Impact of Defense

[jay]

Fangraphs recently released an article projecting the Twins defense to rank 29th in the league. The defense was widely regarded to be sub-par last year and it seems that most people agree the defense will be below-average again this year.

 

The question becomes -- how much of an impact does that have?

 

The projections from Fangraphs show a difference of 65 runs from the best defense to the worst. However, to paraphrase another poster, these types of projections are never going to show as much variation as end of year figures given that some teams will outperform and others will underperform. Projections aren't geared to capture the extremes.

 

Two of the more common defensive measures, UZR and DRS, have calculated the difference between the best and the worst team defenses within the 2011-2014 time period to range from 113 runs to 195 runs. The average difference has been calculated as 147 runs. 

 

So, what say ye? Does that feel like too much, too little or just right? How much impact does defense have? 

[Brock Beauchamp]

I really struggle with the concept that defense can contribute more than half a run a game. I'm not saying it's impossible, just really hard to swallow.

 

Most defensive plays are gimmes or at the most, a marginal effort. Half a run a game equates to what, an extra double and a single in every game (using rough estimation here)? Or maybe roughly three singles. That's roughly one additional base hit every 12-14 PAs. I don't see it.

 

With that said, half a run a game is huge. That's several wins over the course of a season, enough to make a mediocre team good and a good team great.

[Brock Beauchamp]

To go a bit further, let's just say that the best defense saves 1.5 singles per game. At 0.5 runs per game saved, that's a fair number because for all intents and purposes, home runs do not count on defense and they're the only way to get one run from a single plate appearance. So to average half a run, you're going to need a bit more than a single.

 

In 2014, the Kansas City Royals gave up 1386 hits, good for fifth best in the AL.

 

So we'll add 234 hits to the Royals' 2014 total, saying they had the worst defense in the league (instead of the best).

 

That puts them at 1620 hits for the season.

 

In 2014, the Minnesota Twins gave up the most hits in the AL at 1588. These hypothetical Royals would have given up 1/4 of a hit more per game than the 2014 Twins, making them the worst in the AL by a healthy margin.

 

These numbers just don't make any sense and this is using (admittedly rough) math based on giving up just half a run more per game. Some of you are voting for one full run a game, or double that number.

 

Then we get into situations where defense has no impact whatsoever. Say a pitcher walks a guy, gives up a line drive that no defender on the planet could catch, and then serves up a homer. That is a pretty common scenario in baseball.

 

Three runs. The defense had zero ability to counteract any of those runs. Once we start whittling away at the scenarios where the defense could not positively nor negatively affect the outcome of runs being scored, the numbers go into insane WTF ludicrous territory.

 

Defense is important. It saves runs... but these numbers don't make a lick of sense if you take a moment to apply common sense to the metrics.

[jay]

 

Half a run a game equates to what, an extra double and a single in every game (using rough estimation here)? Or maybe roughly three singles. That's roughly one additional base hit every 12-14 PAs. I don't see it.

 

According to the linear weights you're referring to, I believe a half run a game is just an extra single in every game.  That multplies your PA math by 3.

 

Here's the link to tangotiger's work on linear weights, the scores for each event and the weighted average for each (based on the probability of each event): http://www.tangotiger.net/lwbymob.htm

[nicksaviking]

I picked zero simply because defensive measurements are still so entirely subjective.  It all basically boils down to saying Player X missed a play that Player Y would have made and it cost the team a run.  I don't know that Player Y would have made that play, as my machine that turns hypothetical judgments into fact recently broke. 

[Hosken Bombo Disco]

 

I picked zero simply because defensive measurements are still so entirely subjective.  It all basically boils down to saying Player X missed a play that Player Y would have made and it cost the team a run.  I don't know that Player Y would have made that play, as my machine that turns hypothetical judgments into fact recently broke. 

 

So that was you…

[Otto von Ballpark]

 

I really struggle with the concept that defense can contribute more than half a run a game. I'm not saying it's impossible, just really hard to swallow.

You might be right.

 

26 out of 30 teams were within 0.54 defensive runs per game of each other last year, by B-Ref.  (25 teams within 0.59 by Fangraphs.)

 

By comparison, to include 26 teams in the range for batting runs at B-Ref, the runs per game gulf widens to 0.86.  Pitching?  0.88 runs per game.  (Fangraphs drops the pitching range to 0.63 due to their use of FIP, but it is still higher than fielding.)

[Brock Beauchamp]

 

According to the linear weights you're referring to, I believe a half run a game is just an extra single in every game.  That multplies your PA math by 3.

 

Here's the link to tangotiger's work on linear weights, the scores for each event and the weighted average for each (based on the probability of each event): http://www.tangotiger.net/lwbymob.htm

I later adjusted my math in the second post to reflect 1.5 singles. I realized my math was bad there and was going from memory.

[Hosken Bombo Disco]

I appreciate you trying to apply the math to this, because it's hard for me to get my head around.

 

There were several plays last year. Joe Mauer hitting a line drive 8 or 10 feet inside the left field line, yet Alex Gordon was basically standing right in that spot. Is that fielding or just the shift? Danny Santana missed some balls in center early on but improved pretty well. Our right field was a mess all year long. What is the value of an outfielder with a good arm? Hard to tell. Feels so subjective and hard to measure. But defense has got to be worth something.

[jay]

 

I later adjusted my math in the second post to reflect 1.5 singles. I realized my math was bad there and was going from memory.

 

I can appreciate what you're trying to do there by simplifying and applying real scenarios, but I think you're leaving out a lot of places where defense impacts the outcome. 

 

A single is one possible outcome with the outfielder that let a ball fall in front of him that someone else would have caught or the infielder who let a ball get by that someone else would have stopped. However, you're also going to commonly have situations where that missed ball results in multiple bases, which have a significantly higher run expectation and add up much faster. You're also going to have situations where what would have been a single or double gets stretched into more due to lack of arm or range.

 

Equally as important, the factor that is hard to consider in your simplified version is all the events that happen after an out is truly missed. We see this a portion of this in unearned runs, but I think we'd mostly all agree there's more to defense than errors.

[Brock Beauchamp]

 

You might be right.

 

26 out of 30 teams were within 0.54 defensive runs per game of each other last year, by B-Ref.  (25 teams within 0.59 by Fangraphs.)

 

By comparison, to include 26 teams in the range for batting runs at B-Ref, the runs per game gulf widens to 0.86.  Pitching?  0.88 runs per game.  (Fangraphs drops the pitching range to 0.63 due to their use of FIP, but it is still higher than fielding.)

Half a run makes sense to me. In no way can defense have more of an impact than offense or pitching. There are simply too many things going against defense:

 

1. The gulf between best and worst defender is generally lower than best and worst hitter or best and worst pitcher.

 

2. Many defensive plays are gimmes. Going from memory, we'll say that 33% of all balls in play are flyballs. Pretty much every defender on the field will catch 75% of those flyballs that stay in the park. The same applies to infield grounders and other "gimme" plays that defenders make all day long.

 

3. Defense has zero impact on walks, zero impact on strikeouts, zero impact on HBP, and virtually no impact on home runs. Taking the AL average of BB/SO/HBP/HR, each team racks up about 11.8 of those per game when combined. That's roughly a dozen times per team per game the defense has no impact on what's happening.

[Otto von Ballpark]

 

To go a bit further, let's just say that the best defense saves 1.5 singles per game.

Why hits?  Why not just runs?

 

Adding 0.5 runs per game to the 2014 Royals would swing their run differential from +27 to -54.  It would drop their Pythag by 9 wins.   It would have taken them from the 4th best fielding team in MLB (by B-Ref) to the 5th worst.

 

Is that so unreasonable an outcome?

[Otto von Ballpark]

 

I appreciate you trying to apply the math to this, because it's hard for me to get my head around.

 

There were several plays last year. Joe Mauer hitting a line drive 8 or 10 feet inside the left field line, yet Alex Gordon was basically standing right in that spot. Is that fielding or just the shift? Danny Santana missed some balls in center early on but improved pretty well. Our right field was a mess all year long. What is the value of an outfielder with a good arm? Hard to tell. Feels so subjective and hard to measure. But defense has got to be worth something.

That's definitely interesting if you are trying to apportion credit between players (and even coaches/scouts).  But just for making the calculation on a team level, it doesn't really matter if it was positioning or speed or skill -- it just happened.

[Brock Beauchamp]

 

Why hits?  Why not just runs?

 

Adding 0.5 runs per game to the 2014 Royals would swing their run differential from +27 to -54.  It would drop their Pythag by 9 wins.   It would have taken them from the 4th best fielding team in MLB (by B-Ref) to the 5th worst.

 

Is that so unreasonable an outcome?

I used hits because others have used the runs argument. I'm giving an actual real world example how the math doesn't hold up. The Royals' pitching staff would have gone from very good to essentially tied for the Twins solely based on defense.

 

That doesn't make sense. The Royals had good starters. They had an elite bullpen. Yet you take away that defense and suddenly, they're a bottom five pitching team?

 

It doesn't add up.

[Otto von Ballpark]

 

Half a run makes sense to me. In no way can defense have more of an impact than offense or pitching.

Roughly half a run seems to span the vast majority of teams, but there will be a few outliers, just like batting and pitching.  And outliers don't invalidate the calculation.

 

1. The gulf between best and worst defender is generally lower than best and worst hitter or best and worst pitcher.

Seems true, but don't know if that's the right measure to calibrate this calculation, though.  All of the best hitters and pitchers are virtually always starting in MLB.  All of the best fielders almost certainly are not.  And the worst hitters and pitchers often get sent to the bench, minors, or DL pretty quick.  The worst fielders in that group that makes it to MLB probably stick around in MLB longer.

[jay]

 

We see this a portion of this in unearned runs, but I think we'd mostly all agree there's more to defense than errors.

 

Which brings up an interesting point.

 

Even just the defensive miscues that are reflected as the total of unearned runs shows a difference between the best and worst team defenses.  That difference has averaged 46 runs over 2011-2014.  And that's just on the blatant misplays categorized by a scorer as an error...

 

It'd be pretty groan inducing for someone to try to tell me defense does less than that ancient measure.

[Brock Beauchamp]

 

Roughly half a run seems to span the vast majority of teams, but there will be a few outliers, just like batting and pitching.  And outliers don't invalidate the calculation.

True, but as outlined in another thread, the gulf was as much as 125 runs a season in some metrics.

 

When a metric outputs that kind of calculation, we're kinda obligated to question its validity.

 

When the largest gulf between most and least runs allowed in the AL was ~160 runs and the gulf between defense is 125 runs while that defense metric has to fit inside the runs allowed metric (because pitching and defense are intertwined), the numbers enter surreal territory.

 

For all intents and purposes, the metrics are arguing that defense is worth more than pitching (125 defense fitting inside 160 runs allowed overall gap), yet defense doesn't touch the ball a significant portion of the time.

[Otto von Ballpark]

 

I used hits because others have used the runs argument. I'm giving an actual real world example how the math doesn't hold up. The Royals' pitching staff would have gone from very good to essentially tied for the Twins solely based on defense.

 

That doesn't make sense. The Royals had good starters. They had an elite bullpen. Yet you take away that defense and suddenly, they're a bottom five pitching team?

 

It doesn't add up.

Every one of the Royals 2014 starters outperformed their FIP.  They were 13th by FIP-, only 4% ahead of the Twins.  And their starters pitched over twice as many innings as their relievers.

 

Replacing one of the best defenses in MLB with one of the worst defenses in MLB might have that affect on such a staff.  I think you're discounting what it means to be a top-5 defense and a bottom-5 defense.

 

The Royals bullpen was a lot better, but bullpens pitch fewer innings.  It probably had an effect on their Pythag luck (Royals +5 Pythag wins, Twins -5).

 

If the 2014 Twins had that defense, and that bullpen/Pythag luck, they absolutely would have been a contender.  (Probably would have lost in the first round of the playoffs, though. )

[Brock Beauchamp]

 

Replacing one of the best defenses in MLB with one of the worst defenses in MLB might have that affect on such a staff.  I think you're discounting what it means to be a top-5 defense and a bottom-5 defense.

I'm really not.

 

The runs allowed gap in AL teams was 160 runs.

 

Could half of those runs (or, as some are arguing, more than half) be attributed to defense? Are some actually arguing that defense is more valuable than pitching?

 

That's my problem. I don't think anyone would argue that pitching is more valuable than defense yet the metrics some are touting suggest just that. The metrics don't work in real world application unless we're arguing "this team is better than that one", which I can get behind.

 

Again, this doesn't mean defense isn't valuable or that it can't drastically impact a season... But these run numbers are simply absurd.

[jay]

 

Roughly half a run seems to span the vast majority of teams, but there will be a few outliers, just like batting and pitching.  And outliers don't invalidate the calculation.

 

I think you make a valid point on outliers.  They are the hardest to consider, hardest to fathom and hardest to believe (from an accurancy standpoint) all at once.

 

If we remove the 5 best and 5 worst defenses from 2011-2014.  UZR and DRS saw an average of 65 runs separating those remaining 20 teams.  That's less than the half run per game being discussed. 

 

That's saying the difference between the 6th best defense and the 6th worst defense is equivalent to one single each game OR a double every other game OR the good defense preventing an extra base (hold a guy to a single instead of a double) and the bad defense allowing an extra base (stretch a single to a double) each game.  Any of those singularly happening roughly calculates out to 65 runs across the entire season. 

 

Seems entirely reasonable to me.

 

[Otto von Ballpark]

 

When the largest gulf between most and least runs allowed in the AL was ~150 runs and the gulf between defense is 125 runs while that defense metric has to fit inside the runs allowed metric (because pitching and defense are intertwined), the numbers enter surreal territory.

Actually, both WARs (Fangraphs and B-Ref) including a defensive adjustment in their pitching WAR.  Apologies if I confused the issue previously!

 

Fangraphs just starts from FIP, I think.  B-Ref starts with runs allowed.  (I think? )

[TheLeviathan]

 

Actually, both WARs (Fangraphs and B-Ref) including a defensive adjustment in their pitching WAR.  Apologies if I confused the issue previously!

 

Fangraphs just starts from FIP, I think.  B-Ref starts with runs allowed.  (I think? )

 

Those are questions I share and I appreciate your effort in the other thread to mine better data with a more reasonable conclusion.

 

You have to remember where some of this started - a figure (120 runs) that's absurd on the face of it.  I think the arguments centering more around 60-70 runs seems much more reasonable.  I can get behind that.

[gunnarthor]

I voted in the 40-80 range group.  I'll just say what I've said before, defense is important but right now it's impact is being overstated (and the measurements are unreliable).  

 

 

[jay]

 

 

The runs allowed gap in AL teams was 160 runs.

 

Could half of those runs (or, as some are arguing, more than half) be attributed to defense? Are some actually arguing that defense is more valuable than pitching?

 

Where is that number coming from? The gap between the Twins (777) and Athletics (572) in runs allowed was 205 last year. 

 

But, why would we use the gap in AL teams for one year when we're looking at the gap in defenses across all of MLB? 

 

The average gap in runs allowed across MLB from 2011-2014 is 302. The average gap in runs allowed for just AL teams from 2011-2014 is 242. 

 

You know better than to use a sample size of one anyway...

[jay]

Brock, I think you were looking at the AL difference in runs scored.... that's a very different matter than runs allowed when we're talking about defense.

[Brock Beauchamp]

I don't count NL teams because they play by different rules. It muddies the conversation, as NL teams score less runs because they deem it necessary to walk a clownshoe to the plate every ninth appearance.

 

Use AL-only or NL-only, I don't care... But when you're calculating offense in any way, mixing the two gives skewed results. The AL will dominate the top while the NL dominates the bottom of the list.

 

The overall run gap was me looking at the wrong column, nothing more. 200 runs versus 160 doesn't change the argument significantly, as some metrics are still putting 50% of the runs allowed gap on the defense.

[jay]

 

You have to remember where some of this started - a figure (120 runs) that's absurd on the face of it.  I think the arguments centering more around 60-70 runs seems much more reasonable.  I can get behind that.

 

The extreme measurements on each end are inherently likely to be towards the extreme bounds of the measurement error range in their respective directions.  I have no knowledge of how big those are calculated to be, but it makes sense that the real gap from one end to the other should be considered as some degree less than today's metrics show.  I just don't know how much less...

 

I'm totally comfortable with believing that range is closer to 90, 100, even a bit more... but that's me.

[TheLeviathan]

 

The extreme measurements on each end are inherently likely to be towards the extreme bounds of the measurement error range in their respective directions.  I have no knowledge of how big those are calculated to be, but it makes sense that the real gap from one end to the other should be considered as some degree less than today's metrics show.  I just don't know how much less...

 

I'm totally comfortable with believing that range is closer to 90, 100, even a bit more... but that's me.

 

I just think there are a host of other factors in RA differentials that get minimized if you think defense accounts for that much.  If we round things at 200 runs as the difference between top and bottom and use the 120 from Fangraphs - that means defense alone is accountable for 60% of that differential.  I just can't buy that.  30-40%?  Seems more reasonable to me.

[Otto von Ballpark]

 

That's my problem. I don't think anyone would argue that pitching is more valuable than defense yet the metrics some are touting suggest just that. The metrics don't work in real world application.

I don't think anyone is saying that.

 

By Fangraphs, the Royals were 141 runs above replacement on offense, 171 runs above replacement in pitching, and 68 runs above replacement (average) in fielding.

 

Subtracting 80 runs from any of those is going to hurt.  Not sure that it says anything about which is relatively more valuable.

[jay]

 

I just think there are a host of other factors in RA differentials that get minimized if you think defense accounts for that much.  If we round things at 200 runs as the difference between top and bottom and use the 120 from Fangraphs - that means defense alone is accountable for 60% of that differential.  I just can't buy that.  30-40%?  Seems more reasonable to me.

 

Let's say the plus end is something like +70 and the minus end is -70 (on average) -- for a difference of 140 runs. If we want to call the real difference 60 runs, we're saying the measurement error is extreme. I don't think it's perfect by any means, but I don't think it's quite that bad. 

 

Even if we reduce each of those extremes by 20%, we're still at a difference closer to 110.