Poll: Overall Ratings for Pitchers

[RonCo]

Perhaps a role-based fatigue system would be useful, too. In other words, a pitch thrown in a relief role consumes more energy than a pitch thrown in a starter's role.

[battists]

Originally Posted by RonCo

A rewrite of the system to include individual pitches and dynamic stamna is the only option that makes functional sense. Anything else is just hiding the problem under a different rug.

It's the only option that makes functional sense if you agree that:

a. individual pitches are the way you want the game to go, and
b. dynamic stamina is the best representation of pitcher stamina, and
c. you agree that the current model is flawed enough to really require a change, and
d. the value gained by going with this sort of a model outweighs the value of the current model sufficiently enough to justify the amount of time that would be spent coding it

I'm not certain those are all givens, RonCo. Not saying that they shouldn't be, but I don't think it's as black-and-white as your post indicated.

Steve

[andymac]

Originally Posted by Dark Horse

i disagree with this completely. The pitcher’s ratings should not change because his role changes. And, really this is where fatigue should play a major factor. A starter after so many pitches should gradually fatigue and should rating points that way. A reliever will come in (in most cases) at full strength and never have to lose anything because of fatigue.

i know that pitchers will often times have better statistics, but its not because their stuff is SO much better. It’s the advantage of being in the role.

Fatigue & usage are both part of the equation but there is no way that they make up the entire difference of pitchers' effectivness in a starting role compared to relief...that gaps are just too large.

- Starters are expected to give teams, at least five innings. So if a starter doesn’t have his best stuff he may have to suffer through five or more innings to save his bullpen or because the bullpen is already tired. If a reliever (with the exception of closers) doesn’t have his best stuff, the manager is gonna’ come get him. A MR can come into a game and walk a batter then gives up a double. The manager can and will get him out and bring another reliever. If the next pitcher comes in and strands the runners, that pitcher’s ERA is 0.00. But a starter is not going to get that type of help until the middle innings.

I don't consider ERA when comparing the effectivness of a pitcher so strand rates and such can be thrown out. Strikeout rates are the most telling but also home run rates are worth comparing. I don't believe that relievers typically get pulled because they don't have their "best stuff". They are brought in to get guys out and either they do or don't...outings are not typically cut short (at least not by # of batters faced) based on their stuff. Sure occasionally starters "take one for the team" but those instances are rare.

- Starters have to pitch to everybody in the opposition’s lineup. And they have to do it at least twice. A manager can spot relievers to face only certain batters. A mediocre reliever isn’t gonna’ face Albert Pujols more than once. And, in that instance, he’s probably gonna’ be ordered to walk him, either intentionally or unintentionally. A mediocre reliever will often pitch more often with a platoon advantage. So if you have a mediocre LHP and he’s a starter facing the Cardinals he may have to pitch to Pujols and Rolens with a man on first and second with nobody out in the third inning. But if you take that same mediocre LHP and put him in the bullpen, he’s only gonna’ face LHBs and MAYBE some light hitting RHBs.

Again, I agree somewhat with what you are saying but I don't think it makes up the entire story like you are ensenuating. If a pitcher comes in to specifically pitch to certain players they are going to have more success against those batters than if they were starting for a number of reasons. I don't really think fatigue is one of those (at least not in general). I think they are as follows.

1. In a relief role the batter likely hasn't seen that pitcher on that day, so the batter doesn't have that advantage.

2. In a relief role the pitcher has the ability to prepare for specific matchups instead of an entire team.

3. In a relief role a pitcher does not need to conserve energy or save certain pitches for later at-bats.

I agree that the best way to implement things to reflect these things is not to just increase stuff for relief pitchers, that is only part of it. There also should probably be code to give a batter an advantage based on the number of times he has faced a pitcher during a game (as well as during a career per The Book) along with some sort of advantage given for preparation which would make more sense if there were seperate pitch types & such. However, the stuff increase is the only one of these options that seems like it could be done fairly easily without changing some parts of the code from scratch.

[andymac]

Originally Posted by battists

It's the only option that makes functional sense if you agree that:

a. individual pitches are the way you want the game to go, and
b. dynamic stamina is the best representation of pitcher stamina, and
c. you agree that the current model is flawed enough to really require a change, and
d. the value gained by going with this sort of a model outweighs the value of the current model sufficiently enough to justify the amount of time that would be spent coding it

I'm not certain those are all givens, RonCo. Not saying that they shouldn't be, but I don't think it's as black-and-white as your post indicated.

Steve

I certainly agree with all four .


*edit* - Although I don't necessarily think that individual pitch ratings are needed (as I'm sure it would cause many fundamental changes to the game engine as well). I'm not sure that the time needed to do it would reflect the improvement it would make...the key for me is getting the pitchers to develop into their roles realistically not be forced into them by a randomly generated endurance rating.

[disposableheros]

based on:

role - no
endurance - yes, but less than it is currently

so, my vote was neither.

[APujols5]

Originally Posted by battists

Honestly, I think it's very unlikely that Markus would revamp the entire pitching model for this coming version, which is what would be involved in adding specific pitch ratings, or implementing many of the excellent suggestions in this thread. This is why, in my initial posts, I limited myself to the very basic notion of removing endurance from the overall rating calculation.

But, feel free to carry on. I'm not Markus. But, I also didn't want to give anyone the impression that Markus told me "Yeah, let's tear down the whole thing and start from scratch, just tell me what you want to do."

Steve

Too bad because to become realistic this would be a good thing

[molarmite]

I voted for neither because I think ratings should have an impact, but not as great as they did in this year's version. I would much rather have Player A who has skills of 80/68/80 with a 55 endurance than Player B with 55/60/55 and 80 endurance. As of now, Player A's rating would be around a 29 while Player B's rating would be around a 58. I think that you should be able to send a player through workouts to increase arm endurance to a SP level where he can throw at least 7 strong innings. You see people all the time in the MLB switching from MR to SP, this should be accurately modeled in OOTP 2007.

I think you could also add different ratings to add to the overall ratings for a pitcher. These ratings could be like andymac suggested, arm fitness, or a mental strength to help factor how many innings he can go. Mental state of mind is a huge part of pitching, a rating like this could be added to OOTP and help portray the overall ratings a little bit better.

I also think that different pitches should have some sort of an effect. I'm not sure how much of a role they play currently, but I think that if a pitcher has a dominant pitch such as a Tim Wakefield knuckleball or a Barry Zito curveball, or a Johan Santana changeup, that it should boost a pitcher's ratings like stuff, control and movement up a little bit. But like I said, I'm not sure how this currently works so it might already do it this way.

Perhaps even instead of basing player ratings on roles, you could have a seperate rating for their speciality such as pitching to left handed hitters in one inning. You would use him if you needed a pitch for one inning against 2 or 3 left handed hitters. Or have a pitching rating for a groundball pitcher that could get you out of a jam with 1 out and a runner on first and second. I think you get my point with that whole thing.

That's all I can think of for now, but I'm sure I'm missing something.

[Curtis]

I find it interesting that over in the other thread on this topic almost no one agrees with me, but in this one almost no one disagrees.

Conclusion: Segregation still exists in baseball.

[DaddyO]

There is more than just stamina and conditioning that allows a pitcher to perform well as a starter. Basically, some pitchers do not have the pitches, the command, or the makeup to work their way through a batting order three times. Many can fool them once, but after that they cannot keep hitters off balance. They are one trick ponies. Unless you have exceptional stuff and command you can't get away with this as a starter.

The programming solution would have to be long term, after 2007, and I agree that each pitch should have ratings. Pitchers with three good pitches they can usually command, or pitchers that can change speeds well are candidates for starters. Roles have to be learned in the minor leagues, etc.

[RonCo]

Originally Posted by battists

It's the only option that makes functional sense if you agree that:

a. individual pitches are the way you want the game to go, and
b. dynamic stamina is the best representation of pitcher stamina, and
c. you agree that the current model is flawed enough to really require a change, and
d. the value gained by going with this sort of a model outweighs the value of the current model sufficiently enough to justify the amount of time that would be spent coding it

I'm not certain those are all givens, RonCo. Not saying that they shouldn't be, but I don't think it's as black-and-white as your post indicated.

Steve

The only item on your 4-point list that actually sideswipes the modelling of the game is (b). All others are a matter of effort/complexity the design team is willing to bite off.

As far as point (b), I don't think there can be much argument that stamina is developed in a starter by pitching more innings. Maybe someone will make a counter argument, but I can at least fall back on a hundred or so years of Spring Training seasons of anecdotal evidence that guys who throw 2-3 innings at a time can gain stamina enough to throw 6-7+.

I make point (a) because of how I think the game of baseball works. Hence, a valid argument for NOT modelling the game in the way I'm suggesting is if the design team feels that I'm wrong, and that a starting pitcher is NOT defined by the number of quality pitches he throws. I do belive that is the fundamental difference between a starter and a reliever--starters have three or more major league quality pitches. Relievers don't. I'm willing to have someone tell me I'm wrong here, but I don't think my position is too controversial among people who follow baseball.

If you'll give me that I'm right here, then the only prudent way to model the algorithims is on a per-pitch basis--not because I think it's the "way to go" but because modelling the system with anything of a higher order than this is going to result in confusion and a lower-quality model.

Regarding point (c), I think it's fairly obvious that the pitching model is flawed, and the mere fact that we're talking about it and that the thread has brought out so many ideas suggests that it is flawed enough to change.

As far as point (d) is concerned, the value of modelling something well is that it becomes robust and doesn't require doing it over and over again. As I think Einstein said, everything should be made as simple as it can be, but no simpler. Making the model more "simple" than pitch quality drives problems. I'm sure Markus is smarter than I am. But I've thought about this for a very long time. For the life of me I can't think of a way to model this process in a way that is directionally accurate without getting to the level of detail defined as Individual Pitch Quality. Working at higher abstraction means you have to fudge things, and fudging things leads to people saying "you know, this just doesn't seem quite right." Which is where we're at now.

Personally, rather than open this question to the OOTP public (or rather than listen to me, even), I would suggest that the design team get with some friends at a place like Baseball Prospectus or some baseball think-tank that actually quantifies baseball, and determine what the right algorithm really ought be. I mean, what the heck do I know? I'm just a weekend hobbyist SaberGuy. Why not find someone who can really help answer the question?

My .02, of course. It's probably worth the phospors I've used to write it with...

[fhomess]

The problem with moving the pitchers to be rated on a per pitch basis is that you then have a slew of other things you have to model. Pitchers now have to work with catchers and managers on pitch selection. Batters have to be modeled with the ability to recognize different types of pitches. Some hitters have trouble with breaking stuff but can smash fastballs. Others can hit all of them equally well. Then you'd get into pitch sequencing. The order and location of pitches affect a batter's ability to hit it. A 78 MPH changeup that is immediately preceded by a 95 MPH fastball is next to impossible to time correctly. It opens up a huge area of development that I think would be good for the game, but I don't think OOTP2007 is the right time to add it. All of this would add tremendous complexity, which means there'd be countless opportunity for things to go wrong. There's too many other things to fix for Markus to incorporate this at this point in time. This is something that would need to be planned well ahead and plenty of time and energy devoted to understanding it properly.

Don't get me wrong. I think it's a great idea in the sense of creating a realistic baseball sim. I'd love to see it in there, but I would want to see it done right.

[scefalu]

Originally Posted by fhomess

The problem with moving the pitchers to be rated on a per pitch basis is that you then have a slew of other things you have to model. Pitchers now have to work with catchers and managers on pitch selection. Batters have to be modeled with the ability to recognize different types of pitches. Some hitters have trouble with breaking stuff but can smash fastballs. Others can hit all of them equally well. Then you'd get into pitch sequencing. The order and location of pitches affect a batter's ability to hit it. A 78 MPH changeup that is immediately preceded by a 95 MPH fastball is next to impossible to time correctly. It opens up a huge area of development that I think would be good for the game, but I don't think OOTP2007 is the right time to add it. All of this would add tremendous complexity, which means there'd be countless opportunity for things to go wrong. There's too many other things to fix for Markus to incorporate this at this point in time. This is something that would need to be planned well ahead and plenty of time and energy devoted to understanding it properly.

I think you nailed it. Doing this right would require more research than can be done prior to the next release. This would have to be planned one or two versions ahead of its introduction.

[RonCo]

I doubt that m(any) of us knows how long it would take to make a model at this level. I could be easier than you think. In fact, data of this type could already be available. I strongly suspect it is. I know I could write code to do this generically pretty quickly (as in days, not weeks), so if there's data available...

The most important part of my suggestion was the last part.

[fhomess]

Originally Posted by RonCo

I doubt that m(any) of us knows how long it would take to make a model at this level. I could be easier than you think. In fact, data of this type could already be available. I strongly suspect it is. I know I could write code to do this generically pretty quickly (as in days, not weeks), so if there's data available...

The most important part of my suggestion was the last part.

Yep. They need to sit down with some smart people and get it right.

As for how long it would take to code properly, it would be quite a long time. I've done some coding myself, and sure, it would be easy to get a generic model in a few days. Getting that generic model ready to be shipped as part of the rest of the game would take weeks.

[Syd Thrift]

I pretty much agree with andymac 100% on this as I have in the past. All I have to add is, if Markus is going to continue the silly "only half of all pitchers have the stuff to be starters" thing WRT endurance, of course you need to differentiate between role. For one thing, endurance is a huge factor in determining your starters. For another, it doesn't matter pretty much at all in determining relievers. You really do need to know how your 6th starter ranks against the other relievers in terms of his ability to pitch well for an inning, not 6.

[ctorg]

I'm glad to see this being discussed. I really like andymac's ideas. I've long found the whole pitcher classification system in the game to be one of it's bigger flaws. Pitchers, from what I can see, don't have something that would equate to a pitch count endurance. They can lay it all out for one inning and be superb, or they can pace themselves over six innings, in which case they usually aren't quite as good.

It would be nice if someone with access to the proper stats could run a study to see how the average pitcher behaves differently as a reliever. It seems to me that a pitcher's effective stuff rating should go up, and his effective movement rating should do the same, to a lesser degree. I think effective control would be about the same. I say "effective" because I don't think the numbers should actually change for the guy. It's just that situationally, he'd get bonuses/drawbacks. Or maybe pitchers should have two sets of ratings, one as a starter and one as a reliever. Maybe that's going too far.

Good starters usually make great relievers. Good relievers can sometimes make pretty good starters, but they often don't have enough variation of pitches to be effective for long, and pacing themselves can take something off their pitches that they relied on. I'm not sure how the game would take that into account.

[CBL-Commish]

Originally Posted by RonCo

I doubt that m(any) of us knows how long it would take to make a model at this level. I could be easier than you think. In fact, data of this type could already be available. I strongly suspect it is. I know I could write code to do this generically pretty quickly (as in days, not weeks), so if there's data available...

The most important part of my suggestion was the last part.

Coding it is one thing. Making it match up to realistic recent data is another level.

But that's just the tip of the iceberg. What effect would this have on past seasons? Real players? Deadball era replays? What about players whose pitch selection changes over their careers? What about guys who get called up to the majors from AA ball and take three months or a year to adjust to major league pitching? You'd need to be able to model aging and development based on how players throw or hit individual pitches.

To get individual pitch effects in the game you'd have to make up an awful lot of stuff for anyone who hasn't played very recently. Making roster sets would probably become exponentially harder. For anything outside of players with the detailed data available (with probably only encompasses very recent major league players) you'd have to figure out ways to fudge everyone so their in-game results looked realistic. Lahman imports would further deviate from reality, as you've added some very key database fields that don't exist.

I think this is one of those things that falls into "yea, it might be nice, but it's very unwieldly to implement."

[ctorg]

Originally Posted by CBL-Commish

Coding it is one thing. Making it match up to realistic recent data is another level.

But that's just the tip of the iceberg. What effect would this have on past seasons? Real players? Deadball era replays? What about players whose pitch selection changes over their careers? What about guys who get called up to the majors from AA ball and take three months or a year to adjust to major league pitching? You'd need to be able to model aging and development based on how players throw or hit individual pitches.

To get individual pitch effects in the game you'd have to make up an awful lot of stuff for anyone who hasn't played very recently. Making roster sets would probably become exponentially harder. For anything outside of players with the detailed data available (with probably only encompasses very recent major league players) you'd have to figure out ways to fudge everyone so their in-game results looked realistic. Lahman imports would further deviate from reality, as you've added some very key database fields that don't exist.

I think this is one of those things that falls into "yea, it might be nice, but it's very unwieldly to implement."

A couple of years ago I posted that at some point, Markus (with help from us, I'm sure) will have to make a decision about this game. Will it be the best game at creating a fictional but realistic universe or will it limit that by continuing to make itself compatible with real life data and history?

There are many people who play the game primarily for the historical sim aspect, but of course there are many other games that are better than OOTP for doing accurate historical sims. But you don't want to lose those people. I myself love being able to import from Lahman for hypothetical leagues of various kinds.

For the game to be able to meet its full fictional potential, though, it may at some point have to give up on being a great historical game. Hopefully there will be a way to automate things to make the game make sense for both, but I think the game's strength is its fictional abilities, so it needs to do what it can to create a believable model of reality and worry later about how to get historical data to work with that. Perhaps there is a way. I hope so.

[thbroman]

Originally Posted by ctorg

Good starters usually make great relievers. Good relievers can sometimes make pretty good starters, but they often don't have enough variation of pitches to be effective for long, and pacing themselves can take something off their pitches that they relied on. I'm not sure how the game would take that into account.

One obvious device for doing this that occurred to me would be to produce an "effectiveness rating" based on the number of different pitches that the pitcher can command and his control. This would yield a number that is then divided over the number of innings in his role. A closer would get the full number for 1 inning and then start losing it fast. A set-up reliever would get it for 1+, a middle reliever for 2+ or 3, and a starter for 6+ (depending on conventions of the historical age being played). I'm sure there are lots of objections to the specific proposal, but I offer it only as one way of reproducing the "pour-it-on" versus "pacing" mode of pitching.

Originally Posted by ctorg

A couple of years ago I posted that at some point, Markus (with help from us, I'm sure) will have to make a decision about this game. Will it be the best game at creating a fictional but realistic universe or will it limit that by continuing to make itself compatible with real life data and history?

There are many people who play the game primarily for the historical sim aspect, but of course there are many other games that are better than OOTP for doing accurate historical sims. But you don't want to lose those people. I myself love being able to import from Lahman for hypothetical leagues of various kinds.

For the game to be able to meet its full fictional potential, though, it may at some point have to give up on being a great historical game. Hopefully there will be a way to automate things to make the game make sense for both, but I think the game's strength is its fictional abilities, so it needs to do what it can to create a believable model of reality and worry later about how to get historical data to work with that. Perhaps there is a way. I hope so.

Amen.

[RonCo]

Originally Posted by fhomess

Yep. They need to sit down with some smart people and get it right.

As for how long it would take to code properly, it would be quite a long time. I've done some coding myself, and sure, it would be easy to get a generic model in a few days. Getting that generic model ready to be shipped as part of the rest of the game would take weeks.

I suppose we could share resumes, I've done a bit of coding in my life, too.

Writing this algorithm isn't that hard. The only thing I think is in question is the availability of the data. But since there's a bunch of data available publicly today to support the idea, I would be shocked if even better data weren't available from "experts."