The future of OOTP: Meaningful pitches

[DiMaggio5CF]

Here is what is going to take OOTP to the next level . . . meaningful pitches.

What makes a starter a starter and a reliever a reliever? Endurance? Sure, but what determines endurance. Are people born with an ability to throw only a certain amount of pitches at a time, or is that endurance a product of training and conditioning? Is a pitcher born as either a starter or a reliver, or is their role dependent on something else?

I contend that the difference between starters and relievers is their available arsenal -- specifically, how many pitches they throw well. A starter probably needs three different pitches; as a starter, a pitcher faces the same hitters a number of times in a game, and therefore must mix up his pitch selection, which means that he needs three Major League quality pitches. A reliever, on the other hand, probably faces a specific hitter only once per game, and so the need to mix up pitch selection isn't as important; a reliever needs two good pitches, or one dominant pitch and at least average pitch.

So how can OOTP better handle this? First of all, get rid of endurance ratings. Have every single college and high school pitcher enter professional baseball with the ability to be a starter (assuming, of course, that he was a starter in amateur ball) and determine endurance by how a player is used. The more innings a pitcher throws per season in his career, the more capable he is of becoming a starter. For example, a short reliever who logs 60 innings per season probably won't be able to convert into a 200-pitch starter, but a long reliever who routinely sees upwards of 100 innings out of the bullpen will probably be able to handle the increased workload fairly well. And the more time that a pitcher spends as a reliver, the harder it will be for him to jump into the starting rotation; a guy who has been a reliever for two seasons will be able to handle the jump better than a veteran who hasn't thrown more than 70 innings in the past ten years. Going the other way, a pitcher who is accustomed to being a starter will be able to easily convert to being a reliever, but he might struggle throwing on back-to-back days, and it might take a season before he can handle a few days of work in a row.

Second, go to ratings for each individual pitch. For each pitch, give a velocity range, a control rating, and a rating for movement. Take this chart as an example (ratings based on 1-100):

Justin Johnson(L, 23 years old)

Pitch 1: Fastball
Velocity: 93-96
Control: B+
Movement: B (9 to 3)
Comment: His bread-and-butter, Johnson can throw his fastball early for a strike and blow it by hitters for a strikeout. It has a slight movement in on LHB, jamming them.

Pitch 2: Slider
Velocity: 85-87
Control: C
Movement: A+ (2 to 8)
Comment: Because of the sharp break away from LHB, this pitch is killer on lefties. It can get out righties as well, but they are more likely to lay off it for a ball.

Pitch 3: Curveball
Velocity: 81-83
Control: F
Movement: B
Comment: The curveball has good break for Johnson, but he just can't control it, resulting in many walks, wild pitches, and cookies right down the middle. If he can harness this pitch, it has the potential to be a plus Major League curve, but right now it does more harm than good.

Pitch 4: Split-Finger
Velocity: 88-91
Control: D-
Movement: C
Comment: Johnson's control struggles with this pitch, and it sometimes doesn't break as sharply as it should, resulting in a lot of hanging splitters, which usually go for homeruns.

In addition, give a blurb about the pitcher's ability vs RHB and vs LHB from either a scout or, if he's on your active roster, the pitching coach -- and for youngsters, include whether he would be more effective as a starter or a reliever. For example: "Because of the break away from LHB, Johnson would make a great left-handed specialist. He only has two plus pitches, so he would best be used as a reliever."

For a fastball, speed would be the primary factor of quality, with movement as the secondary factor. For breaking balls, it would be the opposite, with movement (amount of break) as the primary factor and speed as the secondary factor. The third factor for breaking balls would be the differential in velocity; when comparing two pitchers with the same quality of fastball and changeup, the pitcher with the bigger difference in speed will be slightly more effecitve. Control, for each pitch, is simply an ability to hit spots; poor control leads to a lot of walks and a lot of hard-hit mistake pitches. Additionally, a fastball will generally have better control than any breaking pitch, and when a pitcher falls behind in a count (usually due to poor control) he'll almost always go to the fastball, which will usually be of poorer quality, and which will usually be hammered by good fastball hitters, which we'll get to later.

Instead of giving L/R rating splits, just make the pitcher better against hitters to which the ball breaks away. It's easier to hit a ball breaking toward you than it is a ball breaking away, so it stands to reason that hitters for which the ball breaks away will have a harder time. The better a pitcher's breaking ball is, the more pronounced that differential will be; a fastball pitcher will be pretty equal against lefties and righties, but a pitcher whose best pitch is a breaking ball will be much better against one or the other.

Additionally, if you choose to go with this pitcher as a reliever, you can opt to disable some of his pitches; the advantage is that he doesn't need the extra pitches and you'll optimize his performance by having him throw only his best pitches, while the disadvantage is that his ratings for the disabled pitchers will decrease and make him less effective if you later put him in the bullpen.

On the other hand, if you choose to put him in the rotation, it would be best to allow him to throw all of his pitches. With meaningful pitches, pitch selection would matter, and if batters see the same two pitches over and over, they'll adjust and cream even the best pitches. So even though Johnson might get creamed when he throws his splitter, it would still be better than throwing just the two best pitches. If he threw just the two best pitches, he would get outs at first but then hitters would adjust the second and third times through the lineup -- which is why he would be better as a reliever.

So does available arsenal mean that pitchers will be starters or relievers forever based just on what they can throw? Of course not. During Spring Training and a player's time in the minors, pitchers can develop, and so can their stuff. During Spring Training, you would be able to target certain areas to work on specifically, whether it be a specific pitch or even a specific aspect of a pitch (for example, the control of a pitcher's curveball). In the minors, the minor league pitching coach will work on whatever pitches are enabled. If you want Johnson to be a starter, keep him throwing all pitches (or at least three or four) and the pitching coaches will try to work on all of them; some pitches might develop while others might not, and coaches might just not be able to develop all his pitches, pretty much making the decision for you. If you want him to be a reliever, disable some pitches and have the pitching coach focus on just the hurler's one or two best pitches.

So what does this mean for hitters; after all, it doesn't make sense to re-vamp the pitching aspect without doing anything to the hitting aspect. Well pitch selection is important for hitters, too. Some pitchers are fastball hitters; others are breaking ball hitters. Naturally, when the Yankees face fireball Brad Lidge, Gary Sheffield will probably hit the ball harder than Alex Rodriguez will.

So the answer is to give batters rating against the fastball and the breaking ball. Look at this as an example (1-100 ratings):

Joseph Anderson (R, 30 years old)

vs RH Fastball
Contact: 75
Power: 60
Eye: 80
Strikeout: 70

vs RH Breaking Ball
Contact: 60
Power: 90
Eye: 45
Strikeout: 45

vs LH Fastball
Contact: 60
Power: 70
Eye: 85
Strikeout: 75

vs LH Breaking Ball
Contact: 50
Power: 100
Eye: 55
Strikeout: 55

So what can we tell about this batter? First of all, he makes better contact against righties, but hits for more power against lefties; additionally, he sees the ball better from lefties, which means more walks and fewer strikeouts against left-handed pitchers. Second, he makes better contact with the fastball, but he hits the curveball for better power; additionally, he sees the fastball better, which means that he's more likely to take a breaking ball for a strike and/or chase it out of the zone for a strike. In short, if you're willing to give up a homerun for a better shot to get an out, go with a left-handed junk baller; if you're willing to give up a hit to keep the ball in the park, go to a righty with a good heater.

Additionally, a batter will have hot spots and cold spots on the strike zone. Maybe he can crush anything down the middle, no matter what it is, but if that ball is down and away, he just can't make good contact; if his strengths are spread out throughout the zone, go with a stuff pitcher who can get him out with his pitches. But if the batter has distinct hot and cold zones -- if his cold spots are very cold -- go with a pitcher who has good control and can just put the ball where the batter can't get good wood on it.

Of course, a pitcher might not throw to a batter's weakness. It depends on the differential between the pitcher's strengths and the batter's weaknesses. If a batter is a slightly better curveball hitter, but the pitcher's fastball is immensely better than his breaking pitch, the pitcher will rely heavily on the fastball. On the other hand, if the pitcher's fastball and slider, for example, are relatively close in quality, he'll defer to the hitter's weakness in choosing his pitches.

Now obviously, a pitcher won't go to one pitch for an entire at-bat just because the scouting report says to, unless it happens that the ball is put in play on the first or second pitch of the at-bat. But a good pitcher following a good scouting report will throw other pitches just to mix it up, mostly out of the zone trying to get the hitter to chase, and he will rely on the primary pitch as his out pitch. Early in the count, you'll see the primary pitch to try to get an easy out, and when the pitcher has two strikes on the hitter, you'll see that primary pitch again as the pitcher goes for the strikeout. Of course, when the numbers dictate a mismatch, you might see a quick three-pitch K with the same pitch three in a row.

So what does all of this do? Well on the surface, it puts more emphasis on scouting. Now you need good scouts to do good homework on your opposition. Not only does the scouting report tell you which pitcher or hitter to go to for a specific game or a specific in-game situation, but it also tells the pitcher on the mound which pitches to throw to each hitter. Get a bad scouting report -- either from a bad scout or because you didn't pay enough attention to your upcoming opponent -- and you'll be in the dark.

But more than that, it makes the game more real. Baseball is about matchups. Does Mark Loretta hit all the pitchers that are overall worse than him and make an out against all the pitchers that are overall worse than him? Or does he hit the pitchers that he matches up well against and go down against pitchers that he matches up against poorly?

And as a side note, pitch selection will influence stolen bases; a runner will find it easier to successfully swipe a bag when a curveball is thrown and more difficult when a fastball is thrown. This also puts more pressure on you as the manager. When you see a situation where the pitcher will probably rely heavily on a breaking pitch, call for the steal; but when it's likely that you'll see some fastballs, keep your runner parked.

Pitches determine a lot in baseball. They determine matchups between individual pitchers and hitters (and therefore an individual pitcher's performance against a specific team), stolen base attempts, and a pitcher's effectiveness as a starter or a reliever.

I don't know anything about coding. But Markus, this is the next step for OOTP. Figure out how to do this, and you will be a baseball sim master. I know that this is a huge leap, and OOTP2006 is pretty much done. But if you can make a big change for OOTP2007, I really think that this could revolutionize the game and bring it to the next level.

[mauboy]

I agree with everything in this thread..

Nice work! I hope that someday it will be like this, too!

[Deft]

Been lobbying for this for awhile. Nice layout of how to implement it. Expect Big City to blast this for being too complex.

[sfgiants88]

That will require an absurd amount of micromanagement, and I will never do it.

However, if it is an option, and it doesn't take development time away from other features, then I don't see a problem with implementing it.

[Skipaway]

I don't think there are enough studies about what's suggested here to come up with reasonable ways to generate players and accurately simulate them. Or you just want complicated cosmetic stuff?

I don't see a way to have this helping the game to be more real.

[marc]

I like this a lot, but I realize it would have to be optional.

[mrbill]

Going to harken back to BBPro days, where pitchers had ratings for each pitch, as much as I love OOTP, I think BBPro's baseball model was really damn good, it was still as simple as a half dozen ratings to model each player, but I felt it was more complete and accurate than OOTP. In BBPro, hitters had to use their "eye" (represented by "contact" in BBPro) to try and determine the pitch coming. It used the "Tecmo Bowl" approach, if you guessed the pitch right, you got a big bonus for making contact on that particular pitch. If you guessed wrong, it was harder to make contact.

I believe the model was that the higher the pitches' ratings, the harder for the hitter to distinguish the pitch, which is very important in timing your swing and knowing where in the zone to swing. But, the 2nd time facing the pitcher, the hitter got a bonus to determining the pitch. I guess that bonus would have to be scaled to match the number of pitches the pitcher could throw, as to reward those pitchers with 3+ pitches.

This meant a reliever could do very well with 2 high rated pitches, they'd be extremely hard to figure out on the first time around. But, the 2nd time around, the hitter would get enough of a bonus to start figuring out the pitches more often. A starter with 4 pitches would get away with hiding his pitch selection to later in the game, even though those pitches may not be as spectacular as the reliever's on the first pass.

In spring training, you had to allocate points to each pitch, and each pitch had a potential. If a pitcher had more pitches, you had to devote more resources to developing/maintaining all 3-4 of them, and could put less into arm strength or control. This meant you could really tweak a reliever's 2 pitches, control and arm strength, but at the cost of letting endurance and a 3rd/4th pitch slide over his career. This gives the pro/con balance of starter and reliever and still defines a distinct "endurance" rating. But, at least it gives a real benefit to having a guy just relieve, in that its easier to maximize his potential in other areas. You can theoretically take advantage of that in OOTP now as well, but in BBPro, it actually was very difficult to develop any ratings if you tried to maintain 4 pitches. OOTP gives you no real benefit for sacrificing endurance, because most players development isn't tied to spring training.

I felt it added very little complexity, and in OOTP's case, you'd replace "Stuff" with a "Stuff per pitch" set of ratings. Although, Stuff and Movement are rather blurred on what they mean, but I imagine Velocity becoming meaningful in this situation to model pitch speed, which would mostly affect the speed of fastballs, in fact, speed probably isn't that important but the difference in speed between fastball and off-speed pitches, but top speed of a fastball is the most important aspect there. Then you let each pitch rating be how hard that pitch is to distinguish, calling that "Stuff". Finally, you just give a pitcher the ability to avoid homeruns illogically, and you call that Movement

[Skipaway]

Originally Posted by mrbill

This meant a reliever could do very well with 2 high rated pitches, they'd be extremely hard to figure out on the first time around. But, the 2nd time around, the hitter would get enough of a bonus to start figuring out the pitches more often. A starter with 4 pitches would get away with hiding his pitch selection to later in the game, even though those pitches may not be as spectacular as the reliever's on the first pass.

Problem is, there aren't actual numbers or correllations of what's described here. After all, OOTP has always been about models that could generate realistic simulation results. It was never about if the model could be easily described by what could have happened in real life, but if the model could generate believeable stats.

It's good to think about ways to make the model itself closer to what could be easier explained by conventional wisdom, but it's impossible to do that unless we actually have some kind of data to establish what described above. How else are we going to come up with the algorithms to code?

[mrbill]

Conventional wisdom certainly says its easier to hit a pitcher the 2nd time a batter sees him in a game.

Can no statistics be found to support/debunk this claim?

We need the Mythbusters!

[Skipaway]

Originally Posted by mrbill

Conventional wisdom certainly says its easier to hit a pitcher the 2nd time a batter sees him in a game.

Can no statistics be found to support/debunk this claim?

We need the Mythbusters!

It's not about if the batter would have it easier the second time. It's about how is that correllated to the pitches. It is easy to just give pitchers a rating that would reflect how they would handle hitters the second time or third time around, but it's not easy to tie that to their pitches.

[mrbill]

This is true, and I'm not exactly sure that BBPro did this either, but certainly the idea can just be applied that the more pitches a pitcher can throw, the less likely the pitcher will be able to recognize the pitch on the 2nd time facing that pitcher. Or you might set it up that if a hitter saw a particular pitch the last time around, he'll be more likely to pick up on it this time.

You might generalize that to say that a hitter has X chance of "seeing" the pitch correctly. The factors that contribute to that are the ability of the pitcher to hide that pitch (the/a pitch rating), the number of different pitches the pitcher throws, and the number of times a batter has seen that pitcher this game.

This doesn't seem like a useful model until you make it a real con to have more pitches, and this would probably have to take effect in spring training, where you have to choose between a pitcher maintaining/improving 3-4 pitches or having him work on other areas of his game.

On another note, from what I remember (I really should reinstall to check out their settings file) BBPro also modeled velocity change between pitches as having an effect on batter's performance. If a hitter saw a flamethrower last time up, and a knuckleballer came in from the pen for this at bat, it'd make it harder for him to hit as he made that adjustment. In fact, I believe that even worked within the at-bat. Seeing fastballs bumped up the "expected velocity" of the next pitch for the batter, so a pitcher could "set up" a batter with some fastballs then an off-speed pitch.

I really liked that aspect, but of course, I have no idea if that's grounded in statistics, conventional wisdom, or was just completely made up.

[Skipaway]

I know there are studies showing the longer the delay between pitches, the less likely are hitters going to be fooled by speed changes. So apparently there should be more factors than just the speed differences and pitch types. And some pitchers just have different approaches in different situations, which would be yet another factor.

I think the problem would be the disconnect between the known factors and the statistical outcomes. Even if the pitch type effects are modeled perfectly, there will still need some kind of generic cover-it-all rating to represent whatever still unknown.

[mrbill]

Well, there's precedent to at least have that generalized rating, home field advantage being one.

If there's a penalty to seeing a pitcher a 2nd time, add that as a global effect. If some pitchers can be better a 2nd time around (because of # of pitches, something about their style, etc), add it as a new rating to pitchers.

I think isolating the most important features of a pitcher is what OOTP is trying to do, simplify a guy to a half dozen numbers. Movement really is aching for replacement by GB%, and "Obscuring pitches on a 2nd/3rd/4th pass through the lineup" seems a viable candidate to take its place. It could supplement endurance in deciding whether someone could start effectively.

But I totally redirected this thread...

I agree that pitch ratings have some place in real life, but I'm not sure how the complexity helps create more intricate pitchers than saying pitchers have a general Stuff and Control rating. To me, that gets in line well behind modeling starter/reliever performance more agreeably. Of course, the two would be tied, if # of pitches and/or their quality was what determined if someone could be as effective starting as they were against the first batter they saw. And the # of pitches and their quality would both be something that has a potential and actual rating. I just read today about a major league pitcher adding a new pitch to his repetoire, and that should make him more effective.

[Skipaway]

Yeah, I think giving pitchers a rating about how good they are at facing hitters multiple times would be easier to study, easier to model, and supplement/replace some of the current ratings better. There are sites already providing related stats, and we probably can come up with a reasonable scale and distribution for this rating. At least a way to distinguish starters and relievers.

As for really getting into detailed pitch type issues like the original suggestion, I surely would like that, but it's probably impossible to do realistically at this stage. Some kind of fun and made-up feature more likely.

To make it more complicated, I would believe the pitch type and pitch ratings should be something related to the groundball percentage and control. Some pitches might just be harder to control in general, and some pitchers might be more likely to induce groundballs. There are also arguments about if some pitch types would be more likely to hurt the pitchers. It's probably unreasonable to treat those as ratings independent from any other pitching ratings, therefore making it even harder to model player creation. We might again have problems similar to a lot of flyball pitchers not giving up any homeruns in OOTP6.

[mbd]

Originally Posted by mrbill

Conventional wisdom certainly says its easier to hit a pitcher the 2nd time a batter sees him in a game.

Can no statistics be found to support/debunk this claim?

We need the Mythbusters!

There is a SABR study that supports that wisdom. I love the Mythbusters but their science is suspect at times.

[mbd]

Originally Posted by Skipaway

It's probably unreasonable to treat those as ratings independent from any other pitching ratings, therefore making it even harder to model player creation. We might again have problems similar to a lot of flyball pitchers not giving up any homeruns in OOTP6.

Mogul already does this and I'm seeing realistic results so it can't be that hard. I'm 57 games into my first season all played pitch-by-pitch and the stats are realistic as far as I can tell. The pitch selection and aiming and pitch guessing by batters all work really well. That is there's no "cheat" that I can find. Mogul's biggest problem is that it doesn't record batter vs. pitcher matchup stats which is what would make this whole system really interesting.

[mrbill]

Originally Posted by mbd

The pitch selection and aiming and pitch guessing by batters all work really well.

I think that model has been around a long time, I am curious if Markus is familiar with Sierra's BBPro line and how it modeled the game, just as a reference.

I never really played around with it much (I upgraded to W2K just after I discovered it), but I do remember looking over the config file, so many tweakable options, I can imagine it was very possible to do realistically. It also made a lot of sense when you read the variables in the config file, very physical attributes, and very detailed.

[Russ]

Originally Posted by mbd

Mogul already does this and I'm seeing realistic results so it can't be that hard. I'm 57 games into my first season all played pitch-by-pitch and the stats are realistic as far as I can tell. The pitch selection and aiming and pitch guessing by batters all work really well. That is there's no "cheat" that I can find. Mogul's biggest problem is that it doesn't record batter vs. pitcher matchup stats which is what would make this whole system really interesting.

I just would like to point out that this guy is a loser.

[Dreslough]

Originally Posted by mrbill

I felt it added very little complexity, and in OOTP's case, you'd replace "Stuff" with a "Stuff per pitch" set of ratings. Although, Stuff and Movement are rather blurred on what they mean, but I imagine Velocity becoming meaningful in this situation to model pitch speed, which would mostly affect the speed of fastballs, in fact, speed probably isn't that important but the difference in speed between fastball and off-speed pitches, but top speed of a fastball is the most important aspect there. Then you let each pitch rating be how hard that pitch is to distinguish, calling that "Stuff". Finally, you just give a pitcher the ability to avoid homeruns illogically, and you call that Movement

What most scouts call "stuff" is really a combination of Movement and Velocity -- basically how hard that pitch is to read correctly, and to hit. (Although PECOTA throws Control into the formula for "Stuff" which is pretty confusing.)

So if I'm reading you correctly you have the right idea here about giving each pitch a rating for "Stuff" and then giving the pitcher himself overall ratings for Control and Velocity. We thought about giving 3 ratings for each pitch: Stuff, Control and Velocity but it doesn't make much sense. Almost all sliders are about 5-10 mph slower than the pitchers fastball. If we let the game create fictional players with sliders faster than his fastball it just wouldn't make sense. The same is true for control in that every pitcher has a harder time throwing strikes with his curveball than his fastball. That's why you see fastballs on 3-and-0 counts.

[OakDragon]

Originally Posted by Skipaway

I don't think there are enough studies about what's suggested here to come up with reasonable ways to generate players and accurately simulate them. Or you just want complicated cosmetic stuff?

I don't see a way to have this helping the game to be more real.

In fact, it's impossible. Accurate statistical simulation and meaningful pitch selection are contradictory. The stats achieved in real life depend on pitchers making certain pitches and batters reacting certain ways to them. Changing the pitches will change the statistics (assuming the computer hitting AI is at least slightly intelligent). To give a simplistic example, if someone were to pitch a game consisting of all fastballs, batters' averages would skyrocket. As user-controlled pitches become more meaningful, statistical simulation becomes less accurate. With accurate statistical simulation, pitch control is meaningless by definition.

edited to add: Of course, this is all assuming we still want our simulations to be statistically realistic. If we don't, then bring on the user-controlled pitches....

edited again to add: ... which of course will lead down the same road all other non-statistically accurate sports games go... they eventually become too easy to beat.