Measuring Baseball players across eras and the problems with OPS+ and league average measures in the
Skin2
Posts: 1,259 ✭✭✭
in Sports Talk
There are many ways to measure baseball players, and one of the best ways is to measure them against their peers, and in this way, it is possible to measure players from different eras. The theory being that, since they are being measured against their peers, they are being measured against players who got to play in the same environment. From there, the players who shine the brightest among their peers, are usually viewed as the brightest players(even across eras).
Stats like OPS+(and nearly all the other advanced stats), measure the player against the league average...and they are very accurate stats for measuring offense.
However, it is my contention that measuring vs the league average(peers) isn't exactly accurate, because not all peers are created equal, and not all players benefit from the various environments that go into creating every players performance.
In measurements such as OPS+ the elite hitters of the live ball era benefited more from the environment that created the offensive explosion in the mid 90's, as opposed to the elite hitters of the more 'stable' environment of the preceding era.
You will see that the elite players were able to separate themselves from league average in the live ball era to a greater degree, as opposed to the era just before, mainly because they had more 'lesser' players to compete with, and because they were more equipped than the peasants of the league to take advantage of the reasons for the offensive explosion.
There are many reasons why there was an offensive explosion starting in the mid 90's; the introduction of a lively ball, building of smaller parks, bringing the fences in closer on existing parks, adding four teams worth of expansion players, shrinking of the strike zone, Body armor for hitters, no inside pitching allowed, and probably more.
Those things will help guys like Thome more so than the Felix Fermine types. What happens is that this allows the elite to separate further from the peasants...and since the league average represents the contributions of the peasants...it is easier for the elite to separate from the league average, hence higher OPS+ for players in the live ball era.
This explosion began in 1993 and took off in 1994, seemingly overnight. Steroids are a gray area, as they were present both before and after the offensive explosion. There is no exact time frame when they took hold, and it was a very gradual process. Also, the offensive explosion happened nearly immediately.
On to the meat of the results. Below are the average OPS+ of the top ten hitters in MLB for the given year. Instead of using just one year examples, I grouped them into five year periods(with specific break points for the five years). This way, it gives more data and leverages the expected one year blips on the radar...makes for a better sample size.
Average of the Top Ten League Leaders in OPS+
'98-'02....172
'93-'97....172
Those eight years above are the heart of the live ball era and the start of expansion.
'88-'92....160
'83-'87....153
'77-'82....157
'72-'76....158
As you can see, in the eras just preceding the live ball era, the elite hitters were not able to separate themselves from the league average like the elite hitters of the live ball era did(for reasons pointed out in beginning explanation). Since it is over several years with several different players on the leaderboards, it helps leverage out the career year(or one player) skewing the results.
The one thing to pay attention to mostly is the era just before and after the live ball era, because the league consisted of many of the same players in both eras. I put the other years in there to see some consistency through the years.
My explanation in the beginning says that the peers are not all equal, hence the reasons for the elite players to more easily separate from the league average. Listed below you will see the average of the five WORST hitters in the league for the specific time periods. The players listed are the worst full-time players(ones with enough PA to qualify for batting title).
We will call these guys the peasants of the league.
Average of the peasants of the league OPS+
'98-'02...61.3
'93-'01...60.8
'88-'92...66.2
'83-'87...65.8
'77-'80...63.8
As you can see, the peasants in the live ball era are worse than the peasants in the preceding era. What this does, is that the peasants in the live ball era are dragging down the league average OPS, thus making it easier for the elite hitters to separate from the league average.
In the 'normal' preceding era, it is harder to separate from the league average because there are less peasants in the league dragging it down, and the environment isn't tailor made to the attributes of the elite hitters(like it is in the live ball era).
Of course, there are even more peasants that are not full time players(and more just above these peasants), and they drag it down more than just the bottom few are doing.
This is why even in a measurement like OPS+, it has to take this into account when measuring across eras. There are other problems in all stats(such as how to tackle career length), but that is another topic.
As an example of how the OPS+ cross era measurement can be skewed, look at these two players and their best season in OPS+
1984 Eddie Murray.....157
1999 Rafael Palmeiro 160
If the average fan looks at each of those years and uses OPS+ to measure who had the better season, it will look like Palmeiro did.
However, a little logic should kick in when you realize that Murray's figure represented the best of ALL of MLB in 1984, while Palmeiro's being the sixth best in MLB for his season. That should tell you which one carried more weight.
OR, looking at the average of the top ten figures above, and comparing them to a pool of players more similar to them:
The top ten average for the league in Murray's time frame was 153...with Murray being at 157
The top ten average for the league in Palmeiro time frame was 172...with Palmeiro being at 160.
So when you compare them against similar players(elite hitters), Murray is ABOVE average, while Palmeiro is BELOW....a far different perspective when the peasants are out of the picture!
When you add the peasants of the league, thus altering the fairness of the peers(with Palmeiro having more peasants to compete against), it makes it look like Palmeiro was better.
This is not an advertisement against the use of OPS+, it is a great stat for comparing players within the same eras(and same career length). Nearly all advanced measures compare against the league average, so they will run the risk of this problem too.
The same can be said for another 'live ball' era, and that is pre war of 1941 and before (where more peasants were around as even though there were less teams, the available population to draw from was far lower), and also elite talent was excluded from the league due to segregation. So even doing THIS method above with a guy like Ruth won't do justice, because his league's top ten is missing players that weren't allowed to play.
Stats like OPS+(and nearly all the other advanced stats), measure the player against the league average...and they are very accurate stats for measuring offense.
However, it is my contention that measuring vs the league average(peers) isn't exactly accurate, because not all peers are created equal, and not all players benefit from the various environments that go into creating every players performance.
In measurements such as OPS+ the elite hitters of the live ball era benefited more from the environment that created the offensive explosion in the mid 90's, as opposed to the elite hitters of the more 'stable' environment of the preceding era.
You will see that the elite players were able to separate themselves from league average in the live ball era to a greater degree, as opposed to the era just before, mainly because they had more 'lesser' players to compete with, and because they were more equipped than the peasants of the league to take advantage of the reasons for the offensive explosion.
There are many reasons why there was an offensive explosion starting in the mid 90's; the introduction of a lively ball, building of smaller parks, bringing the fences in closer on existing parks, adding four teams worth of expansion players, shrinking of the strike zone, Body armor for hitters, no inside pitching allowed, and probably more.
Those things will help guys like Thome more so than the Felix Fermine types. What happens is that this allows the elite to separate further from the peasants...and since the league average represents the contributions of the peasants...it is easier for the elite to separate from the league average, hence higher OPS+ for players in the live ball era.
This explosion began in 1993 and took off in 1994, seemingly overnight. Steroids are a gray area, as they were present both before and after the offensive explosion. There is no exact time frame when they took hold, and it was a very gradual process. Also, the offensive explosion happened nearly immediately.
On to the meat of the results. Below are the average OPS+ of the top ten hitters in MLB for the given year. Instead of using just one year examples, I grouped them into five year periods(with specific break points for the five years). This way, it gives more data and leverages the expected one year blips on the radar...makes for a better sample size.
Average of the Top Ten League Leaders in OPS+
'98-'02....172
'93-'97....172
Those eight years above are the heart of the live ball era and the start of expansion.
'88-'92....160
'83-'87....153
'77-'82....157
'72-'76....158
As you can see, in the eras just preceding the live ball era, the elite hitters were not able to separate themselves from the league average like the elite hitters of the live ball era did(for reasons pointed out in beginning explanation). Since it is over several years with several different players on the leaderboards, it helps leverage out the career year(or one player) skewing the results.
The one thing to pay attention to mostly is the era just before and after the live ball era, because the league consisted of many of the same players in both eras. I put the other years in there to see some consistency through the years.
My explanation in the beginning says that the peers are not all equal, hence the reasons for the elite players to more easily separate from the league average. Listed below you will see the average of the five WORST hitters in the league for the specific time periods. The players listed are the worst full-time players(ones with enough PA to qualify for batting title).
We will call these guys the peasants of the league.
Average of the peasants of the league OPS+
'98-'02...61.3
'93-'01...60.8
'88-'92...66.2
'83-'87...65.8
'77-'80...63.8
As you can see, the peasants in the live ball era are worse than the peasants in the preceding era. What this does, is that the peasants in the live ball era are dragging down the league average OPS, thus making it easier for the elite hitters to separate from the league average.
In the 'normal' preceding era, it is harder to separate from the league average because there are less peasants in the league dragging it down, and the environment isn't tailor made to the attributes of the elite hitters(like it is in the live ball era).
Of course, there are even more peasants that are not full time players(and more just above these peasants), and they drag it down more than just the bottom few are doing.
This is why even in a measurement like OPS+, it has to take this into account when measuring across eras. There are other problems in all stats(such as how to tackle career length), but that is another topic.
As an example of how the OPS+ cross era measurement can be skewed, look at these two players and their best season in OPS+
1984 Eddie Murray.....157
1999 Rafael Palmeiro 160
If the average fan looks at each of those years and uses OPS+ to measure who had the better season, it will look like Palmeiro did.
However, a little logic should kick in when you realize that Murray's figure represented the best of ALL of MLB in 1984, while Palmeiro's being the sixth best in MLB for his season. That should tell you which one carried more weight.
OR, looking at the average of the top ten figures above, and comparing them to a pool of players more similar to them:
The top ten average for the league in Murray's time frame was 153...with Murray being at 157
The top ten average for the league in Palmeiro time frame was 172...with Palmeiro being at 160.
So when you compare them against similar players(elite hitters), Murray is ABOVE average, while Palmeiro is BELOW....a far different perspective when the peasants are out of the picture!
When you add the peasants of the league, thus altering the fairness of the peers(with Palmeiro having more peasants to compete against), it makes it look like Palmeiro was better.
This is not an advertisement against the use of OPS+, it is a great stat for comparing players within the same eras(and same career length). Nearly all advanced measures compare against the league average, so they will run the risk of this problem too.
The same can be said for another 'live ball' era, and that is pre war of 1941 and before (where more peasants were around as even though there were less teams, the available population to draw from was far lower), and also elite talent was excluded from the league due to segregation. So even doing THIS method above with a guy like Ruth won't do justice, because his league's top ten is missing players that weren't allowed to play.
0
Comments
<< <i>the introduction of a lively ball >>
Has the "lively" ball ever been confirmed? Just asking, you guys follow the sport more closely than I do.
<< <i>
<< <i>the introduction of a lively ball >>
Has the "lively" ball ever been confirmed? Just asking, you guys follow the sport more closely than I do. >>
No official confirmation.
So, if we were going to use OPS+ to compare players from different eras, is there any way we could quantify the differences? Or is this something you should just keep in mind when looking at the stat? In other words, is there some kind of formula you could think of that could show how much better Murray's 157 OPS+ was than Palmeiro's 160?
<< <i>Interesting . . .
So, if we were going to use OPS+ to compare players from different eras, is there any way we could quantify the differences? Or is this something you should just keep in mind when looking at the stat? In other words, is there some kind of formula you could think of that could show how much better Murray's 157 OPS+ was than Palmeiro's 160? >>
I was working on that several years ago, and was even looking at tying sociological impacts on the game as well...especially with regard to the pre-war era. I simply don't have that much time to devote... so I haven't really went further.
In a short answer, there is no hard formula, and probably more just to keep in mind. However, looking at the five year charts above, and using them kind of like a ballpark factor(like I did with the Murray 157 vs Palmeiro 160), I think one can get a decent formula out of it.
As you can see, Murray's 157 was better than average, while Palmeiro's 160 was worse than average. Who said a player always had to be compared to league average?? It shines more light on the subject, and compares players to a more fair pool of players, as opposed to a much varied(and unfair or tainted) pool of players.
<< <i>In measurements such as OPS+ the elite hitters of the live ball era benefited more from the environment that created the offensive explosion in the mid 90's, as opposed to the elite hitters of the more 'stable' environment of the preceding era. >>
Wouldn't that be a reason why stats like OPS+ is so good? Baseballreference.com has Palmerio as 45 runs better than average in 1999; they have Murray as 41 runs better than average in 1984. If we believe the accuracy of the formula, that means had both the Rangers and Orioles replaced those players with league average hitting the changes in the score board and the standings would have been nearly identical
Any of the problems you describe only come when you try to give these stats more meaning than what they are really measuring
<< <i>
<< <i>In measurements such as OPS+ the elite hitters of the live ball era benefited more from the environment that created the offensive explosion in the mid 90's, as opposed to the elite hitters of the more 'stable' environment of the preceding era. >>
Wouldn't that be a reason why stats like OPS+ is so good? Baseballreference.com has Palmerio as 45 runs better than average in 1999; they have Murray as 41 runs better than average in 1984. If we believe the accuracy of the formula, that means had both the Rangers and Orioles replaced those players with league average hitting the changes in the score board and the standings would have been nearly identical
Any of the problems you describe only come when you try to give these stats more meaning than what they are really measuring >>
Not if you are trying to compare Palmeiro to Murray it isn't.
<< <i>Any of the problems you describe only come when you try to give these stats more meaning than what they are really measuring >>
Not if you are trying to compare Palmeiro to Murray it isn't. >>
If the Rangers and Orioles replaced both those players with league average hitters, the change in runs scored for the teams would have been very close. The change in wins from the loss of those runs would have been very close. Do you really disagree with that comparison?
<< <i>
<< <i>Any of the problems you describe only come when you try to give these stats more meaning than what they are really measuring >>
Not if you are trying to compare Palmeiro to Murray it isn't. >>
If the Rangers and Orioles replaced both those players with league average hitters, the change in runs scored for the teams would have been very close. The change in wins from the loss of those runs would have been very close. Do you really disagree with that comparison? >>
That doesn't determine who was better. When one has an inferior set of leaguemates(dragging down the league average) as replacements(Palmeiro), then he will be viewed incorrectly as better.
Also, it isn't law that there has to be a league average player as replacement. As you can see in the top tens, it is easier to find a star as a replacement. It may have to wait till the off season, but there are more players that could replace that level of performance of Palmeiro(as evidenced by the top tens).
<< <i>That doesn't determine who was better. >>
Which doesn't mean the problem is with the statistic; again, the real problems come when you try to give the stat more meaning than what it is measuring
<< <i>When one has an inferior set of leaguemates(dragging down the league average) as replacements(Palmeiro), then he will be viewed incorrectly as better. >>
Who views Palmeiro as better than Murray? If there is anyone, it is an extremely small minority
Collecting 1970s Topps baseball wax, rack and cello packs, as well as PCGS graded Half Cents, Large Cents, Two Cent pieces and Three Cent Silver pieces.
One thing steroids don't explain is the 'overnight' jump in offense. Such changes as steroid usage occur over a lengthier period of time, and in the case of MLB, offense took a jump in 1993, then in 1994 jumped into the stratosphere and stayed there until 2010.
There were steroid users before 1993, and after. The usage of steroids happened over time(for example, Bonds didn't start using until the late 90's), yet the jump in HR's and offense happened in one year and stayed consistent until 2009.
Steroid testing began in 2003 and players are still using steroids now(even with the testing).
It would seem that equipment would be the biggest culprit in the jump, most notably the ball. What else could cause such a jump that happens in one year(which stays then for another 16), and then drops again overnight, and is staying consistent again.
Equipment seems the most logical, with other factors such as expansion, batter body armor(which is now no longer allowwed), or the strike zone, either enhancing or furthering it(though random outputs on a yearly basis could hide something that is actually there making an effect).
The same thing occured when the live ball era began around 1920 when a new ball was introduced. There was a pronounced jump with the new ball(and new in game ball policy), then other factors(such as Ruth showing the effectiveness of the HR) that carried the HR output even further in another stage.
Unless Axtell was right, and Selig had a meeting in the winter of 1993 and passed out roids to everyone
1993, of course, coincided with addition of the Marlins and Devil Rays, and by extension, the employment of 25 pitychers who would not have otherwise been good enough to pitch at the MLB level. Expansion is another contributor to the inflation of offensive numbers, along with smaller ballparks and a livelier ball. The impact of PEDs on these results are more difficult to ascertain, for many reasons, some of which you pointed out, but cumultaively, guys using contributed to the inflated numbers we witnessed throughout the 90s and 2000s.
Collecting 1970s Topps baseball wax, rack and cello packs, as well as PCGS graded Half Cents, Large Cents, Two Cent pieces and Three Cent Silver pieces.