at it — but ends up outside it. Because of the movement it becomes a swinging strike or feeble grounder instead of a ball. If you don't have a good defensive catcher, much movement + wild outside the zone = many wild pitches or passed balls. (Just my opinion, but I suspect this is why many AAA and AA teams keep a veteran catcher on the roster — so pitchers like that don't have their development stunted by watching too many balls roll to the backstop, thereby making them afraid to throw their breaking pitches. Something else that could be factored into pitcher development…)

[CMH]

Well, I would think that the above should definitely find it's way into OOTP in the near future.

It would make every pitch per at-bat yield even more realistic results.