Baseball Savant Infield Outs above average explained

[HowAboutThat]

If Martin was -5,  and Iglesias is +12, is it possible to translate that, at least approximately, into wins? 

 

[Tony-OH]

11 minutes ago, Philip said:

If Martin was -5,  and Iglesias is +12, is it possible to translate that, at least approximately, into wins? 

 

For me personally, I find the "wins" metric to be a flawwed. Even WAR doesn't really mean you will win "x" amount of games more or less. This is why defensive metrics have traditionally used runs saved, but I've never been convinced that it very accurately depicts actual runs saved.

I believe OAA is just a way to see whether players are better or worse then average compared to their peers. I like the percentile metric to give you an idea where they stand against their peers, however it doesn't necessarily tell you  how much a 9th percentile SS hurts you vs a 60th percentile SS.

I also like using the success rate added to see the real impact. If they have a -5% success rate added, that means they would get 5% less outs than an average fielder at his position. Over 100 chances that's 5 outs.

[DirtyBird]

Christian Walker #15

[HowAboutThat]

@Tony-OH

Fair enough, so that means that Martin is bad, but we don’t know how much his badness cost us.

 I’m actually surprised at how disappointed I am to learn that Martin was one of the worst. 

Did they do historical rankings?? I was always a huge fan of buddy Bell and of course Adrian Beltre is never to be replaced, and we had a couple of pretty good guys too.

[Frobby]

10 hours ago, Philip said:

If Martin was -5,  and Iglesias is +12, is it possible to translate that, at least approximately, into wins? 

 

I will give you a different answer than Tony, but it’s based on theory more than actual game observations.   

First, you need to recognize that Iglesias played a lot more games than Martin.    He played 1169 innings at short compared to Martin’s 785, about 50% more innings.    So if you assumed Martin’s time was increased to Iglesias’ level, his -5 becomes -7 or -8.    Let’s be generous and call it -7.    That makes the difference between Iglesias and Martin 19 outs.

Calculations done for other defensive metrics show that on average one out equates to .6 runs.    That average is calculated by running some math on every base/out state, and looking at the frequency of how often they occur, and doing a weighted average.    In other words, making a tough play with the bases loaded and two outs has a different impact on runs scored than making that same play with two out and nobody on base.    But when you look at the weighted average of all situations, the impact of making vs. failing to make an out is roughly .6 runs.    So, 19 outs equals 11-12 runs.

Finally, as a general rule, 10 runs greater or fewer over a season leads to one more or fewer win.    Again, that’s an average.    Depending on exactly when those runs were prevented or allowed, the impact could be much greater.    But applying that average, the difference between Iglesias and Martin should lead to about one extra win.    Obviously, the reality could be different depending on the game situations when these plays occur.    There are times where one play decides the outcome of the game, and other times where the play has no influence on the outcome whatsoever.    

[Number5]

2 hours ago, Frobby said:

I will give you a different answer than Tony, but it’s based on theory more than actual game observations.   

First, you need to recognize that Iglesias played a lot more games than Martin.    He played 1169 innings at short compared to Martin’s 785, about 50% more innings.    So if you assumed Martin’s time was increased to Iglesias’ level, his -5 becomes -7 or -8.    Let’s be generous and call it -7.    That makes the difference between Iglesias and Martin 19 outs.

Calculations done for other defensive metrics show that on average one out equates to .6 runs.    That average is calculated by running some math on every base/out state, and looking at the frequency of how often they occur, and doing a weighted average.    In other words, making a tough play with the bases loaded and two outs has a different impact on runs scored than making that same play with two out and nobody on base.    But when you look at the weighted average of all situations, the impact of making vs. failing to make an out is roughly .6 runs.    So, 19 outs equals 11-12 runs.

Finally, as a general rule, 10 runs greater or fewer over a season leads to one more or fewer win.    Again, that’s an average.    Depending on exactly when those runs were prevented or allowed, the impact could be much greater.    But applying that average, the difference between Iglesias and Martin should lead to about one extra win.    Obviously, the reality could be different depending on the game situations when these plays occur.    There are times where one play decides the outcome of the game, and other times where the play has no influence on the outcome whatsoever.    

I'm not a mathematician, so I can't back my opinion up on this, but .6 runs per out seems extremely high to me.  My common sense tells me that there isn't an average of 1.8 runs scored every half inning, even with juiced up balls.  Again, I have no training or experience in this field at all, but it just looks way off to my simple mind.  Can you explain this to me in terms that I can understand?

[Can_of_corn]

1 minute ago, Number5 said:

I'm not a mathematician, so I can't back my opinion up on this, but .6 runs per out seems extremely high to me.  My common sense tells me that there isn't an average of 1.8 runs scored every half inning, even with juiced up balls.  Again, I have no training or experience in this field at all, but it just looks way off to my simple mind.  Can you explain this to me in terms that I can understand?

How much would scoring increase if it was four outs per inning? 

[Number5]

Just now, Can_of_corn said:

How much would scoring increase if it was four outs per inning? 

I have no idea.

[Can_of_corn]

3 minutes ago, Number5 said:

I have no idea.

I don't either but I bet it would be significant.

Stuff like this I like to just assume that the guys that like crunching numbers did it correctly or the other guys that like crunching numbers would have corrected them.

[Number5]

28 minutes ago, Can_of_corn said:

I don't either but I bet it would be significant.

Stuff like this I like to just assume that the guys that like crunching numbers did it correctly or the other guys that like crunching numbers would have corrected them.

I hear ya, but that just seems way too high to me.  Seems like .6 runs would be closer to the number for the entire half inning, not one out.  That would be 5.4 runs per team per game by simple math ( 9 X .6.)  I'm sure the formula that whoever came up with .6 runs per out came up with is far more complex than that.  I've found in life that there are times that more complex isn't always better.  It would be nice to at least be given a reasonable explanation of how they arrived at the number and a reason that the actual number of runs scored versus outs made would be an incorrect way to do it.  Again, I'm no mathematician, but I do like to be able to at least have something make some kind of sense to me.

[Frobby]

On 1/16/2020 at 11:08 AM, Number5 said:

I'm not a mathematician, so I can't back my opinion up on this, but .6 runs per out seems extremely high to me.  My common sense tells me that there isn't an average of 1.8 runs scored every half inning, even with juiced up balls.  Again, I have no training or experience in this field at all, but it just looks way off to my simple mind.  Can you explain this to me in terms that I can understand?

I know it seems counterintuitive, since there are 27 outs in a game and only 4-5 runs scored on average, so that seems like more like .2 runs per out.   

But have a look at this base-out matrix, which shows the expected number of runs scored in an inning depending on which bases are occupied and how many outs there are.   This matrix is based on actual game data, not theory.   https://www.google.com/amp/s/thebaseballscholar.com/2017/08/14/sabermetrics-101-re24/amp/

Let’s take the simplest example, the leadoff batter.    If he grounds out (so now it’s bases empty, one out), the expected runs in the inning are .243.    If he’s safe at first (so runner on first, nobody out), the expected runs are .831.    So, the increase in expected runs between making that play or not is .590.     

There are other situations on the chart where the differential is higher or lower than that.    On the high side, let’s say you’ve got runners on 1st and 2nd and one out.    Make a play there to record an out and hold the runners, and the expected runs are only .343.    Fail to make that play and so it’s bases loaded and only one out, and the expected runs increase to 1.520.

Anyway, someone did a weighted average of all these scenarios and came up with .6 for an average.    If I remember correctly.   
 

 

[Just Regular]

I was curious how much of baseball OAA might hint is fielding, so spreadsheeted the last 3 years for which Savant has published both IF and OF.

The Astros were 1st at +71

The Orioles were 30th at -103

So 174 outs at the estimated 0.6 runs/out gives 104 runs, suggesting ~35 runs/season defensive difference between the best and worst teams.

Across the 3-years, the Astros outscored their opponents by 739 runs and the Orioles were outscored by 620.  104/1359 gives 7.5%, leaving 92.5% for batting, pitching and catcher defense.

The Astros especially lorded over the AL here - Brewers, Reds, Cubs, Diamondbacks, Braves and Rockies were the next six.  Angels got AL second place at +31, and the third place AL finishers were down at +17.

[Frobby]

32 minutes ago, OrioleDog said:

I was curious how much of baseball OAA might hint is fielding, so spreadsheeted the last 3 years for which Savant has published both IF and OF.

The Astros were 1st at +71

The Orioles were 30th at -103

So 174 outs at the estimated 0.6 runs/out gives 104 runs, suggesting ~35 runs/season defensive difference between the best and worst teams.

Across the 3-years, the Astros outscored their opponents by 739 runs and the Orioles were outscored by 620.  104/1359 gives 7.5%, leaving 92.5% for batting, pitching and catcher defense.

The Astros especially lorded over the AL here - Brewers, Reds, Cubs, Diamondbacks, Braves and Rockies were the next six.  Angels got AL second place at +31, and the third place AL finishers were down at +17.

As you note, OAA doesn’t account for catcher defense.    It also doesn’t account for certain elements of outfield defense, i.e. cutting off balls in the gap/down the lines or throwing. Outfield OAA relates purely to how many balls get caught.    So, you can probably assume that the difference between the best and worst defensive teams is greater than 35 runs.   By Fangraphs’ reckoning, it’s more like 75 runs (about 15-20 of which is catcher framing).    

[weams]

 

[tntoriole]

Stats or not...this play is sick.  I am looking forward towatching this guy pick it.