An interesting read

[USAFChief]

http://www.slate.com/articles/sports/the_next_20/2016/09/fire_joe_morgan_and_the_moneyball_revolution.html

[Mike Sixel]

I note that there were no quotes from old timey people.....I think I didn't see any.

[TheLeviathan]

Just goes to show that metrics, like with many other things, rarely stand the test of time as infallible.  They're always going to be improved upon and weaknesses will be exposed.  Humility about their certainty is always wise.

[wsnydes]

 

Just goes to show that metrics, like with many other things, rarely stand the test of time as infallible.  They're always going to be improved upon and weaknesses will be exposed.  Humility about their certainty is always wise.

As a guy that doesn't get into many of the advanced metrics, I can't agree with this enough.  Old fashioned eye tests and gut feeling still need to have a place.  Stats are just another tool to use in evaluating talent.

[big dog]

It's also why statisticians say you can't prove anything, although you can show that a theory does or does not have empirical support given existing data and statistical methods.

[JustinCB]

Just goes to show that metrics, like with many other things, rarely stand the test of time as infallible.  They're always going to be improved upon and weaknesses will be exposed.  Humility about their certainty is always wise.

The point is, it's important to be open minded and willing to change.

[TheLeviathan]

 

The point is, it's important to be open minded and willing to change.

 

Except there is a strong, vocal part of the metric side of the argument that doesn't concede the idea that there is still room for improvement.  Or that their stats may not tell the whole story.  Pitch framing was the article's example, but it goes for pretty much all of them.  Don't think you have it all figured out and too often, that is exactly the opposite of what the statistics crowd wants you to believe.   (In part, i imagine, because selling it as anything short of infallible sort of undermines the value of the stat)

 

There is a lot of hubris that often looks foolish in the long run as better ways to analyze are developed.

[gunnarthor]

I remember way back when in the 90s when ESPN was barely a thing and we could have message boards and rants and such.  Tom Kelly was complaining about Todd Walker's defense at second base and someone in the press - it might actually have been Neyer - ripped Kelly for worrying about defense when a guy was hitting as good as Walker.  And now, years later, we see that he was nearly a -2 WAR defender.  

 

 

[Mike Sixel]

 

Except there is a strong, vocal part of the metric side of the argument that doesn't concede the idea that there is still room for improvement.  Or that their stats may not tell the whole story.  Pitch framing was the article's example, but it goes for pretty much all of them.  Don't think you have it all figured out and too often, that is exactly the opposite of what the statistics crowd wants you to believe.   (In part, i imagine, because selling it as anything short of infallible sort of undermines the value of the stat)

 

There is a lot of hubris that often looks foolish in the long run as better ways to analyze are developed.

 

Can you name one? I can't. Not one, not anymore. That quote was from a long time ago. Can you find a recent one, from anyone in the stats community that matters? Because I can still find them from the other side, every day, most broadcasts.

[TheLeviathan]

 

Can you name one? I can't. Not one, not anymore. That quote was from a long time ago. Can you find a recent one, from anyone in the stats community that matters? Because I can still find them from the other side, every day, most broadcasts.

 

I think pretty much every discussion about WAR hedges on it, here and in most places.  I'm also pretty confident that in 20 years we'll look back and scoff at the idea WAR told us much of anything useful.  

 

I agree, however, that the old-school guys are bad about it too.  There is an unnecessary combination of aggression and defensiveness from both sides.

[Mike Sixel]

Really, I can't recall one person saying WAR is perfect. As a matter of fact, they mostly all say they are tinkering with it, wondering how it can be better, and that in any event, it is NOT the only thing to think about. We are clearly reading different websites. 

[gunnarthor]

 

Really, I can't recall one person saying WAR is perfect. As a matter of fact, they mostly all say they are tinkering with it, wondering how it can be better, and that in any event, it is NOT the only thing to think about. We are clearly reading different websites. 

Generally, a guy like Schonfield will have an article at the beginning of the season that says WAR has faults.  And then, for the rest of the season, he relies a lot on WAR.  Now, we all do that, but it is irksome.  

[USAFChief]

 

Really, I can't recall one person saying WAR is perfect. As a matter of fact, they mostly all say they are tinkering with it, wondering how it can be better, and that in any event, it is NOT the only thing to think about. We are clearly reading different websites.

Mike, I doubt it would take me long to find a quote on Fangraphs arguing Player A deserves MVP over Player B, based solely on WAR.  In fact, arguing that using anything BUT WAR to determine MVP is...dumb.

 

Would you disagree with that?

[TheLeviathan]

 

Really, I can't recall one person saying WAR is perfect. As a matter of fact, they mostly all say they are tinkering with it, wondering how it can be better, and that in any event, it is NOT the only thing to think about. We are clearly reading different websites. 

 

You also strawmanned that point quite a bit.  There is certainly a heavy emphasis on WAR.  You see it all over this site in fact.  People who will agree that the defensive component of WAR is best taken in multiple year averages and rarely reliable as a data set even in single years, and then still use that very same statistic to measure players against each other on a year to year basis.

 

And if you dare to point out that this doesn't make sense, you're basically shunned for not understanding it.

 

I think I do understand WAR.  It's why you'll never hear me use the stat unless i'm comparing players who play the same position over the course of multiple years.  Beyond that the stat is given to large swings of accuracy.

 

Look out in the SABR community and you see little regard for that.  In fact, you see open hostility to anyone questioning the metric.

[Platoon]

I was surprised there was ever a discussion on hots streaks? Or did I misread that part. Averages are what they are, averages over a longer time frame, year career etc. Streaks are in essence SSS, and have existed forever. As to clutch hitting I have long been a believer that clutch hitters are somewhat of an illusion. How so? I think that most of them simply perform to their overall career stats, in important situations. Basically not letting the magnitude of the event over stress them, and letting themselves "slow down", so as not to over react and do something they can't. For if simply trying harder would be enough to give you better results, then everyone would do it, and do it in every game, as trying is simply effort compounded. As to pitch framing. My neighbor and long time friend was a D2 caliber catcher. He caught for years in men's amateur ball, and then umpired for something like 26 years to levels in the state tournament range. Men's, high school and Legion. Ho told me years ago how he showed pitches to an umpire by glove postion, and lack of glove movement, and not being greedy on the ones you can't save. And he said that the same applied to an umpire. A catchers low stance, and how he sets up and receives matter, a lot. While it took stats to more widely make the concept acceptable, cathers and umpires have known it for years, although I doubt most umpires are willing to admit it.

[Mike Sixel]

Why does a hot streak end, if it is a not just random, but an actual thing? That was the question.....if something causes you to be hot, and keeps you hot (i.e., is a streak, not random), why would it ever end? That was the root question, I believe.

 

Like, Baltimore won a boatload of 1 run games, and their fans claimed it was a skill. Then it stopped happening, even with pretty much the same players and coaches....was it random? A streak? A skill?

[Craig Arko]

Why does a hot streak end, if it is a not just random, but an actual thing? That was the question.....if something causes you to be hot, and keeps you hot (i.e., is a streak, not random), why would it ever end? That was the root question, I believe.

 

Like, Baltimore won a boatload of 1 run games, and their fans claimed it was a skill. Then it stopped happening, even with pretty much the same players and coaches....was it random? A streak? A skill?

Nonlinear stochastic process.

 

You asked.

[Thrylos]

Here is the thing (and this article perfectly describes it) :

 

There are phenomena, like pitch framing, that existed for ever and there are measures that try to quantify that phenomena.  Going from qualitative ('he is a great framer') to quantitative ('his framing skills produced x more strikes than the average catcher') is a hard work.  And it gets harder as some of the data points (like umpire strike zones) are subjective and movable targets.

 

Are all metrics perfect?  No, there is no such a thing. But of course they are better than "gut feelings" and "eyeballing" because there has been a whole lot of effort to go from qualitative to quantitative and from subjective to objective.  As there will be more and more data points with the new systems installed that collect raw data, metrics will improve.

 

But sitting down and blasting a metric without even bothering to understand how it is derived and what it is trying to do, is plain lazy.  People put some work in developing those, at least try to figure out what a metric is trying to do, and if you try to blast it, blast it with objective criteria (eg. "this particular parameter is not a true measurement of that") vs. subjective (eg. "my eyes tell a different story".)

 

WAR has been killed to death and misused and abused.  It is a great metric to show overall performance over a long period of time (a season, a career etc.) and it is better than batting average, home runs and RBI that have been used in the good old days...  Now is someone is trying to use WAR for things it was not meant to be, like looking at potential future performance, shame on them, not on WAR...

 

The beauty of baseball is in the fact that there is a great degree of uncertainty that cannot be measured/quantified (how do you quantify UCL conditions or "stuff that happens".)  And this uncertainty over 162+ games almost makes predictions based on metrics, practically invalid.

 

Which is the fun part of it

 

On the other hand, metrics are getting better in describing what happened, when and why.  Time will tell, but I do foresee a new generation of metrics with all those data points, based on if/when type of situations improving over a lot of metrics like WAR, UZR, etc

 

Just my couple of Lincolns

[Hosken Bombo Disco]

"The jocks vs. nerds dynamic that took root in the 1990s and 2000s encouraged smart people to pick sides when the meaningful divide was the curious vs. incurious."

 

Sounds right to me.

[jimmer]

Concerning defensive metrics.  

 

There's a difference between looking at defensive metrics for a longer period than one season to determine a person's actual defensive ability and looking at defensive metrics a season at a time to see how they performed for THAT season.  So yes, you can absolutely use defensive stats for this year alone to tell how a person did THIS year alone. Dave Cameron just addressed this recently when someone asked a similar answer.  He said that how a player performs on defense fluctuates season to season (just like offense and pitching does), and that season stats still tell what happened that season even if it doesn't give the players' true defensive abilities.

 

And, in that way, defensive metrics are the same as hitting and pitching stats, both traditional and advanced one. Danny Santana's 2014 offense happened, even though that wasn't his true skill level.  A player can under-perform and out-perform their abilities and norms offensively, on the mound, AND defensively.  Cant just look at the multiple years disclaimer given for defensive metrics and say the individual season stats mean doesn't tell us how they did that season.  

[TheLeviathan]

 

 Cant just look at the disclaimer given about defensive metrics and say the individual season stats mean doesn't tell us how they did that season.

 

This isn't particularly accurate.   WAR, on Fangraphs, uses UZR does it not?

 

If so, it isn't nearly as reliable as you are making it out to be.  Fangraphs says so itself.  It's not "unreliable" (as the article says) but it also is significantly less reliable than even OPS.  And, depending on the position the player mans, more limited for some players than others.  

 

And yet, despite these demonstrable limitations, we toss year to year WAR comparisons around (across different positions) like we're totally unaware of these issues.

 

And so help you if you besmirch WAR even in the slightest....

[biggentleben]

I do see where WAR is cited in MVP arguments by those who know that WAR is not a finished metric, but it's the best we have right now. That is always how I utilize WAR, and frequently I will cite whether I'm using WARP, bWAR, or fWAR. It's actually something my editor and I made a point of bringing up with other writers on our site.

 

I've had some great conversations with writers who are on both sides of sabermetrics (write scouting reports and you seem to get far ends on both arguments!), and I think in all seriousness, most of those who are very into finding advanced metrics use what is there, hoping a better one is out there that will give a better, more accurate, more full evaluation of a player, while the other side assumes that just using the one that currently exists as believing that is the end-all, be-all statistic. I think it's a bad perception issue, frankly.

[jimmer]

 

I do see where WAR is cited in MVP arguments by those who know that WAR is not a finished metric, but it's the best we have right now. That is always how I utilize WAR, and frequently I will cite whether I'm using WARP, bWAR, or fWAR. It's actually something my editor and I made a point of bringing up with other writers on our site.

 

I've had some great conversations with writers who are on both sides of sabermetrics (write scouting reports and you seem to get far ends on both arguments!), and I think in all seriousness, most of those who are very into finding advanced metrics use what is there, hoping a better one is out there that will give a better, more accurate, more full evaluation of a player, while the other side assumes that just using the one that currently exists as believing that is the end-all, be-all statistic. I think it's a bad perception issue, frankly.

I have been going away from using WAR when comparing position players who play the same position. I usually look at wRC+ and DRS together cause I like DRS better than UZR.  When comparing value of position players to players from other positions I like WAR because it takes into account positional adjustment (meaning if a 1B and SS have the same numbers, it will show the SS to be the more valuable player).

 

Basically, once you know that no matter what you use it won''t be 100% perfect, you just look for the ones that are the best available at the time.  Like I know UZR and DRS aren't perfect, but I definitely think they are better than using errors and fielding %.

[jimmer]

 

This isn't particularly accurate.   WAR, on Fangraphs, uses UZR does it not?

 

If so, it isn't nearly as reliable as you are making it out to be.  Fangraphs says so itself.  It's not "unreliable" (as the article says) but it also is significantly less reliable than even OPS.  And, depending on the position the player mans, more limited for some players than others.  

 

And yet, despite these demonstrable limitations, we toss year to year WAR comparisons around (across different positions) like we're totally unaware of these issues.

 

And so help you if you besmirch WAR even in the slightest....

About UZR, it says: 'Once you get to one and three-year samples, it’s a relatively solid metric but defensive itself is quite variable so you need a good amount of data for the metrics to become particularly useful.'

 

Says right there a one year sample is a solid metric and Cameron has said the same thing in his chats.  Thing is, it's not when trying to figure out a player's true talent level because, like offense, people have good and bad years on defense. But it is okay to use to look at how a player did for a specific year.

 

Now if you're pointing out that the UZR primer says it's not 100% accurate, well, few things are. I've never said they are 100% accurate or that WAR is (or that wRC+ is).  Only that they are the best we have to try and determine overall value of a player.  But I've gotten to the point where, depending on where I chat, I'm not going to try and continue to convince people to go away from their devotion to traditional stats, cause most like to say since the advanced ones aren't perfect, they shouldn't be used while ignoring how flawed the traditional ones are at telling the story of a player's talent. Traditional stats are a cozy blanket coated in tradition and memories.

 

I'll just point out why I think a good chunk of them shouldn't be counted on to tell the story they've been led to believe they tell them and why I think another stat does it better. Nothing more to do after that.  

 

And I don't know why OPS is being brought up at all.  OPS isn't a defensive stat.  But what has been pointed out is that OBP is very undervalued in the equation for OPS.

[TheLeviathan]

You realize "relatively solid" relevant to other defensive metrics is a pretty low bar right?

 

I don't rely on traditional stats and prefer a variety of metrics (including the ones you suggest ), but I don't prop WAR on such a pedastal as is commonly done.

[Mike Sixel]

There is an article on FG right now about injured players, looking at stats, and trying to judge true talent levels.....

[USAFChief]

 

About UZR, it says: 'Once you get to one and three-year samples, it’s a relatively solid metric but defensive itself is quite variable so you need a good amount of data for the metrics to become particularly useful.'

 

Says right there a one year sample is a solid metric and Cameron has said the same thing in his chats.  Thing is, it's not when trying to figure out a player's true talent level because, like offense, people have good and bad years on defense. But it is okay to use to look at how a player did for a specific year.

 

Now if you're pointing out that the UZR primer says it's not 100% accurate, well, few things are. I've never said they are 100% accurate or that WAR is (or that wRC+ is).  Only that they are the best we have to try and determine overall value of a player.  But I've gotten to the point where, depending on where I chat, I'm not going to try and continue to convince people to go away from their devotion to traditional stats, cause most like to say since the advanced ones aren't perfect, they shouldn't be used while ignoring how flawed the traditional ones are at telling the story of a player's talent. Traditional stats are a cozy blanket coated in tradition and memories.

 

I'll just point out why I think a good chunk of them shouldn't be counted on to tell the story they've been led to believe they tell them and why I think another stat does it better. Nothing more to do after that.  

 

And I don't know why OPS is being brought up at all.  OPS isn't a defensive stat.  But what has been pointed out is that OBP is very undervalued in the equation for OPS.

Recent views on that subject aren't uniform:

http://www.fangraphs.com/community/when-slugging-percentage-beats-on-base-percentage/

 

[Thrylos]

 

Recent views on that subject aren't uniform:

http://www.fangraphs.com/community/when-slugging-percentage-beats-on-base-percentage/

 

The author fails to mention that the +/- 5% deviation in that graph is not significant based on sample size... 

[USAFChief]

 

The author fails to mention that the +/- 5% deviation in that graph is not significant based on sample size... 

Please explain.

 

The sample size is 1911-2015.

[Thrylos]

 

Please explain.

 

The sample size is 1911-2015.

 

Horizontal.  I mean for each column.  Vertically.  The sample is about 500 people or so whose SLG% and OBPs are measured.