Assessing (real life) baseball player value

[Buccaneer]

In the card forum, someone started a thread about how Jim Rice should be an easy HOFer (off-year next year) and that Robin Yount is perhaps the most overrated HOFer. Without arguing whether Rice or Yount deserves to be in or not, here was one of his responses:

I am familiar with OPS and personally think it's a bunch of sabermetrics BS. To me, there's no better gauge than BA/HR/RBI. You can cut up stats, adjust for ballparks (ridiculous idea), or do whatever, but if you look at the history of baseball, look at the greatest hitters throughout the different eras, you'll find that BA/HR/RBI is as good a gauge as anything. Remember, Bill James invents statistics to sell books... People have been following and rating ballplayers based on BA/HR/RBI as far back as you can go, and not because we're any more mathematically skilled at assessing player value and can come up with "more accurate" statistics. It's because Bill James wants to sell books. Which is not to dismiss it as meaningless, but is to reject the idea that BA/HR/RBI is in somewhat insufficient in judging value.

Does anyone here even remotely agree with this? (This is probably a bad place to ask since most of us here are stats-heads anyways.)

[Chubby]

that's stupid, look at HR/BA/RBI without looking at park effects or the mound height etc? dumb dumb dumb.

[Crapshoot]

In the card forum, someone started a thread about how Jim Rice should be an easy HOFer (off-year next year) and that Robin Yount is perhaps the most overrated HOFer. Without arguing whether Rice or Yount deserves to be in or not, here was one of his responses:


Does anyone here even remotely agree with this? (This is probably a bad place to ask since most of us here are stats-heads anyways.)

No - because he argues that the idiocy of past judgement is reason enough to continue along those veins. OPS correlates far better with run scoring than any of the measures above, and there is a better predictor of runs - which is the nutshell of offense.

[dawgfan]

What a maroon. Fortunately, his kind is slowly but surely being replaced by those of us raised on Bill James that understand how to use statistics in a more meaningful way to understand baseball.

I've maintained for a long time that a college education is incomplete unless it requires students to take a statistics and probability class. There are a lot of people out there that make uninformed decisions because they just don't understand statistics and probability and how to apply them in their life.

[LastWhiteSoxFanStanding]

Which players in the past have had good BA/HR/RBI totals but poor OPS's? I would be curious to know.

[dawgfan]

Which players in the past have had good BA/HR/RBI totals but poor OPS's? I would be curious to know.

Well, if you have good BA and HR numbers it's almost impossible to not have good OPS numbers, but depending on how much the guys walks he could have significantly lower OPS numbers than another guy that had similar BA/HR numbers but a lot more walks, i.e. Andre Dawson vs. Frank Thomas.

[LastWhiteSoxFanStanding]

I know you just came up with those players off the top of your head so I'm not purposely being a jerk or anything but if you look at the stats of Dawson and Thomas. Dawsons 162 game average was 279/29/98. Thomas is 308/37/121. So again you can use these numbers and not OPS to differentiate between the two.

I really am curious is if theres a case where OPS can reveal something that BA/HR/RBI can't.

[LastWhiteSoxFanStanding]

To answer my own question:Juan Gonzalez 162 game average compares of that to frank thomas 295/42/135. However his OPS is 904 while Thomas is 996. And I think that is a significant difference.

[ISiddiqui]

I think the concensus is the card forum poster is an idiot who can't get rid of his preconcieved 'traditional wisdom' outlook. While those with high BA/HR will tend to have high OPS, some with lower BA may surprise in their OPS numbers. The problem is that the original stats for baseball were created by Henry Chadwick, a Brit, who really didn't know the intricacies of the game. Therefore he though walks were a waste and didn't count them and thus Batting Average being hits / at bats.

James, obviously, doesn't invent stats to sell books (in the beginning, what like 30 people bought his abstract). OBP is far more relevent than BA. BA is basically meant to give a number that shows how often a guy can get on base. Obvious, an On Base Percentage that takes walks into account does a better job. And SLG is a better measure of power than just HRs. SLG for one also takes doubles into account. OPS is far superior.

If you want to comprise, you can do the whole BA/OBP/SLG. But he probably wouldn't like it.

And don't get me started with RBI... that greatly depends on the team you have around you.

[dawgfan]

I know you just came up with those players off the top of your head so I'm not purposely being a jerk or anything but if you look at the stats of Dawson and Thomas. Dawsons 162 game average was 279/29/98. Thomas is 308/37/121. So again you can use these numbers and not OPS to differentiate between the two.

I really am curious is if theres a case where OPS can reveal something that BA/HR/RBI can't.

Just for clarification I was thinking of individual seasons rather than career averages.

What OPS reveals that BA/HR/RBI doesn't is a lot actually. BA is much less useful than OBP which is half of OPS. BA and HR are a big part of the SLG equation, the other half of OPS, but not all of it.

Let's say two players have identical BA/HR/RBI lines. What you still don't know is how often the player got on base (i.e. didn't make an out), which is a more critical measure than BA alone. Say player A walked rarely while player B walked a ton - player B carries more value.

The other big difference is that OPS is a rate as opposed to a total. Let's say a guy misses 30 games with an injury - his totals will suffer (i.e. HR and RBI from your example) but it doesn't affect his rates (i.e. OPS).

OPS obviously isn't the only way to rate a player's effectiveness - it's a simple tool. A more effective version would be 2 parts OBP and 1 part SLG since OBP has a greater correlation to runs scored than SLG. And totals do matter - a guy that's in the lineup 150 games a season has some advantages over a player in the lineup only 120 games a season.

RBI only tells part of the story as well - a player hitting cleanup in a high-powered lineup and with guys in the first 3 spots with high OBP is going to have more opportunities to drive in runs than a guy hitting 6th in a weak lineup with mostly low OBP guys hitting ahead of him. A better gauge is to look at how well a player converted his RBI opportunities.

[John Galt]

Tony Perez was a recent player that epitomized this split based on using sophisticated stats (although his HR total sucks under any system for a 1B).

http://www.baseball-reference.com/p/perezto01.shtml

[SackAttack]

The thing about OPS is that it's effectively tracking two stats:

1) the rate at which the player gets on base

2) How many bases his at-bats will be worth to his teams. A player with a SLG of .500 (which is half of the OPS equation) is advancing himself an average of a base every two at-bats. He also advances any runners already on base, although OPS doesn't do a very good job of tracking how far. That .500 SLG is basically worth a single every two at-bats (or a double every 4, or a HR every 8, however you want to calculate it, but let's stick with singles for the purpose of this paragraph), which is fine. But if you have a runner on first who takes third on the single, or on 2nd who scores on the single, OPS isn't factoring that in.

That's what we have other stats to track, I suppose, but I think my ideal stat would include an "bases advanced per runner" metric along with OBP and SLG. RBI helps with this to a degree, but only a degree: it tracks the runners that score on the player's hits, but doesn't do anything to track runners who advance from first to second, first to third, or second to third in the process.

I think OPS is more valuable than BA/HR/RBI, but I don't think it's as complete a statistic as it could be, either.

[Arles]

I agree with Josh here. OBP seriously overvalues walks, IMO. I was having this discussion in OOTP regarding Jose Cruz Jr. Over the past month, he had an OBP of .400 - which made most at OOTP say he was doing very well. Yet, he had an average of .180 and a SLG of .330 - but walked a ton. Walks, while nice, do not score runs (unless the bases are loaded)- they simply give you a "mulligan" of sorts and put the onus on the next player to get that run in. I'm not sure how desireable it would be to have 3-4-5 guys that walk a ton but have a low average. They would look nice on paper, but I have a feeling there would be a lot of guys left on base.

That said, I think OPS is probably one of the better stats we have in baseball, but I would prefer one that put more of an onus on extra base hits than one that focused on walks if someone were to change it.

[clintl]

Walks are good, but they're often not as good as a single. On a single, you have the opportunity to move players up more than one base.

My main problem with OPS, though, is that it double-counts hits. I think that's where the notion that SLG is less important than OBP comes from. A better measure might be something like (TB + BB + HBP)/PA. Maybe call it something like SLG+.

[SackAttack]

Walks are good, but they're often not as good as a single. On a single, you have the opportunity to move players up more than one base.

My main problem with OPS, though, is that it double-counts hits. I think that's where the notion that SLG is less important than OBP comes from. A better measure might be something like (TB + BB + HBP)/PA. Maybe call it something like SLG+.

I gave some thought to that notion the other night, but it was late, and I didn't get very far with it. I was thinking something along the lines of ISO (SLG - BA) + OBP, but it occurs to me that even there, extra-base hits get double-counted.

The solution might be a continuing to use SLG in conjunction with a sort of isolated OBP - (BB + HBP)/(PA - H) - but I'm not sure how effective that would be.

[SackAttack]

Dola,

In looking that over again, that looks like TB + BB+ HBP/PA, only phrased more stupidly.

[QuikSand]

The most interesting thing to me is that this guy:

(a) holds such an opinion (which isn't all that unusual, really);
(b) feels compelled to write and defend this opinion on a baseball discussion site; and
(c) despite these grave shortcomings, still writes in complete and fairly intelligible sentences.

[QuikSand]

Walks, while nice, do not score runs (unless the bases are loaded)

No, but they do have a very meaningful and positive correaltion with the team eventually scoring runs... which is really what a good metric here ought to help us represent.

Honestly... a guy who hits a lead-off triple doesn't score a run on that play, either. But any fool (I think) could agree that this situation has a strong relationship with his team eventually scoring one or more runs in that inning. So, it would make sense to give credit to the guy who hit the triple, and not just the galoot who knocks the sac fly.

[dawgfan]

I agree with Josh here. OBP seriously overvalues walks, IMO. I was having this discussion in OOTP regarding Jose Cruz Jr. Over the past month, he had an OBP of .400 - which made most at OOTP say he was doing very well. Yet, he had an average of .180 and a SLG of .330 - but walked a ton. Walks, while nice, do not score runs (unless the bases are loaded)- they simply give you a "mulligan" of sorts and put the onus on the next player to get that run in. I'm not sure how desireable it would be to have 3-4-5 guys that walk a ton but have a low average. They would look nice on paper, but I have a feeling there would be a lot of guys left on base.

That said, I think OPS is probably one of the better stats we have in baseball, but I would prefer one that put more of an onus on extra base hits than one that focused on walks if someone were to change it.

I dunno - OBP is basically how often does a batter not make an out in a plate appearence (with the exception of not counting times reached base via error). Regardless of whether you reach base via walk, a hit-by-pitch or a base hit, the most important thing is you're not making an out. This is why of the traditional baseball stats, OBP carries the highest correlation with runs scored, which is the whole point.

You can make arguments as to the relative merits of walks vs. singles, and I would agree that singles probably carry a little more value than walks, it's not as cut and dried as clintl would portray. I think the advantage singles carry in how far they advance a baserunner on average vs. a walk (<1 vs. 1) is tempered by the greater fatigue applied to the pitcher in drawing a walk (more pitches on average) as well as arguably a greater psychological blow in surrendering a walk vs. a hit.

[dixieflatline]

I am familiar with OPS and personally think it's a bunch of sabermetrics BS. To me, there's no better gauge than BA/HR/RBI. You can cut up stats, adjust for ballparks (ridiculous idea), or do whatever, but if you look at the history of baseball, look at the greatest hitters throughout the different eras, you'll find that BA/HR/RBI is as good a gauge as anything. Remember, Bill James invents statistics to sell books... People have been following and rating ballplayers based on BA/HR/RBI as far back as you can go, and not because we're any more mathematically skilled at assessing player value and can come up with "more accurate" statistics. It's because Bill James wants to sell books. Which is not to dismiss it as meaningless, but is to reject the idea that BA/HR/RBI is in somewhat insufficient in judging value.

As others have stated this is just crazy. The two things you want in a stat is something that does a good job of correlating with runs scored and something that correlates well with next year(projectability). OPS is far from perfect but BA/HR/RBI does a horrible job in both of these cases. In fact there are many stats now that are much better than OPS mostly because the advancement in computing. If I have a new stat I want to test all I have to do is download the entire history of baseball stats and check it myself. This was completely impossible just ten years ago.

I agree with Josh here. OBP seriously overvalues walks, IMO. I was having this discussion in OOTP regarding Jose Cruz Jr. Over the past month, he had an OBP of .400 - which made most at OOTP say he was doing very well. Yet, he had an average of .180 and a SLG of .330 - but walked a ton. Walks, while nice, do not score runs (unless the bases are loaded)- they simply give you a "mulligan" of sorts and put the onus on the next player to get that run in. I'm not sure how desireable it would be to have 3-4-5 guys that walk a ton but have a low average. They would look nice on paper, but I have a feeling there would be a lot of guys left on base.

I also think this is just crazy. If a player's first plate apperance of the year is a walk then his OPS is 1000. If his first PA is a single his OPS is 2000. How much do you want a walk to count? Sure a walk only scores a run if the bases are loaded but getting men on base is the first step to scoring runs. Another "hidden" value of a walk is the wear on the opposing pitcher. The pitcher had to throw at least 4 pitches and he didn't get anyone out. I would love it to have a 3-4-5 combo that walk a lot even if they have a low average(a high slugging would be a must for this player though). And because OPS correlates pretty well to runs scored I am confident that my team would score a lot of runs on the field not only on paper.

Go ask the Reds how having Adam Dunn and his .258/.391/.605(AVE/OBP/SLG) in the middle of their order is working for them.

[Arles]

I am not saying walks are not valuable, I am simply stating I think their impact is overvalued by many. A guy that walks all the time but does not get extra base hits has minimal value UNLESS he has a bunch of guys that can knock him in.


Honestly... a guy who hits a lead-off triple doesn't score a run on that play, either. But any fool (I think) could agree that this situation has a strong relationship with his team eventually scoring one or more runs in that inning. So, it would make sense to give credit to the guy who hit the triple, and not just the galoot who knocks the sac fly.
Yes, but it's much harder to score after a walk than it is after a triple. At somepoint, you are going to need a hit (or error) to score someone who walked with all else being equal. Equating walks and extra base hits in OBP makes that stat less valuable to me. OPS is better, but it double counts singles (which are only slightly better than walks).


I also think this is just crazy. If a player's first plate apperance of the year is a walk then his OPS is 1000. If his first PA is a single his OPS is 2000. How much do you want a walk to count? Sure a walk only scores a run if the bases are loaded but getting men on base is the first step to scoring runs.
As I said earlier, I don't mind OPS, I just think OBP is a little overvalued. The fact that someone could walk 30 times (with no hits) in 100 ABs and be "equal" to someone who hits 30 HRs (with no walks or other hits) seems a little silly to me.


Another "hidden" value of a walk is the wear on the opposing pitcher. The pitcher had to throw at least 4 pitches and he didn't get anyone out.
Again, walks have value, I just don't think OBP is a great indicator of value all by itself.

I would love it to have a 3-4-5 combo that walk a lot even if they have a low average(a high slugging would be a must for this player though).
I think you are making my point here. It's why I think slugging is a much better determination of a player's value than OBP or Avg. BTW, I like Avg much worse than OBP.

Go ask the Reds how having Adam Dunn and his .258/.391/.605(AVE/OBP/SLG) in the middle of their order is working for them.
But Dunn is the perfect example for someone I would value very highly as his is one of the league leaders in SLG. Another example is Manny Ramirez. By OPS, Manny (.961) isn't that much better than Nick Johnson (.946). Yet, Manny is slugging .589 and Johnson is slugging .512. IMO, Manny is a much more valuable hitter, yet the impact of OBP on OPS minimizes that.

[dawgfan]

I think you are making my point here. It's why I think slugging is a much better determination of a player's value than OBP or Avg.

Except that it's not - OBP has a higher correlation with runs scored than SLG (or BA). Part of the problem comparing the two (OBP & SLG) is that they overlap by both including singles. A better comparison is OBP vs. ISO (isolated power) since there's no overlap.

[SackAttack]

Except that it's not - OBP has a higher correlation with runs scored than SLG (or BA). Part of the problem comparing the two (OBP & SLG) is that they overlap by both including singles. A better comparison is OBP vs. ISO (isolated power) since there's no overlap.

I dunno. Maybe I'm looking at it all wrong, but let's take some extreme examples here.

A team full of players who never get hits, but walk 40% of the time will be gaining an average of .4 bases per trip to the plate. In order to gain the four bases necessary for a run, you need 10 walks.

A guy who bats .250, with no walks and a .500 SLG, he's getting two bases every four at-bats. On average, he's going to gain those same four bases in eight at-bats, rather than the ten necessary to do it with walks.

OBP has a higher correlation, but you have to remember that in the real world, part of that OBP is going to be the player's SLG.

By comparing ISO with OBP, there's still overlap - the difference is that the overlap is in the extra-base hits, rather than in the singles. OBP is counting the average of bases gained per plate appearance, while ISO is counting bases gained per plate appearance with the exception of singles, walks, and HBP.

It just seems to me like the better approach is going to be SLG + Isolated OBP, rather than OBP + SLG or OBP + Isolated Power.

[Arles]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs)

Here, extra base hits count the most, then singles and finally walks. But, singles take the biggest hit when compared to OPS.

Using this formula, the top 10 hitters are:
1. Derrek Lee
2. Adam Dunn
3. Albert Pujols
4. Manny Ramirez
5. Travis Hafner
6. Miguel Cabrera
7. Morgan Ensberg
8. Alex Rodriguez
9. Jason Giambi
10. Andruw Jones

Basically, the puts more emphasis on guys who get more extra base hits, a decent number of walks and fewer singles. Guys hurt in this rating are Brian Roberts, Nick Johnson and Moises Alou. Whereas guys like Dunn, Manny, Jim Edmonds and Andrew Jones get a boost. It's by no means perfect, but gets a little closer to what I think of from a value standpoint.

[dawgfan]

I dunno. Maybe I'm looking at it all wrong, but let's take some extreme examples here.

A team full of players who never get hits, but walk 40% of the time will be gaining an average of .4 bases per trip to the plate. In order to gain the four bases necessary for a run, you need 10 walks.

A guy who bats .250, with no walks and a .500 SLG, he's getting two bases every four at-bats. On average, he's going to gain those same four bases in eight at-bats, rather than the ten necessary to do it with walks.

This is because in case #1, the ISO + OBP = .400, while in case #2 the ISO + OBP = .500; in other words, not a fair comparison.

OBP has a higher correlation, but you have to remember that in the real world, part of that OBP is going to be the player's SLG.

True, as I pointed out above.

By comparing ISO with OBP, there's still overlap - the difference is that the overlap is in the extra-base hits, rather than in the singles. OBP is counting the average of bases gained per plate appearance, while ISO is counting bases gained per plate appearance with the exception of singles, walks, and HBP.

Not true. Isolated power is SLG - BA, i.e. the extra base value above singles a hitter generates. ISO removes the value of singles from the equation. Which is perfect because all hits are valued the same as a single in OBP; thus, OBP and ISO carries no overlap.

It just seems to me like the better approach is going to be SLG + Isolated OBP, rather than OBP + SLG or OBP + Isolated Power.

As shown above, SLG + isolated OBP will yield the same result as OBP + ISO but is easier to calculate (since ISO is a relatively common stat to find).

[dawgfan]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs)

Here, extra base hits count the most, then singles and finally walks. But, singles take the biggest hit when compared to OPS.

You might think this is better, but it's an opinion not supported by facts. Simply put, OBP shows the highest correlation to runs scored of any of the traditional stats, moreso than SLG and moreso than BA.

[Arles]

You might think this is better, but it's an opinion not supported by facts.
What metric are you using to determine how "good" it is?

Simply put, OBP shows the highest correlation to runs scored of any of the traditional stats, moreso than SLG and moreso than BA.
Of course it does. You can only score once per PA and OBP values all "on base" actions equally. So, I would expect a better correlation between OBP and runs scored by that player than SLG or AVG and runs scored - both of which do not account for bases gained by walks (and are therefore not accounting for a decent number of runs scored).

My point is simply that judging the numbers of runs someone scores is not enough to show value to me. I am more interested in the value they provide as a hitter. And, to this point, I haven't been sold on OBP showing that value in a vaccum.

[korme]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs)

Here, extra base hits count the most, then singles and finally walks. But, singles take the biggest hit when compared to OPS.

Using this formula, the top 10 hitters are:
1. Derrek Lee
2. Adam Dunn
3. Albert Pujols
4. Manny Ramirez
5. Travis Hafner
6. Miguel Cabrera
7. Morgan Ensberg
8. Alex Rodriguez
9. Jason Giambi
10. Andruw Jones

Basically, the puts more emphasis on guys who get more extra base hits, a decent number of walks and fewer singles. Guys hurt in this rating are Brian Roberts, Nick Johnson and Moises Alou. Whereas guys like Dunn, Manny, Jim Edmonds and Andrew Jones get a boost. It's by no means perfect, but gets a little closer to what I think of from a value standpoint.
Given this formula, if you lead the league in homeruns you are the best player in the league apparently. Adam Dunn isn't close to being the second most valuable position player in the league.

[dawgfan]

Of course it does. You can only score once per PA and OBP values all "on base" actions equally. So, I would expect a better correlation between OBP and runs scored by that player than SLG or AVG and runs scored - both of which do not account for bases gained by walks (and are therefore not accounting for a decent number of runs scored).

My point is simply that judging the numbers of runs someone scores is not enough to show value to me. I am more interested in the value they provide as a hitter. And, to this point, I haven't been sold on OBP showing that value in a vaccum.

When I say runs scored, I'm talking about per team - team OBP correlates better with team runs scored than team SLG.

If you have an ESPN Insider account, you can dig around Rob Neyer's archives to get more info on this claim. You might be able to find info for free on the web that also shows these sabermetric studies.

[Arles]

When I say runs scored, I'm talking about per team - team OBP correlates better with team runs scored than team SLG.
Again, I do not doubt that given OBP takes into account players that reach via walk. And, for an entire team over the course of a season, those walks will end up amounting to a large number of runs which would not be registered by SLG. I am simply trying to determine a single player's "value". Given that goal, OBP (no extra base hits), SLG (no walks) and OPS (premium on singles) leave me a little lacking.

Given this formula, if you lead the league in homeruns you are the best player in the league apparently.
No, this formula puts on value on the % of ABs that end up with extra base hits, while also rewarding players for taking walks. Andruw Jones is leading the league in HRs and he shows up 10th on the list. Jim Edmonds, on the other hand, does well because of his high # of walks (61) and % of extra base hits per AB (14.6%). But, by HR totals, he is not even in the top 25.

Adam Dunn isn't close to being the second most valuable position player in the league.
First of all, this does not account for defense - so keep that in mind. Next, if you go by OPS (as many do), he ranks #5 in MLB. The reason he does slightly better on my formula is that he has a high % of extra base hits per official AB (16.3%). Others like Pujols (13.8%) and A-Rod (11.9%) are much lower.

It's simply a different way to look at a hitter. What I was trying to do (and think it does) is put an onus on getting extra base hits when a player does not walk. All the while, still giving minor rewards to players that do get walks and singles. To me, this rewards extra base hits better than OBP, accounts for walks (unlike SLG) and doesn't overvalue singles like OPS.

[Vince]

Brilliant discussion so far guys -- keep it up. Hopefully tomorrow when my brain's not whacked from working a short-handed shift, I'll be able to add to it.

I like where Arlie is going a lot, and I think the idea has a lot of merit...with some tweaking and a bit of testing, I think he might have a nice stat there.

[QuikSand]

I dunno - OBP is basically how often does a batter not make an out in a plate appearence (with the exception of not counting times reached base via error). Regardless of whether you reach base via walk, a hit-by-pitch or a base hit, the most important thing is you're not making an out. This is why of the traditional baseball stats, OBP carries the highest correlation with runs scored, which is the whole point.

If you want to really be part of this debate -- read and understand this paragraph. It is the single most important part of understanding how this all-in-one metric concept has to work.

Well said.

[cuervo72]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs)

Doesn't that really overstate extrabase hits by counting them in the SLG and discounting singles? Looks like a stat that would trump up the value of players like oh, say this type of guy:

http://reference.thefobl.com/fbr/players/s/stowerdarrel754.html

[oykib]

I dunno - OBP is basically how often does a batter not make an out in a plate appearence (with the exception of not counting times reached base via error). Regardless of whether you reach base via walk, a hit-by-pitch or a base hit, the most important thing is you're not making an out. This is why of the traditional baseball stats, OBP carries the highest correlation with runs scored, which is the whole point.


If you want to really be part of this debate -- read and understand this paragraph. It is the single most important part of understanding how this all-in-one metric concept has to work.

Well said.

I'm still shocked that this isn't a better understood concept.

Earl Weaver was talking about the value of your 27 outs thirty years ago.

[Huckleberry]

Absolutely agree. Weighting SLG more important than OBP is incorrect in determining value.

I personally use OBP*1.4 + SLG in determining player value (for Diamond Mind leagues, for example). I call it MOPS in my spreadsheets for modified OPS.

[Klinglerware]

You might think this is better, but it's an opinion not supported by facts. Simply put, OBP shows the highest correlation to runs scored of any of the traditional stats, moreso than SLG and moreso than BA.

Let's remember that inferential statistics are not facts, they are figures derived from tools of analysis. These tools of analysis are only as good as the analyst--it is up to the analyst to derive meaning from the numbers he or she is seeing. For example, drowning deaths are correlated with ice cream sales--that doesn't mean that ice cream causes drowning or vice versa. What you need to do is to devise a credible causal mechanism to explain the statistical effects you are seeing. The statistics can guide you, but they cannot build theory on their own.

As for the OBP correlation--as you state, there is a strong correlation between team OBP and team runs scored. That in itself is pretty uncontroversial to me. I would be interested in seeing if anybody has run a correlation study on individual OBP to team runs scored. I would suspect that the correlation is much lower (plus I would like to see the theory on exactly why a player with higher OBPs would be more valuable than players with lower OBPs). Since we are looking at the value of individual players here, analyzing that statistic at the team level may not be appropriate to assessing value at the individual level.

I suspect that there really is no magic bullet statistic that would completely assess all facets of a player's value. Defense and base-running should also be considered (though I believe that SABR-metricians have done studies suggesting that these facets may be overvalued, nevertheless they probably do have some explanatory power), as well as games played (holding other factors constant, a player who plays 162 games is more valuable than a player who only manages to play 100)--none of these facets are captured by hitting statistics. Also, some may argue that return-on-investment calculations might be something to consider in an analysis. The bottom line is that multiple metrics probably should be used in determining a player's value.

That being said, I am quite sympathetic to the SABR-metric views on baseball, and many of the people running these studies are quite bright. My point is that statistics shouldn't be overvalued either--they are tools, not facts...

[Crapshoot]

I'm still shocked that this isn't a better understood concept.

Earl Weaver was talking about the value of your 27 outs thirty years ago.

yup. That's why sacrifice bunts, except in late, one run games, are a god-awful idea (pitcher not withstanding). To echo Quicksand and Dawg, this is the fundemental concept that people keep ignoring.

[Crapshoot]

Let's remember that inferential statistics are not facts, they are figures derived from tools of analysis. These tools of analysis are only as good as the analyst--it is up to the analyst to derive meaning from the numbers he or she is seeing. For example, drowning deaths are correlated with ice cream sales--that doesn't mean that ice cream causes drowning or vice versa. What you need to do is to devise a credible causal mechanism to explain the statistical effects you are seeing. The statistics can guide you, but they cannot build theory on their own.

As for the OBP correlation--as you state, there is a strong correlation between team OBP and team runs scored. That in itself is pretty uncontroversial to me. I would be interested in seeing if anybody has run a correlation study on individual OBP to team runs scored. I would suspect that the correlation is much lower (plus I would like to see the theory on exactly why a player with higher OBPs would be more valuable than players with lower OBPs). Since we are looking at the value of individual players here, analyzing that statistic at the team level may not be appropriate to assessing value at the individual level.

I suspect that there really is no magic bullet statistic that would completely assess all facets of a player's value. Defense and base-running should also be considered (though I believe that SABR-metricians have done studies suggesting that these facets may be overvalued, nevertheless they probably do have some explanatory power), as well as games played (holding other factors constant, a player who plays 162 games is more valuable than a player who only manages to play 100)--none of these facets are captured by hitting statistics. Also, some may argue that return-on-investment calculations might be something to consider in an analysis. The bottom line is that multiple metrics probably should be used in determining a player's value.

That being said, I am quite sympathetic to the SABR-metric views on baseball, and many of the people running these studies are quite bright. My point is that statistics shouldn't be overvalued either--they are tools, not facts...


Actually, statistics like Win Shares or WARP1 (or 3) attempt to calculate just this. More so, defense is highly valued by sabrematicians - especially if you understand the implication of DIPS. While every columnist has been preening about the A's low OBP, they've built a team around a damn good defense, because they recognize the value in that. Defensive statistics like UZR (propietary now sadly- but available from 2000-2003 - Cards fans will be glad to know the brain behind it works for their team) attempt to convert this into runs.

[Buccaneer]

I'm still shocked that this isn't a better understood concept.

Earl Weaver was talking about the value of your 27 outs thirty years ago.
And even going back to well into the 19th century (Pre-Chadwick days), the concept of "making your run" was the key. That was just an extreme version of not making an out. Back then, the most important thing to do was to get on base and to get your butt scored (with the latter being more important) - and not to make an out in any way or form.

[Klinglerware]

Actually, statistics like Win Shares or WARP1 (or 3) attempt to calculate just this. More so, defense is highly valued by sabrematicians - especially if you understand the implication of DIPS. While every columnist has been preening about the A's low OBP, they've built a team around a damn good defense, because they recognize the value in that. Defensive statistics like UZR (propietary now sadly- but available from 2000-2003 - Cards fans will be glad to know the brain behind it works for their team) attempt to convert this into runs.

Okay, fair point on defense. I'm not up to speed on the new research on this, I just remembered Bill James' early work when he thought that the defensive contribution of a player was vastly overstated...

[Huckleberry]

The MOPS rankings are just a hair different than the OPS rankings.

RO RM PLAYER OPS MOPS
1 1 Derrek Lee 1.128 1.303
2 2 Albert Pujols 1.052 1.224
7 3 Jason Giambi 0.991 1.168
3 4 Miguel Cabrera 1.006 1.168
4 5 Alex Rodríguez 0.997 1.163
6 6 Travis Hafner 0.995 1.160
5 7 Adam Dunn 0.996 1.152
8 8 Morgan Ensberg 0.972 1.126
15 9 Nick Johnson 0.946 1.120
10 10 David Ortíz 0.958 1.114
9 11 Manny Ramírez 0.963 1.113

[Crapshoot]

Okay, fair point on defense. I'm not up to speed on the new research on this, I just remembered Bill James' early work when he thought that the defensive contribution of a player was vastly overstated...

No, he probably was right - but that's a far cry from "defense " is irrelvant. Rather, I think sabremetrics tells you facts like Jeter is a god awful defender (last year, he was about average- naturally, he won the gold glove). James was pointing out that Doug Mientweitcz (sp ?) could be Ozzie Smith in the field, and it still doesn't justify him being out there at 1b.

[dixieflatline]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs)

Here, extra base hits count the most, then singles and finally walks. But, singles take the biggest hit when compared to OPS.


This stat is dumb. With this stat a walk counts as 1000. A single counts as 2000(the same as OPS so far). Doubles count as 5000. Triples count as 7000. Home runs count as 9000. I certainly hope even you don't believe that a double is "worth" 5 times more than a walk. Your belief that SLG should count more than OBP in OPS is exactly opposite the best multiplier that correlates with runs(around 1.4*OBP + SLG). If you are interested in building a baseball team arround a stat that doesn't correlate well with runs be my guest. There is nothing that I, or anyone, could say to change you mind on this, or any other topic it seems.

[Subby]

Just playing around, a formula that may be better is:

SLG*2 + OBP - (singles/PAs) Hmmm...you aren't german by any chance are you?

[Arles]

Doesn't that really overstate extrabase hits by counting them in the SLG and discounting singles? Looks like a stat that would trump up the value of players like oh, say this type of guy:

http://reference.thefobl.com/fbr/players/s/stowerdarrel754.html
Not really. Remember, singles are counted fully (from a total base standpoint) three times before being subtracted less than one time. Walks are also counted once in OBP and given another slight bump in the (singles/PA) factor in that if you have more walks, you will lose less here. This is by no means a perfect formula, it just heads towards the path I would like to see for a player's individual value as a hitter.


This stat is dumb. With this stat a walk counts as 1000. A single counts as 2000(the same as OPS so far). Doubles count as 5000. Triples count as 7000. Home runs count as 9000. I certainly hope even you don't believe that a double is "worth" 5 times more than a walk.
Two things here. First, I think we have to look at "chance to score" given the result. Ie, if someone gets a walk - what are the chances they score from 1B. Then, what are the chances they score from 2nd, and chances they score from third. Of course, for a home run, the chance is 100%. so, having a homerun count 9X more than a walk may be a little better if the chance someone scores from 1B is 20% compared to 100% of the Home run.

Second, I'm trying to value players from a hitting standpoint. I can go up and take a few walks by not swinging. Most players can hit a single from time to time. But, fewer players can consistently hit extra base hits. Therefore, there should be premium put on those that have the ability to do so. Now, batting eye is important and it is a factor here, but if you are going to score runs - the quickest way (and most effective) to get them is extra base hits. So, when valueing players, I am looking for a way that will put a premium on guys that when they decide to swing, they get more extra base hits - while still taking into account walks and singles as they do matter.

Your belief that SLG should count more than OBP in OPS is exactly opposite the best multiplier that correlates with runs(around 1.4*OBP + SLG).
Again, this is correlating to runs. The causality is still a little more of a grey area for individual players. If someone gets on base 8 times in 12 ABs with 6 singles and 2 walks, they are in position to score more runs than someone who gets a single, a walk and three home runs. But, does that mean the first person is more valuable? I'm not sure I buy that logic. OBP's claim to fame is its correlation to team runs. I don't know that you can carry that over for the value of individual players. I could very well be wrong, but I am not convinced solely by the studies I've seen attaching OBP to team runs.

If you are interested in building a baseball team arround a stat that doesn't correlate well with runs be my guest. There is nothing that I, or anyone, could say to change you mind on this, or any other topic it seems.
It seems to me that you are doing exactly what the Bill James crowd used to despise 10 years ago - holding on to something that you hold true but has not been definitively shown to be true (ie, OBP being the best evaluator of a hitter's value). And, consequently, are simply unwilling to engage in debate over the value of the concept when its challenged because of different ideas. OBP may very well be the best indicator of an individual player's value. If after this discussion and feedback from the numerous strong minds in this forum that becomes the consensus (and there is relevant data to back it), I have no problem changing my forumula and hyphothesis accordingly. However, you seem intent on shouting down anyone that doesn't treat OBP in the same reverence as you do - which is often very limiting when engaging in these discussions.

[Huckleberry]

I agree that the tone of some of the replies to your post is regrettable. But I have a question for you:

What is the purpose of ranking the value of players if not for the theoretical compilation of a team, roster, or lineup? Clearly if you compiled the best team you could using your formula, then the team would be weighted more heavily toward slugging and less toward OBP. Then the team would score fewer runs than a team compiled through the use of 1.4*OBP + SLG.

That is why it correlates to individual player value.

[Arles]

Maybe the best way to look at these are in examples.

Case 1, you have a player that gets 20 BBs, 15 1B, 5 2B, 0 3B and 5 HRs in 100 ABs (120 PAs).

His AVG is .250, OBP is .375, OPS is .750, "MOPS" is 0.900 and my formula is 1.29. AVG is a woeful evaluator here, while the rest seem OK at first glance.

Case 2, you have a player that gets 15 BBs, 8 1B, 10 2B, 1 3B and 11 HRs in 100 ABs (120 PAs).

His AVG is .300, OBP is .391, OPS is 1.04, "MOPS" is 1.200 and my formula is 1.63.

IMO, the second guy is a much more valuable hitter than the first. Yet, OBP shows only a benefit of 4%. MOPS is better (25%) and OPS is slightly better than that (~28%). But I like that my forumula shows an increase of around 39%. If you have two guys that both get on base 45 times in 120 PAs, but one has 22 extra base hits while the other has 10, it would seem reasonable to me to hold the former guy in much higher esteem than the latter.

[Arles]

I agree that the tone of some of the replies to your post is regrettable. But I have a question for you:

What is the purpose of ranking the value of players if not for the theoretical compilation of a team, roster, or lineup?
Great question and one I should have answered. My goal is to decide an individual player's value as a hitter. Think of it like I had $20 million to lay on one hitter and wanted to know on who my money would be best spent (without factoring in defense or position).

Clearly if you compiled the best team you could using your formula, then the team would be weighted more heavily toward slugging and less toward OBP. Then the team would score fewer runs than a team compiled through the use of 1.4*OBP + SLG.
That may be - as evaluating an entire team is not the goal I am after. Most of the valueing of baseball stats seem to be based on looking at large groups (ie teams or even leagues) and looking for correlations. What I am trying to do is find a number one could use to simply evaluation the hitting ability of one player. Now, to score runs for a team, you need numerous roles filled from leadoff guys to middle of the lineup hitter and lower lineup players. For that reason, using one formula (like mine) to determine the overall ability of an entire lineup is probably not wise. I am simply looking at finding away to evaluate "bang for buck" if I want to bring in one "elite" hitter.

That is why it correlates to individual player value.
It probably does, but just because there is a relationship does not make it the best metric.

[korme]

Maybe the best way to look at these are in examples.

Case 1, you have a player that gets 20 BBs, 15 1B, 5 2B, 0 3B and 5 HRs in 100 ABs (120 PAs).

His AVG is .250, OBP is .375, OPS is .750, "MOPS" is 0.900 and my formula is 1.29. AVG is a woeful evaluator here, while the rest seem OK at first glance.

Case 2, you have a player that gets 15 BBs, 8 1B, 10 2B, 1 3B and 11 HRs in 100 ABs (120 PAs).

His AVG is .300, OBP is .391, OPS is 1.04, "MOPS" is 1.200 and my formula is 1.63.

IMO, the second guy is a much more valuable hitter than the first. Yet, OBP shows only a benefit of 4%. MOPS is better (25%) and OPS is slightly better than that (~28%). But I like that my forumula shows an increase of around 39%. If you have two guys that both get on base 45 times in 120 PAs, but one has 22 extra base hits while the other has 10, it would seem reasonable to me to hold the former guy in much higher esteem than the latter. No doubt that the second player is better. See, you never *just* compare OBP to each other without looking at any other stat. No one in their right mind will look at those two guys and think they are comparable because their OBP is the same. The one thing I will take from that is that the first guy has a phenomenal eye with an OBP about .125 higher than his batting average.

I just think it's odd that you think people might compare those two players as being equally good because their OBP is the same when the second guy's OPS is ridiculously better.

[dixieflatline]

Two things here. First, I think we have to look at "chance to score" given the result. Ie, if someone gets a walk - what are the chances they score from 1B. Then, what are the chances they score from 2nd, and chances they score from third. Of course, for a home run, the chance is 100%. so, having a homerun count 9X more than a walk may be a little better if the chance someone scores from 1B is 20% compared to 100% of the Home run.

Let's look at the expected run matrix indeed. From www.baseballprospectus.com for 2004:


OUTS
RUNNERS 0 1 2
--- 0.5379 0.2866 0.1135
1-- 0.9259 0.5496 0.2460
-2- 1.1596 0.7104 0.3359
12- 1.4669 0.9577 0.4605
--3 1.4535 0.9722 0.3623
1-3 1.8540 1.2236 0.5219
-23 2.1343 1.4717 0.6179
123 2.2548 1.5946 0.8082


No outs nobody on and I take a walk. I've upped my teams expected runs by .388. If I double instead my team's expected runs go up by .6217. You can plug in your favorite situation and do the math. If you then weigh each situation by the likelihood it will occur you can then find just how much a double is "worth" compared to a walk. It clearly will be more than twice as good but it won't be nearly five times as good. I can tell you though it will be much, much, less than you think. In fact there are stats out there that do exactly this. And, by definition, they correlate very well to runs scored. Doubles are great. I love doubles. I also like walks though.

Second, I'm trying to value players from a hitting standpoint. I can go up and take a few walks by not swinging. Most players can hit a single from time to time. But, fewer players can consistently hit extra base hits. Therefore, there should be premium put on those that have the ability to do so.

When you boil everything down this is where we disagree. You like players that have an ability that few other players have(consistently hit extra base hits). It doesn't matter how much this actually helps the team score runs more than the unsexy walk or single. It just matters because few players can do this so it should be "worth" a lot. To me I could care less what my players are doing as long as the team is scoring runs. If that means a lot of HBP that's great. If that means lot's of doubles that's great too. When the season is over and I go to evalute my players the ONLY thing I am concerned with is how much did this player help us score runs. I therefor am going to use a stat that correlates to the team scoring runs. This is NOT OPS by the way because it isn't good enough.

[cuervo72]

Not really. Remember, singles are counted fully (from a total base standpoint) three times before being subtracted less than one time. Walks are also counted once in OBP and given another slight bump in the (singles/PA) factor in that if you have more walks, you will lose less here. This is by no means a perfect formula, it just heads towards the path I would like to see for a player's individual value as a hitter.

So you'd like a full team of Darrell Stowers'? Look at his 2014 campaign... .204 avg, .273 OBP, but he had 42 HR. His formula would be:

.450 * 2 + .273 - 58/672

I have that coming out to 1.087. So for a player with a .723 OPS, he'd be just a little less valuable than Manny Ramirez?

Ok, gotcha.

[SackAttack]

Let's look at the expected run matrix indeed. From www.baseballprospectus.com for 2004:


OUTS
RUNNERS 0 1 2
--- 0.5379 0.2866 0.1135
1-- 0.9259 0.5496 0.2460
-2- 1.1596 0.7104 0.3359
12- 1.4669 0.9577 0.4605
--3 1.4535 0.9722 0.3623
1-3 1.8540 1.2236 0.5219
-23 2.1343 1.4717 0.6179
123 2.2548 1.5946 0.8082


Looking at that matrix, though, you start to get a sense of stolen bases being overrated. A player who steals second after reaching first only increases his team's expected runs by around 0.23, but if he gets caught stealing, he reduces his team's expected runs by 0.64ish.

I guess it comes back down to the idea of if you play for one run, that's what you're going to get.

[QuikSand]

Second, I'm trying to value players from a hitting standpoint. I can go up and take a few walks by not swinging. Most players can hit a single from time to time. But, fewer players can consistently hit extra base hits. Therefore, there should be premium put on those that have the ability to do so. Now, batting eye is important and it is a factor here, but if you are going to score runs - the quickest way (and most effective) to get them is extra base hits. So, when valueing players, I am looking for a way that will put a premium on guys that when they decide to swing, they get more extra base hits - while still taking into account walks and singles as they do matter.

I have read this a couple of times. *pause* Okay, now three times. I still don't see how this explanation is anything but reframing the original question/topic to better fit the answer/assertion you offered.


If we are having a discussion about ways to measure a player's effectiveness as a batter in baseball (and until that post, I thought we all were doing exactly that), then it stands to reason that we want to value the things that correlate to the team scoring runs. This really isn't all that hard to do -- and when you invest some time in it, you find that OPS does a very good job. It can be tweaked a bit to good effect (as some have offered, perhaps weighting OBP a little higher makes it a bit more accurate -- essentially the opposite of what you suggest) but on balance, it's a good indicator of a player's value to his team as a batter.



If you now want to start essentially a second debate about "it's hitting, not batting" or whatever else it is you think your "extra base hits are everything" concept supports -- that's okay, but understand that is all you're doing. It doesn't make your world full of Adam Dunns any more valuable to their actual baseball teams just because you're now succesfully correlating to something other than scoring runs, which is what really matters.

[dawgfan]

Again, I do not doubt that given OBP takes into account players that reach via walk. And, for an entire team over the course of a season, those walks will end up amounting to a large number of runs which would not be registered by SLG. I am simply trying to determine a single player's "value". Given that goal, OBP (no extra base hits), SLG (no walks) and OPS (premium on singles) leave me a little lacking.

A lot of time and effort has already been expended in the pursuit of quantifying an individual player's contributions towards a team scoring runs. Complex versions of Runs Created per 27 outs give us the best results in this quantification.

[cuervo72]

yy, RC/27!

[dixieflatline]

Looking at that matrix, though, you start to get a sense of stolen bases being overrated. A player who steals second after reaching first only increases his team's expected runs by around 0.23, but if he gets caught stealing, he reduces his team's expected runs by 0.64ish.

I guess it comes back down to the idea of if you play for one run, that's what you're going to get.

Yes which is exactly why you had better steal bases at a good clip(at least 2 out of 3) or your just hurting your team. Certainly in a close game in the later innings getting a key run is important but early in the game your not helping you team much at all by stealing second.

[SackAttack]

A lot of time and effort has already been expended in the pursuit of quantifying an individual player's contributions towards a team scoring runs. Complex versions of Runs Created per 27 outs give us the best results in this quantification.

Are there any iterations of RC/27 that make allowances for park factors, though?

[dawgfan]

Let's remember that inferential statistics are not facts, they are figures derived from tools of analysis. These tools of analysis are only as good as the analyst--it is up to the analyst to derive meaning from the numbers he or she is seeing. For example, drowning deaths are correlated with ice cream sales--that doesn't mean that ice cream causes drowning or vice versa. What you need to do is to devise a credible causal mechanism to explain the statistical effects you are seeing. The statistics can guide you, but they cannot build theory on their own.

As for the OBP correlation--as you state, there is a strong correlation between team OBP and team runs scored. That in itself is pretty uncontroversial to me. I would be interested in seeing if anybody has run a correlation study on individual OBP to team runs scored. I would suspect that the correlation is much lower (plus I would like to see the theory on exactly why a player with higher OBPs would be more valuable than players with lower OBPs). Since we are looking at the value of individual players here, analyzing that statistic at the team level may not be appropriate to assessing value at the individual level.

True. However let's remember that baseball stats reflect a closed system - unlike most of the real world, we can quantify all offensive results in baseball. And by looking at things like the expected run matrix, you now have a better picture of the causation of why certain offensive events are worth a certain amount of value.

As to comparing individuals vs. the team, your logic of team OBP vs. individual OBP value differing doesn't make sense. Let's say you take 2 players - one rates higher in a measure like RC/27, the other rates higher in Arlie's formula. Clone those players so you have one team of player A (RC/27) and one team of player B (Arlie's formula) - the team full of player A's will score more runs.

Obviously there will be variety in reality in the offensive contributions of your lineup, but each player constitutes a vertical slice of the equation. What's good for the team is what's ideal for the individual.

I suspect that there really is no magic bullet statistic that would completely assess all facets of a player's value. Defense and base-running should also be considered (though I believe that SABR-metricians have done studies suggesting that these facets may be overvalued, nevertheless they probably do have some explanatory power), as well as games played (holding other factors constant, a player who plays 162 games is more valuable than a player who only manages to play 100)--none of these facets are captured by hitting statistics. Also, some may argue that return-on-investment calculations might be something to consider in an analysis. The bottom line is that multiple metrics probably should be used in determining a player's value.

Well, there have been attempts to collect all facets of the game and give a numerical value to players that assesses their contributions towards their team winning as pointed out above. The best known one currently is Win Shares devised recently by Bill James - it attempts to quantify all facets of a player's contributions, hitting and fielding and give them the appropriate weighting towards contributing to their team's success, as well as provide a balanced way of comparing pitchers with hitters. I haven't bought the book that details the formula, and since I'm not a math savant it would take me a while to digest it all if I did, so I can't give a personal opinion of the accuracy of Win Shares, but my impression is that it's viewed in a pretty favorable light in the sabermetric community.

If we're talking strictly offensive value, the best one seems to be the more complex versions of runs created per 27 outs - this carries an extremely high correlation with runs scored when you assess a team's stats.

That being said, I am quite sympathetic to the SABR-metric views on baseball, and many of the people running these studies are quite bright. My point is that statistics shouldn't be overvalued either--they are tools, not facts...

Very true, and just like any other science it is healthy to view things with skepticism and continue to test the theories with more studies to see if there isn't a better formula or theory. I think there is a fair amount of hubris that has grown around the sabermetric crowd that gets very stubborn in their conviction that they are right and that certain concepts are sacrosanct. This is a dangerous trap that the community needs to be aware of and combat - the temptation to think they've figured things out completely and that existing theories are infallible.

[dawgfan]

Are there any iterations of RC/27 that make allowances for park factors, though?

I'm pretty sure there are - I think what you'd do is evaluate how a park effects all facets of offensive output compared to the league average and then run an individual's output through that filter before calculating RC/27.

[SackAttack]

I'm pretty sure there are - I think what you'd do is evaluate how a park effects all facets of offensive output compared to the league average and then run an individual's output through that filter before calculating RC/27.

What I'm thinking is that the filter should probably only be applied to the player's home statistics, because he will play an inordinate number of games in his home park as compared to any other park, and thus, his RC/27 will be more greatly influenced by his home performance than by the fluctuations of road ballparks.

[dixieflatline]

What I'm thinking is that the filter should probably only be applied to the player's home statistics, because he will play an inordinate number of games in his home park as compared to any other park, and thus, his RC/27 will be more greatly influenced by his home performance than by the fluctuations of road ballparks.

Park effects have been a real issue since baseball went to interleague play. Correctly adjust data for park effects when teams aren't playing a balanced schedule is not very easy.

[dawgfan]

What I'm thinking is that the filter should probably only be applied to the player's home statistics, because he will play an inordinate number of games in his home park as compared to any other park, and thus, his RC/27 will be more greatly influenced by his home performance than by the fluctuations of road ballparks.

I think any sophisticated park effects correction will do this. You also have to make a small adjustment to the road stats though as well, since the road parks won't exactly equal the league average. As dixie points out, making corrections for park effects is rendered more difficult because of interleague play and unbalanced schedules. That said, any moderate effort in this area will go a long way towards correcting imbalances caused by the uniqueness of a player's home park.

[Klinglerware]

As to comparing individuals vs. the team, your logic of team OBP vs. individual OBP value differing doesn't make sense. Let's say you take 2 players - one rates higher in a measure like RC/27, the other rates higher in Arlie's formula. Clone those players so you have one team of player A (RC/27) and one team of player B (Arlie's formula) - the team full of player A's will score more runs.



My point of contention is that you are using a statistical finding from one level of analysis (team) and attempting to use it to make an inference at a different level (individual). The high correlation score between team OBP and team Runs Scored make sense--I don't disagree with that finding. There certainly could be a correlation between individual OBP and team Runs Scored, but I would rather see an actual correlation study run to find out for sure--I have trouble with concluding anything on the micro individual level, based on analyses done at the macro team level...

[Huckleberry]

Thanks for the 2004 matrix. Been a while since I looked at one.

Here are my quick observations:

Stealing 3rd with 1 out makes the most sense of any steal. To check myself I created a matrix of the breakeven steal percentage for each situation (runner in bold is stealing and someone check my work):

OUTS
RUNNERS 0 1 2
1-- 0.7323 0.7306 0.7324
-2- 0.7481 0.6951 0.9271
12- 0.7032 0.7280 0.8947
--3 0.9326 0.7320 0.3254
1-3 0.7588 0.7764 0.8446
1-3 0.9478 0.7499 0.4189
-23 0.9825 0.8263 0.4625
123 0.8595 0.7575 0.5534

[dawgfan]

There certainly could be a correlation between individual OBP and team Runs Scored, but I would rather see an actual correlation study run to find out for sure--I have trouble with concluding anything on the micro individual level, based on analyses done at the macro team level...

That's what the above scenario describes - using the runs created formula (which has an extremely high accuracy in projecting runs vs. actual runs) you can compare lineups of 2 different players and figure out which one would score more runs.

[dixieflatline]

Stealing 3rd with 1 out makes the most sense of any steal.

I agree with this completely. This is the best time to steal. It would be interesting to see a breakdown on when teams try to steal bases and how many of them are situations where it really isn't helping your team all that much.

I wonder how many MLB teams have done this study?

[Arles]

So you'd like a full team of Darrell Stowers'? Look at his 2014 campaign... .204 avg, .273 OBP, but he had 42 HR. His formula would be:

.450 * 2 + .273 - 58/672

I have that coming out to 1.087. So for a player with a .723 OPS, he'd be just a little less valuable than Manny Ramirez?

Ok, gotcha.
Actually, Manny Ramirez would grade a 1.44 in my formula with an OPS of .963. When using OPS, Manny is 24.9% better than Stowers. Using my formula, he is 24.1% better than Stowers. Not much of a difference there between ratings.

I was simply trying to find some way to measure a player's ability as a hitter outside of OBP and OPS. The formula I chose may need serious modification (and it probably does), but I just do not think OPS or OBP reward extra bases enough. Maybe I am valueing extra base hits too highly - it's all possible. I also was not trying to pull a fast one Quicksand by using the terms "hitter and batter" seperately, I was trying to re-iterate that I was only talking about a player's ability as a hitter (not taking defense into account).

At the macro level, I agree that you would be hardpressed to beat OBP when evaluating a team or group of players. I was just hoping that there would be a better way to evaluate the value of an individual hitter than OBP or OPS. It appears that many here do not feel that this the case, but I will still follow in hopes something emerges. But, in the end, maybe some simple combo of OBP and OPS is the best way for evaluating players. It leaves me somewhat unfulfilled, but I guess it will have to do.

[Arles]

As to comparing individuals vs. the team, your logic of team OBP vs. individual OBP value differing doesn't make sense. Let's say you take 2 players - one rates higher in a measure like RC/27, the other rates higher in Arlie's formula. Clone those players so you have one team of player A (RC/27) and one team of player B (Arlie's formula) - the team full of player A's will score more runs.
I certainly agree with this. But my point was what if you could NOT have a team full of player "A"s or player "B"s. Let's say you have 8 "average" major leaguers to go with this one player. Is it still a slam dunk to go with a high OBP or OPS guy over one with a high formula similar to what I posted?

Is it better to have a guy that does very well at reaching 1B when he has a group of players behind him not nearly as accomplished at reaching base? Or is it better to have a guy that is better at reaching 2nd, 3rd or scoring on his own (yet a little worse at reaching 1B) when he has a group of players behind him not nearly as good at reaching base? These are the questions I am not sure on.

But, I will certainly agree that if you could have 9 guys with a high OBP, you would maximize your chances for scoring runs. Still, the real world is often a far cry from that and teams will undoubtedly be playing average to below average guys at certain spots.

Obviously there will be variety in reality in the offensive contributions of your lineup, but each player constitutes a vertical slice of the equation. What's good for the team is what's ideal for the individual.
I'm just not sold on this. If you take it to the extreme and say you have 8 guys with an OBP of .300 - is it better to have the ninth guy have an OPB of .450 but a SLG of .390 or an OPB of .350 but a slugging of .490? Practically, it seems like the ability of the .490 SLG player to put himself on 2nd, 3rd or all the way home with no help from his teammates might prove more valuable than a guy that simply gets on base a little more often.

Well, there have been attempts to collect all facets of the game and give a numerical value to players that assesses their contributions towards their team winning as pointed out above. The best known one currently is Win Shares devised recently by Bill James - it attempts to quantify all facets of a player's contributions, hitting and fielding and give them the appropriate weighting towards contributing to their team's success, as well as provide a balanced way of comparing pitchers with hitters. I haven't bought the book that details the formula, and since I'm not a math savant it would take me a while to digest it all if I did, so I can't give a personal opinion of the accuracy of Win Shares, but my impression is that it's viewed in a pretty favorable light in the sabermetric community.
I have heard of this, but haven't really looked into it. I have to admit I do like the way it sounds though.

If we're talking strictly offensive value, the best one seems to be the more complex versions of runs created per 27 outs - this carries an extremely high correlation with runs scored when you assess a team's stats.
This is another stat I have heard of but am not all that familiar with. Is there a definition of it somewhere?

[Crapshoot]

I'm just not sold on this. If you take it to the extreme and say you have 8 guys with an OBP of .300 - is it better to have the ninth guy have an OPB of .450 but a SLG of .390 or an OPB of .350 but a slugging of .490? Practically, it seems like the ability of the .490 SLG player to put himself on 2nd, 3rd or all the way home with no help from his teammates might prove more valuable than a guy that simply gets on base a little more often.



And in this statement lies the major flaw Arlie - a 450/390/840 guy is more valuable, by far, than a 350/490/840 guy - because he makes less outs. A single is worth about 1.4 times a walk - basic OPS treats it as twice as much, and that needs to be corrected for - thus the (x*OBP) + (SLG) measure that's generally accepted - with x ranging from 1.4 to 1.8

[NYFAN]

I've read a little bit of this topic, and am only partially knowledgable of these stats (I know how some of them are computed... not all)... anyway, I'm a volleyball coach, and we have a very interesting stat, it's called hitter efficiency, TV Analysts love to compare it to BA, but it's really not... anyway it goes like this...

Kills - a hitter hits, and terminates a rally, thus a point for his team
Error - the opposite of a kill (gives a point to the other team)
Zero - The hitter neither scored, nor committed an error
Attempt - Anything that was attempted to cause a kill, and as a result falls in one of the above categories

(Kills-Errors) / Attempts = Hitter Efficiency

While it doesn't attempt to give credit to other facets of the game that may effect their HE (like quality of passing, setting etc...) it gives you a remarkably accurate picture of what the hitter has done, especially in comparison to teammates at the same position (some positions have certain advantages that allow them to have higher %, much like in basketball where big men are taking easier shots than perimeter players).

Anyway, my point is that it seems like devising a strong system for baseball could easily be derived from this system...

(Singles * 1.5 + Doubles * 2 + Triples * 3 + HRs * 4 + BB + HBP - K's) / Total Attempts

Now a few clarifications... I'm multiplying singles by 1.5 to give it a distinct advantage over a walk because a single theoretically moves a teammate between 1 and 2 bases over. The rest of the multipliers are just to give a bases figure, and subtracting K's because at this point in time a K not only doesn't advance a runner, but it doesn't even force the defense to make a play on the ball, meaning you can't even get the benefit of an error. By not subtracting hit balls that were defended, you are basically saying that a batter made contact with the ball, and the defense was able to stop it, thus it's a void play, or a "zero" in volleyball. Dividing it by Attempts gives you your average.

Any thoughts?

[dawgfan]

Here's the Wikipedia page on runs created for those interested: http://en.wikipedia.org/wiki/Runs_created

[Arles]

And in this statement lies the major flaw Arlie - a 450/390/840 guy is more valuable, by far, than a 350/490/840 guy - because he makes less outs. A single is worth about 1.4 times a walk - basic OPS treats it as twice as much, and that needs to be corrected for - thus the (x*OBP) + (SLG) measure that's generally accepted - with x ranging from 1.4 to 1.8
This certainly makes sense, but is Player A worth more simply because he makes fewer outs? Again, if Player A ends up stranded on 1B three straight times with Player B grounding out twice and hitting a HR, wouldn't player B's output be more valuable?

What you are saying is certainly true when viewing an entire time or large collections of players. But, given the team around him, I could see a player with a higher SLG being more valuable than one with a high OBP. Maybe I am making this too hard, but I just don't see how the trend that OBP shows team runs fairly well translates perfectly to the micro level of individual performance.

[dawgfan]

This certainly makes sense, but is Player A worth more simply because he makes fewer outs? Again, if Player A ends up stranded on 1B three straight times with Player B grounding out twice and hitting a HR, wouldn't player B's output be more valuable?

What you are saying is certainly true when viewing an entire time or large collections of players. But, given the team around him, I could see a player with a higher SLG being more valuable than one with a high OBP. Maybe I am making this too hard, but I just don't see how the trend that OBP shows team runs fairly well translates perfectly to the micro level of individual performance.

If you trust the runs created formula to accurately predict actual run output (and the more sophisticated versions get extremely close in this regard) you can experiment - take a lineup full of guys with a certain batting line and see what they produce as a collective in terms of runs. Then start unplugging guys and replacing them with players that have different batting lines. If you want to test your formula, take guys that get a set value with your formula, then start replacing them with guys with a similar overall value but higher slugging numbers and lower OBP numbers, then see what results you get.

[lynchjm24]

Nothing correlates better to runs scoring then

(OBP*1.8)*SLG

It's been shown many times over to be that equation or quite close to it.

[lynchjm24]

This is another stat I have heard of but am not all that familiar with. Is there a definition of it somewhere?

RC/27 is realtively simple. It's just a rate stat of Runs Created per 27 - the offensive equivilent of ERA.

If you create 85 runs and use 540 outs to do it:

85/(540/27)=4.25 runs per 27 outs.

[dawgfan]

RC/27 is realtively simple. It's just a rate stat of Runs Created per 27 - the offensive equivilent of ERA.

If you create 85 runs and use 540 outs to do it:

85/(540/27)=4.25 runs per 27 outs.

I'm assuming he was looking for the formula for runs created, which ranges from the simple version (which is very accurate by itself) to the more complex versions that take into account base stealing, grounding into double plays, etc (even more accurate).

[SackAttack]

This is another stat I have heard of but am not all that familiar with. Is there a definition of it somewhere?

Simple definition: OBP*TB, but as has been mentioned, there are far more complex iterations of the formula floating around out there as well.

[SackAttack]

Dola,

For runs created, that is; that was sort of in response to dawgfan's comment that Arles was looking for runs created.

[lynchjm24]

Very true, and just like any other science it is healthy to view things with skepticism and continue to test the theories with more studies to see if there isn't a better formula or theory. I think there is a fair amount of hubris that has grown around the sabermetric crowd that gets very stubborn in their conviction that they are right and that certain concepts are sacrosanct. This is a dangerous trap that the community needs to be aware of and combat - the temptation to think they've figured things out completely and that existing theories are infallible.


True to a degree. There are certain areas that are much more science and less gray.

I have no doubt that as far as quantifing the offensive contribution of a player sabermetrics does a tremendous job. The offensive metrics tell 99% of the story. If an offensive event has already happened, it can be quantified.

There are other areas where there has been great work, but it's still much grayer. For example, defense. I'm a huge stat guy, but I'm still going to weigh the observations of a player on defense very heavily when making decisions about them.

I think that sometimes the people who follow this closely blur the lines between these sorts of things. Just because runs created or VORP is a very solid measure doesn't mean that UZR is. I think that the very brightest though are smart enough to realize this and aren't as dogmatic as they were in the past.

[dawgfan]

True to a degree. There are certain areas that are much more science and less gray.

I have no doubt that as far as quantifing the offensive contribution of a player sabermetrics does a tremendous job. The offensive metrics tell 99% of the story. If an offensive event has already happened, it can be quantified.

There are other areas where there has been great work, but it's still much grayer. For example, defense. I'm a huge stat guy, but I'm still going to weigh the observations of a player on defense very heavily when making decisions about them.

I think that sometimes the people who follow this closely blur the lines between these sorts of things. Just because runs created or VORP is a very solid measure doesn't mean that UZR is. I think that the very brightest though are smart enough to realize this and aren't as dogmatic as they were in the past.

I agree with the thought that we're much farther along in quantifying hitting contributions in a meaningful way than we are for fielding. Areas I was thinking of though are more along the lines of thoughts on clutch hitting, the extent to which DIPS is true, whether platoon splits are essentially equal among all batters with differences accounted for as just random fluctuation, etc.

On clutch hitting, Bill James was one of (if not the) first people to attempt to prove through data whether such an ability actually existed, and his early attempts found no compelling evidence. Since that time, the concept of clutch hitting has been deemed by stat-heads as a mythical thing, a hoary old concept clung to by old-school baseball types but without merit. More studies have been done since that seem to come to the same conclusion, but pretty much every one I've seen has dealt with the same issues - how do you define clutch hitting (and is it the same thing for every hitter) and small sample sizes. When Bill James in recent years stated he was reconsidering his views on clutch hitting, there was a huge uproar.

On DIPS, when Voros McCracken first published his groundbreaking study on the events pitchers had a great degree of control over and those they didn't, it was a bombshell that quickly converted most of the stat-head community, such that the new gospel was that pitchers have no control over whether a ball put in play against them turns into a hit or an out. This despite a well-researched follow-up study by Tom Tippett showing that, while it's much less of a repeatable skill than strikeouts, some pitchers do have some control over hits allowed. This thought was echoed by Bill James and others (and McCracken himself backed off his initial conclusions somewhat), yet some in the stat-head community haven't gotten the message.

On platoon splits, there was a lot of excitement when James suggested that hitters' platoon splits were all essentially the same if you had enough sample size to look at, and that variations in splits were actually just random fluctuation. I believe he's since backed off from that assertion somewhat, with the current thinking that the concept is more true for left-handed hitters than right-handed hitters, but again this is a recent idea that was embraced by the community and some are having a hard time backing off that initial belief.

What concerns me is that a certain level of arrogance has crept into some members of the stat-head community in thinking that many concepts that have been stated are hard and fast rules, and are very resistant to any reconsiderations of those conclusions by newer studies. Like any science, the theories of sabermetrics should constantly be tested by new theories to see if the existing ones still hold up, or if the newer theories better describe reality. Science and sabermetrics should be as Carl Sagan described a 'Darwinian survival of the fittest theory'.

[cuervo72]

If you trust the runs created formula to accurately predict actual run output (and the more sophisticated versions get extremely close in this regard) you can experiment - take a lineup full of guys with a certain batting line and see what they produce as a collective in terms of runs.


Off on a little bit of a tangent here, going again from reall baseball to OOTP fake baseball....I actually wrote a program that would:

a) parse through past seasons of stats, and tallying them by stat and rating (so all doubles by guys with 7 ratings against RHP go here, all HR by guys with 4 ratings vs LHP go here, etc). Determine percentage of time per AB that said outcome occurs for that rating occurs (I do this from the Joe/BOSI exporter...accounts for league totals better I think).

b) use these composite components to build a profile for a given guy and his ratings sets (vs RHP and vs LHP)

c) for teams, parse the vLHP/vRHP lineups, adding the components together to find team numbers (vRHP numbers are weighted to account for more RHP)

d) determine how many outs the team will make for the number of PA, then adjust them up or down to match ~ 27 * 162

e) plug numbers into RC formula

f) do the same for pitching

g) place leaguewide numbers in a spreadsheet, balance leagues, and figure out pythag (using 1.83 factor) records

(Step d) is at the heart of what what we're discussing here...obviously with more BB and less outs, the team's overall production goes up. I haven't run numbers with all OBP guys vs all SLG guys, so a mix of both may be the ideal. But in general, OBP = desired just because you keep rolling over the lineup)

Some minor adjustments and steps may be in the process that aren't detailed, but that's the basic approach. Works pretty well too. :)

http://www.thefobl.com/forums/showthread.php?t=18514&highlight=faces+projections
http://www.thefobl.com/forums/showthread.php?t=17832&highlight=faces+projections
http://www.thefobl.com/forums/showthread.php?t=16896&highlight=faces+predictions
http://www.thefobl.com/forums/showthread.php?t=16041&highlight=faces+projections

[/end partially self-promotional tangent]

:D