It was August of 2006.
I was in Vermont, visiting my uncle. He and my cousin were having a fight over something, so I was told to just sit in my room and wait it out. There wasn't a TV or computer or really much around me that could keep me occupied.
There was, however, a stack of paper over in one corner.
Being an avid fan of Baseball Prospectus and an amateur post-Moneyball stathead, I had recently been pondering how no one stat really gets to the heart of how good a player is.
Really, name any stat, and I can name plenty of things that it leaves out in its ultimate conclusion.
So with nothing better to do, I sat down in front of this stack of paper with a pen and said to myself "Let's change the world."
(I did seriously say that, I'm not saying it for some dramatic effect.)
It's been nearly two years since that day where I initially started playing with the numbers, and I finally think I've gotten somewhere.
And without further ado, ladies and gentlemen, I give you Ultimate Value Index.
Ultimate Value Index, or UVI, as I'll refer to it from here on out, is a measure of a player's "ultimate value." This might sound like a nebulous concept, but it isn't.
What UVI essentially measures is this: Without outside help, how many bases does a hitter get every plate apperance? How many does a pitcher allow?
Hitters
The formulas for UVI work on different variables, so we'll examine hitting first. UVI for hitters works a lot like slugging percentage, but there are a few differences.
1.) Slugging percentage is based on at-bats; UVI is based on plate appearances.
One of the reasons why OPS is a bad stat that skews toward sluggers is that OBP and SLG are based differently; one on ABs and the other on PAs.
2.) UVI includes baserunning.
That's right, Juan Pierre fans; this gives your boy credit for all those steals and bunts, and penalizes sluggers for all their GIDPs. The fact that UVI combines both offense and baserunning is, in my opinion, a big point in its favor.
First-order hitting UVI
There are three UVIs: first, second, and third-order.
First is simply what the raw data tells us, second is park-adjusted, and third is park and level-adjusted. Since I've briefly explained the concepts above, here is the first-order UVI formula, for all the world to know:
(Total Bases + Walks + HBPs + SB - CS + .25(Bunts + Sac Flies) - GIDP - .1K)/PA
I figured that situational hitting does matter some. If a manager wants a player to bunt, and he does so succesfully, that at least means that a runner advanced on the out. The same goes for sac flies.
I thought since no one can advance on a strikeout, the penalty of .1 bases/K was reasonable. You can agree or disagree on that, but the formula is obviously subject to change if that's wrong. Since a second out is guaranteed on a GIDP, that's -1 extra base. (If any of this doesn't make sense to you, just comment at the bottom and I'll explain).
Second-order hitting UVI
Second-order UVI adjusts for the batter's home park.
To get it, first you divide 1000 by the park factor (NOT THE PARK FACTOR BY 1000!!!). Take that number (let's call it x) and multiply it by the batter's singles, doubles, triples, and homers. Thus:
Second-order singles: (First-order singles)*x
Second-order doubles: (First-order doubles)*x...etc. for triples and HR.
Then just plug the new numbers into the first-order formula.
Third-order hitting UVI
Third-order UVI adjusts for the batter's level and home park. For MLB players, second and third-order UVI are the same.
Essentially, third-order UVI is the result of changing the following attributes: singles, doubles, triples, homers, walks, strikeouts, stolen base percentage, HBPs, and GIDPs. Yeah, that's right, I let them keep their bunts and sac flies.
Minor league translations are very much an inexact science.
I guarantee none predicted Hanley Ramirez to be so good so fast. Same with Ryan Braun. Therefore, I'm not going to give the exact numbers I use for this, because your guess is as good as mine, but I'll basically tell you how it works.
Each one of those attributes has a coefficient assigned to it. Then, to get the third-order stat, you do something like this:
Third-order singles = Coefficient^(levels away from majors)*Second-order singles
Therefore, if you use a coefficient of .9, that means an AAA hitter gets .9 of their singles, an AA hitter gets .81, a High-A hitter .729, and so on. Therefore, this tells you how many singles they would have gotten had they played in a neutral park in the majors.
Like I said, it's tough to really nail down good coefficients because of all the weird career paths in the minors, so there's no sense in telling you the exact ones I use, because they're just educated guesses. Still, a lot of the numbers, when I run them, look pretty accurate, so I'm confident that there aren't any destructive flaws in what I'm doing.
So yeah, that's hitting UVI.
Pitchers
Pitching UVI works quite a bit differently. Some major differences between hitting and pitching UVI are:
1.) Hitting UVI tells you what a player has done; pitching UVI tells you what a pitcher should have done.
It draws on Voros McCracken's DIPS theory to take the elements of luck out of pitching.
2.) Pitching UVI doesn't include baserunning against.
The main reason for this is simply because I don't have baserunning against data. Also, this is really a function of the catcher more than the pitcher, so it would skew things.
It took me the better part of two years to figure pitching UVI out to a level that satisfied me, whereas hitting wasn't too hard. It's pretty complicated, so here's how it works:
All you need to know to calculate a pitcher's UVI is the following: IP, H, HR, BB, K, HBP, and GB%. In return, my system spits out the following: Expected IP, Expected Hits, Expected WHIP, Expected ERA, Expected BABIP, Expected BAA, Expected OBPA, Expected SLGA, Expected OPSA, and Expected UVI.
Several years ago, Voros McCracken proposed DIPS theory, which states that a pitcher has no control over balls in play and that the only things a pitcher can control are walks, strikeouts, and home runs allowed.
Not quite.
It has since been found that a pitcher can also control their groundball rate, and that groundballs have a different BABIP than flies.
So first, from the variables above, we need to figure out how many plate appearances there were against the pitcher. Simply add 3*IP to the hits, walks, and HBP for that.
Then, we can break down the plate appearances into definite-outcome PA's and indefinite-outcome PAs. To do this, just subtract Ks, BBs, HRs and HBP.
For reference, we'll use Erik Bedard's 2007 line. Bedard faced 749 batters last year; he K'd 221, walked 57, gave up 19 HR, and hit 5. Therefore, Bedard had 302 plate appearances that he controlled the outcome of, and 447 that he did not.
Next, take the indefine-outcome plate appearances and multiply them by the pitcher's groundball percentage. Now, you have split it into grounders and non-grounders.
To continue with the Bedard example, his 49.8 GB% gives him 232 grounders and 235 non-grounders.
A grounder has a BABIP of .251 (Trust me, I researched this to death) and about eight percent of grounders go for extra bases. When you factor in triples to the equation, you get something along the lines of 1.095 bases per groundball hit. Multiply that by the BABIP of .251 and it comes out to .275 bases per grounder.
Thus, Bedard should have allowed .275*232, or 64, bases on grounders last year. This would come on .251*232 = 58 hits. The other 174 hits should have gone for outs.
Flies are much trickier, and caused me a ton of trouble, mainly due to the fact that all are not created equal. Liners, high flies, and popups are all dramatically different. However, I'll spare you the story of my trials and tribulations with flies and simply tell you how to get to the UVI.
First, batters hit .3606 on flies, including homers. So take .3606*the flies to see how many flyball hits the pitcher should allow.
In Bedard's case, .3606*235 = 84 hits. Then, subtract the homers allowed, and multiply the remaining number by 1.394, which, according to my calculations (that is the only time in my life I will ever be able to say "according to my calculations"), is the average number of bases on inside-the-park non-groundball hits.
In Bedard's case, he allows 65 non-homer hits on flies, so 65*1.394 = 91 bases on flies.
Got all that?
Thank God I made a spreadsheet to do all this.
OK, now we can finally get somewhere. Add walks, HBP, and 4*HR to the expected groundball and flyball bases (the 64 and 91 in the Bedard example), and you have your expected total bases allowed. Divide that by the plate appearances, and you have UVI!
Bedard's comes out to .391.
To get expected hits allowed: Just add together the groundball hits and flyball hits (the 58 and 84 in the Bedard example).
To get expected IP: (PA's - BB - HBP - H)/3
Expected WHIP: The formula is in the name, guys.
Expected ERA: I just figured four bases to an earned run, and it works surprisingly well.
Getting the AVG, OBP, SLG, and OPS from here is easy enough that I won't explain it. If you need help just ask.
Second and Third-order Pitching UVI
The only thing a park really affects that a pitcher can really control is the homers, so just divide 1000 by the park factor and multiply by the homer total. Then just rerun the stats again.
Third-order stuff is the same as with hitters; just mess with coefficients for BB, K, HR, and GB% in similar way, although this time the coefficients will be greater than 1.
UVI FAQ's
What's a good UVI and a bad UVI?
One of the reasons I like the pitching UVI so much is that the system can spit out the other numbers along with it. Not only do you have the UVI itself, but you have more familiar things to compare it to, like ERA and WHIP. Generally, if the other stats all look good, the UVI will be good as well.
For hitters, it obviously depends on the position. I've run enough UVIs that I would guess "average" is somewhere around .460 or .470. In general, below .400 is very bad and above .500 is very good. Also note that pitchers and hitters have inverse UVI's; that is, the higher the UVI, the better for a hitter, but the lower the better for a pitcher.
Does this have predictive value?
Yes and no. While I don't currently have a formula in place, predicting UVI increases until age 27, stability through 30, and decline from there is easy to see. The tough thing is getting all the other data as well to line up. It's my next big project, but no, there is currently no set system in place.
This sounds pretty interesting, but there's no way in hell I have as much time on my hands to do this as you do. Can I have the UVI spreadsheet?
Sure, just email me at stoltznh@jmu.edu. You need to have a basic knowledge of how to work Excel.
What is (player x)'s UVI?
Feel free to ask me this, but be aware that I can't get to everyone. You can always just run it yourself. If you have any ideas for a study I can do on something, or a question that you think this can solve, I might write up an article on it (just comment here or post on my board).
Where does one find HBP and GIDP and all that stuff to put into the formula?
I'd recommend The Baseball Cube for that, since they have minor league stuff through '07. Baseball Reference is good for the '08 major league data. '08 minor league data on GIDP, HBP, bunts, and sac flies can be a pain to find; check the team's website and hope you get lucky.
Why UVI Tells Us More Than We Already Knew
UVI is the only stat that exists that combines baserunning and situational hitting with pure offense to determine a batter's overall value. It also has a fair amount of in-game tactical value. If the pitcher on the mound has a .4 UVI, and the batter has a .5 UVI, we can estimate that the batter will create an average of .45 bases per plate appearance against that pitcher.
It's the only stat that really transcends the differences between hitting and pitching, so it's a useful comparable for hitters and pitchers. It also works better than VORP because it isn't a counting stat, so it's not about how much you play, but instead, simply how you play.
I hope you've found this both informative and interesting. If you have any questions about this, just comment here, post a note on my profile, or send me an email. Feel free to use UVI wherever you want—I'll even help you out if you want to do a study involving it—but give me credit for my idea.
comments (56) write a comment »
write a new comment
2 months ago
I found this interesting, very much so.
I'd kinda like to see an example though. Like does this really work? Do you have any UVIs that translate to a players success, or is this a rather new thing that doesn't have results yet?
That must have taken you a ton of time though, excellent stuff and hey if this stat catches on, you might go places.
from 2 months ago
I'm working on a followup article with examples right now, and I'll be doing tons of updates.
Yeah I've got data that looks good, only time will tell in the long run, and like I said, doing future predictions as opposed to just keeping track is my next big step--I hope to have that in place before '09.
from 2 months ago
I find myself uninteresting... not the most of all time... At least not yet.
Thanks for detracting from the article Dana
2 months ago
I like this concept. Let's hope that this makes things easier to decide the baseball MVP.
from 2 months ago
Check my followup article for a discussion on last year's NL MVP.
2 months ago
Interesting concept, but I think the premise is based on an assumption that's not always true. It certainly measures what you purport to measure, the number of bases a player gets per plate appearance. That's fine. But I think you start to run into a problem when you tie a player's "ultimate value" to the number of bases they get without appropriately weighting the methods by which they do so.
For example, all other factors being equal, a guy who has one single and zero walks will have an equal UVI to a guy with one walk and zero singles, right? Well, I would argue that they aren't equally valuable; given any random pattern of men on base, I would always prefer to have a single rather than a walk...assuming someone's on, you always have a chance to bat a runner in on a single, rather than on a walk, which only drives in a run if the bases are loaded. Isn't the guy who hits the single in that situation more valuable?
Now, I realize that a guy who's swinging for the base hit instead of walking WILL get himself out sometimes too, when a walk obviously would not. Even so, while I don't know for sure, I'm skeptical if the risk/return on swinging for the single evens out in comparison to taking the walk, which your theory implies. I mean, you can get called out on strikes looking for a walk too. Maybe it does all even out, I'm all ears if you have an explanation.
from 2 months ago
A fair point, but that's kind of what the strikeout penalty is for. If you don't get the bat off your shoulder, no one can advance on the play unless you walk. So I guess I built that in slightly differently than you would expect.
from 2 months ago
Right, but you can strike out without swinging too, so the strikeout penalty hurts both those who swing and miss or those who look for walks and take called third strikes. I mean, that's the way it should be, no doubt.
But my point is that you don't also assign a penalty to swinging and lining out when a walk would have been fine, and you also don't assign a penalty to walking with two outs when the pitcher's batting behind you and you probably should have swung instead. I don't think you can assume that they cancel out perfectly evenly to have walks and singles at a 1:1 value ratio.
from 2 months ago
I think you're right, but two things:
First, free swingers strike out less. Look at Jack Cust. He swings at the fewest percent of pitches out of the zone in the majors (that's actually true, I'm not pulling it out of my ass) and has the highest K rate in the majors. Adam Dunn, same way. Then you get guys like Howie Kendrick and Vlad Guerrero who swing at anything and don't strike out. I'm not saying it's always the case, but in general, selective hitters K more.
Second, yes, even with that, it isn't perfect, but a.) Does it make a big difference and b.) Is there a way to measure it more accurately?
from 2 months ago
I believe that about Jack Cust. His problem is not strike zone judgment, it's his atrocious contact rate. He swings at strikes, but misses them too much.
I think it does make a big difference. I wouldn't be surprised if the singles/walks value ratio should be something like 1:0.75. Over the course of a season and certainly a career, that's a pretty big deal. But again, I don't know. Maybe it is 1:1. I just wanted to know your reasoning.
As for your second question, I think there's a way to measure it more accurately, but it would take a lot of data. Maybe chart out all the outcomes of any pitches while the batter has 3 balls and map it out that way to see if it's better to swing or not. I didn't really think that one through, though...it probably has a lot to do with an individual hitter's contact rate. Just throwing that out there.
from 2 months ago
Right. Yeah, I agree it can be done, but as a college freshman, I don't exactly have all this stuff on my hands. If I did, you can bet I would make some edits there. Pitch f/x might help some with that part of the equation as well, now that more data will come out. But at some point you start forsaking some variables for others if you stray to far, and it gets unbalanced anyway.
My point on the "big difference" thing is just that if it changed Juan Pierre's UVI from .430 to .445 or even .470, it wouldn't make his contract a good call. If UVIs were all off 20 points, and teams used UVIs exclusively when determining free agent signings, it wouldn't cost them all that much economically.
I think the batting side of UVI is more flawed than the pitching side. Pitchers definitely don't have much BABIP control, but there is evidence that hitters have quite a bit. But then some have more than others. Then you'd have to look at a hit-by-hit basis, but how much of that is luck? What defense was playing--how bad were they? You can just keep asking questions all day about it. Hitter profiles are so unique that properly valuing one against another is nearly impossible to pull off flawlessly.
2 months ago
excellent idea. although this might sound stupid and unrelated I did the same sort of thing by creating a complex formula to come up with a player rating for our pong league in college (which counted sinks, saves, +/-, horse trailers, hits, unforced errors, etc) and was thinking of working on a good formula for an mlb player rating but never got around to it. nice work im interested to see who should have won the mvp last season based on your theory. I'll give you like 3 high fives if it comes up with rollins.
from 2 months ago
Look at the followup on second-order UVI. It discusses some of the names there.
from 2 months ago
Also remember this is only for offense. Defense does factor into MVP voting as well, as far as what position you play and how well you play it. I wouldn't judge MVPs based on UVI alone, but as far as non-defensive value goes it pretty much tells you who did better.
2 months ago
I hate to be a spoilsport, but this is pretty much the same as Total Average, which Thomas Boswell came up with about 25 years ago:
http://www.baseball-reference.com/bullpen/Total_Average
http://en.wikipedia.org/wiki/Total_average
from 2 months ago
OK that's similar, but it:
A.) Doesn't have K's, bunts, or sac flies worked in.
B.) Doesn't have an equivalent to second-and-third-order UVIs.
C.) No pitching stuff.
D.) It's not per plate appearance; what the hell does the denominator mean?
That's basically what I came up with the first two days I looked at this stuff sorta. This is more advanced.
2 months ago
OK, but then there's Base-Out Percentage (http://www.tangotiger.net/wiki/index.php?title=Base-Out_Percentage), which actually predates Total Average. In both cases, the denominator is Outs Made instead of Plate Appearances.
The bigger point is that UVI, while interesting, is not exactly revolutionary. Certainly no more so than the latest Runs Created formulas, Equivalent Average, or any other offensive metric from the last fifteen-odd years.
from 2 months ago
I tend to agree with you that offensively, it's just sort of a different way of looking at the stuff. But the big thing is that it
A.) Includes baserunning.
B.) Has some measure of continuity for pitchers and hitters.
C.) Is marketable (It sort of looks like batting average, so maybe traditionalists will catch on).
But OK, I see where you're coming from. Still, given how many people research this, you'd think there'd be no room for improvement left, and I think I've made some improvements over different aspects of those.
2 months ago
Why not just do this?
http://www.hardballtimes.com/main/article/how-to-evaluate-hitters/
2 months ago
Interesting thought. I like the idea. I would like to see some examples. What's A-Rod's UVI? What's Albert Pujols' UVI? What's Johan Santana's UVI?
from 2 months ago
Look at my latest article for some hitting ones. I'll do a pitching one later today.
2 months ago
Dude, I would trademark this before some company/corporation/conglomerate tries to steal it out from under you.
Well done.
2 months ago
Dude, I would trademark this before some company/corporation/conglomerate tries to steal it out from under you.
Well done.
from 2 months ago
Working on it.
And here I have proof I came up with it first.
2 months ago
Some thoughts on this:
-There's an underlying assumption stolen bases have a value similar to a hit or a walk. Unfortunately, all bases are not created equal. A home run has a different run expectation than four singles and ditto compared to four stolen bases. And then there's situational problems, sometimes a SB results in increasing the chance of a win late in close inning games. These are more valuable than 2-out SBs in say the bottom of the second inning in an 10-2 ballgame.
-Did you run any regressions? I'd like to see if UVI reflects and can predict actual runs scored. Since OPS has a high correlation, it is used often. I think Runs Created has a higher correlation (.96 I think, I'll have to look it up) to runs scored.
You should also see if UVI differential can accurately be used to predict winning percentage. This will add gravitas to your stat.
-If your goal was to simply measure the "real" abilities of individual players, you have to find a way to guage its accuracy. See if UVI moves with RC27 or if there's differences. Often stats like these can fall apart at the extremes.
-I think you're going to need to find a way to make pitching and hitting UVI equivalent, so that a higher UVI is good for hitters and for pitchers, this is just aesthetics of course.
-UVI has another failing, which is the lack of fielding. Do we really need another way to measure hitting ability? Between VOPR, OPS, RC27 I think we're covered. But, there are some issues with Win Shares
-Just a question, LD% is also important, is there a reason you didn't try to incorporate it into your pitching UVI?
Stuff I liked:
-New stats are always fun, good work
-DIPS, it's still under a lot of debate but it's good you used it for this stat
-Park adjusted and it includes MLE, awesome, that's something which isn't done often enough
-If UVI has predictability value (that is, if you can take a player's AA UVI at the age of 22 and use it to predict MLB UVI at the age of 28, there could be a great deal of hope for this stat. (Only if it has empirical value for predicting runs scored or winning percentage)
from 2 months ago
Thanks for your feedback, Marty.
To answer your question about LD% and your statement in the first paragraph, it basically boils down to this: I don't have access to the data. LD% is less controllable than GB%, and has more yearly fluctuations. What I wound up doing was figuring out the average percent of non-grounders that are liners, flies, and popups, looking at BABIP data for the three types, and estimating slugging (and consequently UVI) from that. I know that certain bases may be different as far as a runs or wins basis, but figuring out what innings players stole bases and such is just out of my league at this point. Also, wouldn't the expectancy of an SB average out? Do some players steal more late in games really, enough to make a significant difference? I guess a lot of this comes back to what I said to Eugene further up; you're correct, but having this number contain a slight margin of error doesn't cost a team millions of dollars, whereas other stats can.
As far as actually looking at the stat's effectiveness, I've been caught up with two things: college work and the A's system (I'm writing a stats column for Scout.com on the A's system, so I update my spreadsheet with all the hitters and pitchers a lot). Because of those two things, and the fact that pitching UVI especially is a recent creation, I simply haven't had the time yet. In the summer, I'll definitely be looking at a number of different things, like UVI for a team over the course of a year, etc.
Pitching and hitting UVI equivalent? Meh. If that's the worst flaw in the stat, I can live with it.
There are so many defensive metrics out there that just adding another would create more confusion. Plus, once again, I don't know where to get advanced data on a widespread basis. Yes, you can find a few players LD%, but if you can't compare them to the baseball population, what's the point of putting that in?
What does MLE stand for? Haha I'm usually good with acronyms, but have never seen that one.
But yeah, as far as the RC27, predictability, etc. you'll see me putting up various UVI tidbits quite a bit from here on out. One is already up, in fact.
2 months ago
Great formula and I like that you included stats that are other wise not used in "all encompassing" stats.
I have a few concerns:
-GIDP. While I like the idea of punishing players for GIDP, they are like RBIs in that they are based on the number of chances you have to GIDP. A leadoff hitter in the NL, with an extremely low OBP in front of him will have less chances to GIDP, and thus will be rewarded in this formula for not really achieving anything.
-Ks. I think you might be punishing too extremely for Ks, by taking away 10% of a base. I think RC takes away 3% of a base, and I think that might be more accurate than 10%.
-Pitches per plate appearance. I would love to see #p/pa included in the formula if you could do that in anyway. Maybe find the given #p/ap for a player, and divide it by the average #p/ap or something. I love the stat and I think it is too often overlooked and (maybe it's the Moneyball in me) I feel it is essential for working counts and getting to the underbelly of all teams: the middle relief.
Overall I love the stat and the proposed possibility of compatibility between the pitching and hitting aspects...Keep up the good work.
from 2 months ago
I've thought about adjusting GIDP to .25 to be compatible with the bunts and sac flies, but I'm really torn on it. It does reflect a hitter's speed and their ability to hit situationally. I'm definitely on the stats side of the metaphorical stats-scouts debate, but if you're trying to get the ball in the air to avoid the double play, that's kind of important.
Really the whole GIDP/SAC/SF/K thing was just some educated guessing on my part. It's one of the things I'd be fairly prone to editing if I got real proof there was something better.
I'm not sure exactly where pitches per PA would work in, and like a lot of these suggestions, it's a nice idea, but it would help if I could actually get the data.
2 months ago
FINALLY, a formula that takes into count the nuances of the game. The eggheads at BP rarely do (meaning bunts, stolen bases, etc). While this formula still has a few flaws (described by the posters above me), it's still the best formula I've seen to explain the game. Very, very, very well done, Nathaniel.
from 2 months ago
Yeah, I love BP to death, and I usually read a team in the annual every day just to keep my mind fresh, but to be honest I've always found some of their stats to be just that one step away from really getting where they need to be. I think this corrects that. Thanks. I'm glad you appreciate it.
2 months ago
In one word: Impressive.
Obviously your formula isn't perfect yet, but who could expect it to be. I am impressed by the fact that you were able to sit down and come up with this system. It takes into account alot of things that go widely unnoticed in the game today. I like your insight and I would love to see the UVI used some day in the big leagues.
from 2 months ago
Thanks man, it's good to hear I'm not the only one who thinks I've got a clue what I'm doing.
2 months ago
MLE=Major League Equivalencies (taking minor league numbers and converting them to reflect the major leauge equivalent)
Just FYI
from 2 months ago
Ahh yeah. Thanks.
Like I said, mine arent perfect (typically .85 or .9ish coefficients on hitting and 1.1 or 1.2 on pitching, I think HRA might be the most affected) but yeah, what's the point of a stat if it can't tell you where minor leaguers are.
It's just so hard to find something as far as MLE goes that really makes a ton of practical sense. It would take a really great AAA line for the third-order UVI to be pretty good by MLB standards, but then again, that's probably true; it's hard to say.
What I have found in looking at minor league translations that is a positive development is this:
A lot of hitters promoted midseason keep their third-order UVIs after the promotion, which means at least from a minors promotion perspective it looks good.
A lot of the hitters promoted to MLB kept similar UVIs (With the A's, Danny Putnam jumps out as an example after being called up from Midland, although both the AA and MLB lines were small-sample. There are others promoted who perform as expected as well).
But then again, some players play better in MLB than they every did in the minors (The Marlins DP combo is one example), so it's hard to really get a hold on it. Any idea if that matters, and if it does, any idea as to how to account for it? And then there's the cliche of the "athlete turning into a baseball player." You see that with a few guys every year, where they make big breakthroughs. Obviously, statistically accounting for tools is pretty hard to do, so I don't know what to do with that either. But yeah, 3rd-order stuff on both sides is really a work in progress to an extent, although like I said, there are positive signs from it.
2 months ago
I think this is really cool. I, too, feel revolutionized.
Nathaniel, remember that time we went to the Nats game and Juan Pierre played? I thought of that because you mentioned him.
from 2 months ago
Lol, you got an account :)
Yeah lol, that was what, three years ago? I was so oblivious back then on this stuff lol...
But yeah, hella fun.
2 months ago
hey im not exactly sure how you would add in a coefficient for this but do think theres any way to take into account distance above sea level (like multiply home runs by .9 if you play at coors field even though you could definitely get more in depth than that). I'm not sure how you would work this in. but think back to the years when rockies players were in the running for mvp like dante bichette and a lot of people didnt vote for him since he played half his games a mile high. although now you'd also have to take the humidor into account. i dunno just a thought.
from 2 months ago
That's what the park factors are for in second-order UVI.
2 months ago
"UVI is the only stat that exists that combines baserunning and situational hitting with pure offense to determine a batter's overall value"
What about WPA? This statistic is actually based on play by play data which means that a home run hit in the top of the ninth while the team is already winning by 9 runs is not as valuable as a home run hit in the bottom of the ninth while the team is trailing by a run.
Wouldn't you agree that WPA is actually far more successful at measuring what a player does. It even can calculate 'clutch', which idiot sports writers love.
How about WS? Unfortunately this does take into account fielding, and it is a cumulative stat, but...
WSP is a rate stat. It takes into account how a player performs in regards to his teams victories. Again, based on play by play data. Thus, 30HR out of Vlad are much more important than 30HR out of Jason Bay.
While your stat looks nice, it is based solely on what one can view. It is essentially the equivalent of on base percentage or isolate power. One with a little bit of baseball knowledge can conclude whether player A is better than player B with using your measure, but it doesn't TRULY tell us what that player is accomplishing.
Another major error, is that you ignored league effects overall. You state that all triple A players play in the same league against the same team. This is not true. Similarly, it is not true within MLB.
This is a nice article, however, without pumping in every players name over a certain amount of time, there is no telling how accurate your stat is.
from 2 months ago
Not sold on the whole "clutch" thing; since it varies so much, it's best to even it out. Why should Bay be penalized for the Pirates sucking?
Look, you can look at this stuff a ton of different ways, I like to look at it this way, if you don't, no one is making you.
How much would league effects change this? And, isn't the point of stats to level the playing field? Second-order UVI accounts for park stuff anyway.
Yes, I know I need to do stuff with it to see its accuracy; you can bet I will.
from 2 months ago
I'm not sold on clutch either, but the mainstream media gushes over it.
That said, while I disagree with 'clutch' metrics (ie saves, 7th inning + batting average), WPA does take this into account.
League effects are massive! Park effects do not take into account that NL pitcher X is facing a pitcher every time through the order where as AL pitcher Y is facing a DH. They simply take into account how the ball is hit, on average. There is a reason why when AL players move to the NL they gain a lot and vice versa.
The 'just don't look' philosophy doesn't work when you are making a claim that the stat you are inventing will 'revolutionize baseball'. As a fan of the game and of statistical analysis, I am bothered by this claim. Telling me to not look is ridiculous. That would be like Commissioner Bud Selig stepping up to the microphone after the Mitchell Report and saying, "Fans of Baseball, no one is making you come to the baseball games. Our players will do whatever they want to compete at the highest level."
Saying 'just don't look' is like saying 'my Dad is better then your Dad'. If you can't accept criticism's for a flawed system, then you might not want to publish your work on a public domain.
from 2 months ago
As for why Jason Bay should be punished for playing for the Pirates...
Thats not what we are doing here. What we are doing is rewarding a hitter for playing on a good team and making his hits count.
I am curious as to why, if as you claim you are such a BP and SABR follower, you did the following:
1. Ignored VORP, WPA, WSP?
These are all relatively common stats that measure the value of a player based on their play by play data.
2. Ignored the fact that GB%, BABIP, etc are greatly influenced by ballparks. Hell even strikeouts are!
The Hardball Times came to this conclusion about a month ago when they shared all of their split statistics and park factors. Jake Peavy, a great pitcher at home and away, has dramatic splits in his strikeout rate for his career.
from 2 months ago
Ahh, I'm an idiot. For some reason, when you said "league effects, I started thinking about the International League, PCL, Cal League, etc. because I had just been looking at minorleaguebaseball.com.
Sure, I can accept criticisms. Look at my UVI: A Reprise article; I address some of them.
As far as what to do about leagues, it's once again pretty tough. One thing to remember is that NL relievers are almost the same as AL relievers because by the time they come in, they usually face pinch-hitters instead of pitchers. And yeah, for starters it's different though. Do you know of an easy way to figure that out?
Pitchers, like all hitters, are better than others at hitting, so adjusting for that is still difficult.
I'm not about to say this doesn't have flaws. It has a ton of flaws. The point is, in creating a rate stat that's all-inclusive offensively and somewhat compatible with pitching, I've created a foundation for something that could get really interesting.
I do think that what you're saying about the NL is absolutely true, but don't you think that's pretty easy to just mentally take into account without the stat saying it? If the NL pitching UVI winds up, say, .009 lower than the AL one, wouldn't it just be easy enough to say "Hey, they on average deserve an extra .01 added." Not saying that it wouldn't be nice to have in there; I'm just saying, it isn't like that renders the stat useless and wildly incorrect. NL pitchers still get compared to NL pitchers, and likewise for the AL. So there's a bit of a disconnect between them, so what? I'm eighteen years old dude, I've got three more years of college to figure these things out.
But thanks for pointing that out--I never though about incorporating it. Don't say I don't know how to take criticism though--I had just woken up when I wrote that and I was thinking about Minors stuff, so I wasn't really 100% there I guess.
from 2 months ago
As for the other stuff, VORP is a counting stat. Good bench players get undervalued compared to mediocre starters. I'm not quite as familar with WPA or WSP, and I'm sure they have plenty to offer, but they aren't the same thing as UVI, so who's to say UVI isn't a good measure (I'll come back to this in just a sec).
GB%, BABIP, etc. may or may not be influenced by ballparks--research a month ago has yet to be proven or disproven. As for many of the issues you and others have raised about this, the question once again turns to "How can I, amateur college freshman sabermetrician, collect the data for this?" If you have a way of figuring that out, sure, I'll take a look at it. I also think a lot of that stuff could be mental--Peavy may challenge more hitters at Petco because he knows he can get away with it there, so he gets more Ks there or something. Also, is that just a few players or is it real ballpark/team correlations?
Coming back to the first paragraph, here's the thing:
Turn on a broadcast of a game, and what do they show when the player comes up? Batting average.
VORP, WSP, WPA, RC27, etc. have all failed to make it there, for whatever reason. Why not throw something else out there? It's a nice stat because it compares hitters and pitchers, and it sort of looks like batting average, so maybe if it comes at the right time, it can be marketable.
UVI has flaws, but wouldn't you rather see it than AVG/HR/RBI?
And remember dude, I read Moneyball less than four years ago, and picked up my first BP Annual less than two years ago. The fact that I could come up with anything at my age, you have to admit, is pretty damn good. It's not perfect right now, but who's to say I can't gather data over the next three years and emerge with a formula quite a bit more accurate than this one?
from 2 months ago
Nathaniel,
I'll try to respond to everything here in one post. Feel free to email me at bheikoop@baseballdigestdaily.com and we can go further with this.
To league effects.
The Hardball Times found that the move from the AL to the NL added about 2.7 runs and -2.5 in the other direction. This was the Runs that are found within the RC stat.
That said, for pinch hitters, relief pitchers, etc it isn't a matter of facing a pitcher, it is also the matter of facing lesser linups. Simply look at the 1 thru 8 (or 9) on any AL team and compare it to the 1 thru 8 in the NL. Even the worst offensive teams in the AL this year (Baltimore, Minnesota and Oakland) would have a competitive offensive team in the NL. Whereas the Nationals and Giants would look like fools in the AL.
VORP is a counting stat, you are right. And while certain managers may make errors, injuries pop up, over the big picture 5, 10, 15, 20+ years, any inflated VORP number or deflated VORP number will be essentially erased. If a player goes their entire career as a bench player, even if they are performing at an outstanding level, there is probably a reason they are on the bench.
In terms of why WSP, WPA, RC27 and VORP are not utilized more within mainstream media, the answer is simple, in fact, you answered it yourself. That is, many outlets are either not aware of the measures or do not understand them and are unwilling to understand them.
Why for example do these outlets use batting average? Because they understand it.
Why do they make mention of on base percentage? Because they can calculate it themself.
Why do they discuss slugging percentage? Because they think it represents 'power'.
That said, there is an obvious shift within the mainstream media in terms of statistics. Broadcasters are beginning to explain the 'shift', they get on base percentages place v. batting average (although a 300 hitter with a 330 on base percentage still seems better to them then a 250 hitter with a 380 on base percentage).
I think it is only a matter of time before we start seeing ISO and OBP as the two key stats on the scoreboard at any ballpark.
All that being said, the work is interesting, but I don't see it as being groundbreaking. BP has EqA for hitters and NRA for pitchers. The Hardball Times has its Bill James Win Shares Percentage and Tom Tango's WPA/LI.
And while the work is interesting, and I definitely can understand where you are coming from being a student (I'm a graduate student trying to run a blog, write a fantasy article for another website and pump out news for another) we cannot use that as an 'excuse'. The reason being, people in the industry are not going to give us a break because we have other things on our plate, so when we do put out something, it has to be of at least the quality as those who are doing it for a career (or even those who are doing it as a hobby but have a little less on their plate).
As for your hopes of the formula, that is an excellent goal and one I would not suggest you give up on. I do however, think that the formula will end up coming pretty close to that of BPs VORP (although more of a rate) or THTs WPA, WSP.
from 2 months ago
Fair enough in general. Feel free to email me at stoltznh@jmu.edu if you have any suggestions for different things, or find places to find advanced data.
And as for the article title, understand that "UVI: The Stat That Is Kind Of Close To A Bunch Of Other Stuff But Has A Few Flaws" isn't exactly going to get many reads...
2 months ago
Great article and system. I'm having a difficult time allowing myself to grasp it more completely due to the difficulty and the lack of the actual stats sitting in front of me. I think that the most troubling aspect of the UVI is actually FINDING the intricate stats necessary to calculate them.
One other thing, I don't think that article will be your last one to say, "If my calculations are correct."
Great work!
from 2 months ago
Nah, they're not hard to find. Maybe GB% for pitchers (thanks to my BP subscription I can do that), but look in the FAQ's for stuff about where to find most of the stats.
Haha, I hope it won't be the last to say that. Thanks.
2 months ago
I think the problem you are going to encounter with your hitting UVI is that hitters who line and fly out advance runners differently than hitters who ground out. That being said, a fly out to RIGHT with a runner on second is different than a fly out to LEFT. If you punish those who strikeout for not advancing the runner then you have to punish those who line out, those who fly out to left, and those who ground out in front of the runner.
All of that being said, you punish those who hit into double plays but don't penalize those who ground into fielder's choices. These are essentially hitters who grounded into double plays but either because there were two outs our because the fielder couldn't get a good throw off.
Finally, I don't know if you can call your pitching UVI accurate if you can't include baserunning. When it comes down to it, there are some pitchers that are so slow to deliver that it really does hinder a team. Tim Wakefield essentially allowed every hitter that got a single to have a double. To blame the catcher and dismiss it is wrong. Pitchers must pitch out, pick off, hold runners, and make quick deliveries to prevent stolen bases. Also, lefties are much more difficult to steal on than righties. There is a significant difference there.
Other than all of that....interesting. Can't wait to see this with actual numbers.
from 2 months ago
Agreed, but a lot of that stuff is going to even out over time. And yeah with a ridiculous exception like Wakefield, baserunning will get skewed, but then again, knuckleballers can control BABIP, so that makes it more ridiculous. Given that Wakefield, Charlie Haeger, Charlie Zink, and RA Dickey are the only knuckleballers I'm aware of in professional baseball right now, I'm not worried.
Furthermore, pitchers who don't have a ton of velocity ("crafty lefties") usually have a pretty good pickoff move to compensate. There are a few examples (Greg Maddux comes to mind) who are notioriously bad at holding runners, but once again, it's tough to find all the data. If there's a source you have that lists SBA/CSA & pickoffs for every pitcher, I'm all over it and it's in the formula in a couple of hours. But I haven't found one yet.
There's only so much you can do as far as the balls in play go. The hitter's speed is every bit as much responsible for turning a GIDP into an FC as the infield defense. And maybe the flies to right/flies to left thing is true, but that would do a number of things: penalize hitters for handedness, assume they have more control than they do, and once again, send me looking for data that to my knowledge isn't out there.
What I essentially did is take all the variables out there that I could find for pretty much any pitcher and made a formula out of them. If you can find another variable that's important and easy to access for all players, tell me where and I'll happily take a look. I'm always looking for ways to make this better, and as the article states, I've had UVI for two years now. I just finally tweaked the formula enough that it's good enough for me to disclose now. That doesn't mean there still isn't room for improvement.
If you'd like actual numbers, look at pretty much any article I've written since this.
2 months ago
I can see that you put a ton of time and effort into that, and I respect anyone willing to do that. On another note, just being a little picky, when you said no one could predict Hanley and Braun to be so good so fast, is a little tricky, because you are right with Hanley, he was a great prospect that the Red Sox really valued, and no one expected to produce offensively like he has. Braun is a different story, for the people that follow minor league dealings, it was commonly known that Braun had a MLB bat, but his glove was holding him out of the majors. So, I guess you could have expected Braun to be able to hit, with a lot of errors at the major league level.
On another hand the level-adjusting is pretty difficult, because its not only level, but the league. When a guy is consistently going up against top pitchers, in what is considered a pitchers league, his numbers are going to be lower than maybe if he is promoted to a hitters league.
The minors are a little tricky because there are so many variables at hand, but I really commend you for your research.
from 2 months ago
Oh yeah, everyone thought Braun would hit, but I doubt many expected him to lead the league in slugging as a rookie.
Yeah it is tough, because so many more things are at work in the minors. I've gotten some good results with my coefficients though. Only time will tell how everything works out.
Thanks for the compliment. It's not perfect, but it's better than not having it at all.
2 months ago
I think it's really cool that you took the time to do something like this. What I really appreciate about your formula is that it takes a lot of things into account and really tries to put a lot of things together than to just look at one statistic. I'm excited to see this statistic put into play. This reminds me in a small way of Range Factor, just because it takes a look at an aspect of baseball differently, which I think is really cool. Again great job!
from 2 months ago
Thanks a lot.
2 months ago
Nathaniel-
What's taken more time out of your life: devising this formula and utilizing it, or responding to comments on the BR threads?
from 2 months ago
Hahahahahahaha...good question. At this point, it's actually close to equal lol.
write a new comment