2035 BBA Pitcher Ratings (And their stats)

[RonCo]

NOTE: All this is done with data from late July 2035.
--------------

Look, there are a lot of problems with the study I’m about to show you. A pitcher is a composite of all his skills, and looking at just one at a time is always dangerous. Beyond that, a pitcher’s stats are influenced by what batters they face and where they play. Who their catcher is. Probably several other things, too—not the least is that these ratings (as you’ll see) were taken at the end of July, and since ratings change, the stats could be misleading.

And that's before we get into the entire issue of platoon advantage and splits.

So, yes, this is deeply simplified. I get it, and so should you.

Still, there’s perhaps some value here, and if you follow me and the process, you can do the same thing in your own ways to dig into and answer some of those questions all by yourself, if you so desire.

WHAT I’VE DONE

You can do this, too. It’s easy.

Step 1: Go to “Sortable Stats” and make a view that has Pitcher Ratings, and the stats you care about (for this study it was STU, MOV, CON, BF, K, HR, BB).
Step 2: Sort that view on pitchers
Step 3: Open the report in a browser.
Step 4: Copy that report and paste it into Excel
Step 5: Make a quick pivot table with the proper columns.

Easy-peasy.

When you’re done you can create all the tables of data you want. Get creative, even. Look at billions of things, and post weird features of your own.

WHY I DID IT

In this case, I was trying to show something that would suggest what the value is of various pitching ratings in the BBA environment of 2035 (when you go to the OOTP boards you’ll see people looking for a universal constant, but the problem there is that the rating environment of every league is different and that this rating environment changes as time progresses). The question I asked was: How do each of the big three ratings perform regarding to the Three True Outcomes.

So, let’s take a look.

STUFF

Here’s the data:

First, how to read the chart. In the case of Stuff, we see that as of the time I took this data (end of July) 500 batters have faced pitchers with a Stuff rating of 4. Of those, 49 struck out, 19 hit homers, and 20 walked (meaning 412 either got hit, intentionally walked, or put the ball in play. Across the rest of the columns we find this data converted into rates/BF.

Now, what does this tell us.

Well, look at those curves. With each point of Stuff, strikeouts rise with almost perfect linearity. HR rates don’t budge. Walks rise very gently. So, we can say clearly that STU creates strikeouts, and STU has zero effect on HR. The interesting question is whether high STU pitchers, by definition walk more batters. This data says yes, but the next question (why is that happening?) is perhaps more interesting.

I say that because the rate of walk change is low, possibly not even statistically significant. I also suggest a likely cause has nothing to do with stuff directly, but that it’s likely that OOTP creates players with a correlation between STU and CON. We’ll see what CON does shortly.

MOVEMENT

Here’s the same charts for MOV.

Holy sample size, Batman. Look at that weird bump in K/BF for 3 MOV pitchers. Yes, at only 483 BF, I would chalk that up as an anomaly of the FAC (Fancy-Assed Calculator).

Otherwise, though, what do we see here? Strike-outs seem to actually fall a little as movement increases. Again, this could be within the nature of randomness (the curve is not smooth), or it could be that MOV actually reduces STU (unlikely in my mind), or it could be that OOTP player creation drives STU/MOV pairs. You tell me.

Walks might rise a little, but I’d be shocked if that was significant. Then, of course, they fall in the 396 batters faced by a “9” guy. Go figure.

Where we see a linearity, though, is in HR rates. HR/BF rates are nearly a perfect line from .014 (9 MOV) to .61 (2 MOV). I think it’s fair to say MOV has a direct influence on HR. Note, I don’t have GB% in this data set. Knock yourself out. [grin]

CONTROL

Finally, here’s the CON chart.

Again, note the weirdness in the rate data for the far ends (CON =1, 2, 11). But, when we squint our eyes around those issues, we can see that K-rates fall a little as control increases, that HR rates seem to rise just a little (the span of change …. From .24 to .43 … is considerably less than the span of change covered by MOV). So, one might reasonably ponder the question of whether STU and MOV directly affect the results engine, or whether that effect is achieved in other ways. Not that it matters, perhaps. Dunno.

The big take away, though, is, of course in the area of BB/BF, which is, again, quite linear through most of the zone, and quite pronounced. There can be almost no argument to the idea that CON’s primary purpose for a pitcher is to control walks.

Well, duh, right?

HOW WOULD YOU USE THIS?

Heck, I don’t know. Don't ask the hard questions, all right? You could print it out and use it as tinder as far as I can tell. Or you can point at a person silly enough to do all this and laugh at him. That seems appropriate.

You could look at these charts and realize that they are both flawed and snapshots in time—meaning that the data is dirty in ways I’ve mentioned before, and that the data is reliant upon the ratings distributions in the league as we know it today—it has changed before, and will again.

But you can also look at it and do some back-of-the-napkin calculations for valuations of pitchers as we know and love them today. For example, I’ve fairly often heard people around here say that a point of Movement is worth more than a point of Stuff. Is that true?

I don’t know.

Let’s do a quick calculation.

Say I’m a 6/5 STU/MOV pitcher, and I gain one point of each.

K-rate: .128/BF >> .143/BF = .015 K/BF Improvement
Current HR rate: .041/BF >> .035 HR/BF = .006 HR/BF Improvement

If I start 30 games and face 25 hitters a game, that’s 750 batters. So, doing that math:

K: 96 Ks increased to 107.25 (11.25 more)
HR: 30.75 HR reduced to 26.25 (4.5 less)

In my mind, the value of 4.5 fewer HR is almost certainly more than the value of 11.25 K’s.

How about another?

Let’s do that same study if I’m an 8/7 STU/MOV guy.

K-rate: .170/BF >> .195/BF = .015 K/BF Improvement
Current HR rate: .029/BF >> .024 HR/BF = .005 HR/BF Improvement

Doing the math:

K: 127.5 increased to 146.25 (8.75 increase)
HR: 21.75 reduced to 18 (3.5 HR reduction)

Again a point of MOV is probably worth more than a point of STU, but the gap is closer.

Now do the same study for a 9/7 guy vs, a 10/6 guy. The point here is that the valuing of a point of any of the three main ratings is not cut and dried. There is no cookie cutter answer. And then, the question can be made in a different way. If you’re a 6/6/6 pitcher the value of a point of movement might be the difference between a career and no career, whereas the value a point of movement to a 7/8/3 guy is not nearly as valuable as a point of control.

So, is a point of MOV worth more than a point of STU?

Yes.

Maybe.

Often, anyway.

I don’t know.

You tell me.

[niles08]

This is really interesting stuff digging into this. I think it also shows that 6 movement is absolutely the new norm based on the batters faced although 7 isn't terribly behind I suppose.

[RonCo]

Looking at "norms"...

Another interesting piece of data regarding the league. Looking at an average at bat, (again, not taking into account platoon advantage/disadvantage, which has to be included to get "right," but which I don't have in this data) the league average hitter is facing a pitcher with the following ratings:

Stuff: 8.407
Movement: 6.254
Control: 7.129

So an 8/6/7 is your basic, solid major leaguer. Dropping below that is still workable at times, but you're not going to win championships with guys at levels below this if you're using them in primary roles (they can fill gaps, though). Alternately, if a guy has a real hole (say a 4 Control...or, well, pick your poison), he had probably be outstanding, or at least above average, in other areas. Assuming that's true, that pitcher can probably play just fine--though he might be frustrating here and there.

[RonCo]

Stuff: 8.407
Movement: 6.254
Control: 7.129

Another way of looking at this is that given the performance rates in the study, a high 7/5/6 could perform very much like a league average pitcher (especially if they have favorable splits for their usage pattern). So, really, it's possible that a 7/5/6 could provide valuable innings, but it's a risk and if you get pushed into using someone like this, you have to pay very close attention, since it's most likely that the guy is a AAAA kind of player at best.

You won't know until you try, though.

[agrudez]

The two main potential correlations of primary stats with non-primary outcomes (higher STU = higher BB/BF and higher CON = higher HR/BF) seem easily explainable to me.

If you strikeout more batters you face less batters and therefore a constant BB/9 equates to a higher BB/BF.

If you walk less batters you face less batters and therefore a constant HR/9 equates to a higher HR/BF.

Other than that, it seems you've cracked the code that STU=Ks, MOV=HRs and CON=BBs. (/sarcasm, though good natured) It is interesting to note the relational production vs. rating in each pairing, though. (/not sarcasm) I wonder how many people will look at this and say that MOV isn't as important as STU/CON because the delta from low MOV to high MOV in HR/BF is smaller in amplitude than the other 2 stats. I mean, just look at how much flatter that curve looks! (/also sarcasm, a little ill natured this time, though not directed at you)

[RonCo]

The two main potential correlations of primary stats with non-primary outcomes (higher STU = higher BB/BF and higher CON = higher HR/BF) seem easily explainable to me.

If you strikeout more batters you face less batters and therefore a constant BB/9 equates to a higher BB/BF.

If you walk less batters you face less batters and therefore a constant HR/9 equates to a higher HR/BF.

That feels true, but but it isn't. That logic is true for non-HR hits allowed, though--because the hits allowed value comes from the hitter's BABIP rating (mostly) and that will only come into play after the three true outcomes are determined.

Other than that, it seems you've cracked the code that STU=Ks, MOV=HRs and CON=BBs. (/sarcasm, though good natured)

One reason I wrote this is that I'm aware that we have several GMs who didn't know it.

It is interesting to note the relational production vs. rating in each pairing, though. (/not sarcasm) I wonder how many people will look at this and say that MOV isn't as important as STU/CON because the delta from low MOV to high MOV in HR/BF is smaller in amplitude than the other 2 stats. I mean, just look at how much flatter that curve looks! (/also sarcasm, a little ill natured this time, though not directed at you)

That's the value of the little exercises. Converting those rates to numbers (would you rather strike out 11 more hitters over a season or give up 4.5 more homers?) allows a person who follows baseball to make a more quantitative decision. That said, in the game engine itself, the value of a strike out includes a reduction in hits (due to not getting to the hitter's BABIP rating) and a reduction in fielding errors (for the same reason).

So the equation there is 11 Ks, 3 less hits, and maybe a half an error, vs. 4.5 HR. Still a value to movement, but it's not as huge as some make it.

There are hundreds of "what if" games you can play here, and not all of them come out with movement being all-powerful

[RonCo]

Likewise, if you give up fewer HR, you'll give up a few more hits (because the results engin gets to the BABIP determination more often.

Again, I agree movement is the most useful rating for pitchers (you can see that in the FIP calculation, too), but I don't agree that pitchers with 6, 5, and even 4 movement ratings can't be just as successful as guys with holes elsewhere. I admire movement, but I do not worship at its church.

[agrudez]

Since you mention the FIP equation as something you place faith in (as do I), let's look at it in relation to your findings. As you know, the relational weighting of the components in the FIP equation are 1K : 1.5 BB : 6.5 HR - meaning that, (I guess in Bill James' opinion) 1 HR is 6.5x more detrimental than 1 K is beneficial. So that means for a '6' STU to hold the same weight as a '6' MOV, as an example, the delta from 5-6 STU needs to be 6.5x that of the delta from 5-6 MOV, right?

Since 5 STU was an outlier (ie. less than 4 STU) I'll take the average of 4/5 STU to be more generous to it - making that value .0935 and the delta from it and .128 at '6' STU = .0345. For MOV, a '5' gave you .041 and a '6' gave you .035 and, thus, a ratio of .006 - making the ratio of the two deltas 5.75. According to FIP, you needed to get this delta to 6.5 for the two to be equal. Thus, a 6 STU, 5 MOV pitcher is inferior (88.5% the production) to a 5 STU, 6 MOV pitcher (assuming their CON is equal, of course) according to the relational weightings of the FIP components correlated with your findings.

[RonCo]

Since you mention the FIP equation as something you place faith in (as do I), let's look at it in relation to your findings.

I wouldn’t say I have “faith” in FIP. I use FIP as one of several valuable tools to make decisions. I try to remember that FIP’s weighting in the MLB is done around the MLB world and the MLB’s run creation environment—which will not be the same as ours. I’m certain statisticians employed by every MLB team have their own proprietary view of FIP-like stats. FIP is a better projection of future value than ERA, but it's no guarantee and it has flaws and limitations.

That said, yes, I like your math there. The point is that using that approach, you can plug in a whole bunch of scenarios and get varying answers--which help you understand risk and how it varies in each situation.

Another way of looking at your example of 5/6 and 6/5 pitchers is, since the league average pitcher (giving the study the benefit of the doubt) is 8/6/7, then that 5/6 pitcher is league average in movement, but three points below league average in stuff, whereas the 6/5 guy is below league average in both categories. So, it shouldn’t be too surprising that the 5/6 guy is more valuable than the 6/5 guy even before you get to FIPing.

My main point is to put universal rules of thumb in their proper place. They are good places to start your thinking from, but they are not … um … universal. Regardless of how you see FIP and whatnot (can we say WAR?), looking at relative weightings of value in different ways can give a GM a better feeling for what kinds of risks they are taking in any particular situation.

[RonCo]

Also, regarding FIP (and WAR) in the OOTP world, I trust that Markus and Matt have made ballpark modifications that are directionally appropriate, but I do not trust them to have gotten them "right" with regard to the final number. FIP- (and WAR) are ballpark adjusted stats, so you'll not get the answers "right" by just plugging in numbers. Giving up 40 homers in Coors is different than giving up 40 homers in Philadelphia.

[Ted]

I've tried to read most of the comments, but if I missed this and an being redundant, sorry. Ron, you mentioned some things about how OOTP develops rating pairs and I'd have to say from an observational standpoint, the walk rate rising with increasing stuff and the K rate decreasing with increasing movement, and the HR rate increasing with increasing Co, etc are all artifacts, and not related to the actual ratings. This is for two reasons:

1) movement tends to decrease with velocity increases as players develops, there's something going on there. Simmilarly, control does the same.

2) MUCH MORE IMPORTANTLY, the inverse relationships you're observing between a rating and it's non primary outcomes are almost certainly due to the players WE ARE SELECTING TO USE. Basically, there are a ton of very high stuff pitchers wtih garbage control that we're trying to get by with based on their elite strikeout ability. Similarly, there are control guys with bad stuff or bad motion. We won't use a 6/6/6 pitcher, but we will use a 6/8/6 pitcher. Essentially, we have selected this data set. Any time you get to the high end of the curve fpr any one rating, you're going to see the effects of the other two ratings being bad, simply because those data points represent flawed, but usable pitchers.

An interesting sample would be to do this again with guys with "flat" ratings, although sample size will be an issue and the "high seven/low seven" will be an issue. To do it properly you;d have to also assign median values for each stat and scale for standard deviations. In that light, the average starter would be the 8/6/7 guy you mentioned or something close, and you could just look at all the x/6/7 guys for stuff, the 8/x/7 guys for motion, etc. You'd then have to created data sets for steps up and down to next step guy, which is probably something like 9/6.5/7.5 going up and 7/5.5/6.5 going down and find a way to integrate all that data. However, I bet when you did this you'd find linear relationships between a rating and it's primary outcome, and essentially no relationship between it and the other outcomes.

[RonCo]

I bet when you did this you'd find linear relationships between a rating and it's primary outcome, and essentially no relationship between it and the other outcomes.

This is pretty much exactly what you would see.

Which is really the point of this study. Theoretically, you would see a small interaction between movement and Ks as I understand the world today.

I've been testing the player development and creation algorithm for 15 years, and have been involved in much of the design work in the past. As I've told Kyle a few times now...I do know how this part of the engine fundamentally works. I'm not going to write the specifics because that would almost certainly violate my NDA, but this data is repeatable league over league, year over year. The specific values change based on our ratings (and yes selection bias--which is always wrong [see the absolute love of movement over all things...see Contreras, Alfredo]).

I've done the studies you've discussed on hundreds of test leagues. I've written massively detailed scripts to test development patterns of players in all but the last version (I just ahven't had time). I've shared almost everything I know to the level I can share it by doing things like this--use data that every person in the league can get to, and give step-by-step instructions on how to do it.

So...yeah. You're right.

I probably shouldn't have been so obtuse in my write up.

[Ted]

Didn't mean to seem dismissive, if I came across that way. I still appreciate seeing the data, and the work you did. I just figured that since you hinted at it, I'd continue the logical path you started.