Still a lot left to do

[Enscheff]

The simulation tool that was used had a lineup full of guys with below-average power, did it not?

I'm not sure what you're talking about, but this is the lineup optimization tool I use when I want to mess around:

http://www.baseballmusings.com/cgi-bin/LineupAnalysis.py

It only considers OBP and SLG, which is obviously different (maybe only slightly) than plugging in rates for hits, walks, HRs, etc.

 

[Enscheff]

You quoted me and said that is where you start to disagree with sabremetrics in that a 3 WAR player is better than a 2 WAR player based on the dynamics of the team and what said 2 WAR player brings to the table. The 3 WAR player is better, period.

Kemp improved the team because he played at a 1.5-2 WAR level for the Braves when we had Jeff Francoeur and others before him.

It's like you intentionally miss the intent of a comment just so you can nitpick a tiny point that isn't anywhere near the center of the topic being discussed. You consistently do the same thing with Harry's posts as well. Almost like an attempt to show how smart you are by being contrarian, but all you really show is that you completely missed the concept being conveyed.

Regardless of what I quoted, the point was essentially "WAR is WAR, regardless of how it is generated". I disagree. I then literally used "2 WAR" in all my examples to demonstrate players with equal WAR, so I never said a 3 WAR player is better than a 2 WAR player.

I think WAR made up mostly from power (Kemp's case) is more beneficial to a low-power (but average OBP) team like the Braves than just adding his WAR to the team's total would indicate.

I also think defensive WAR is misleading because defensive WAR of individual players can't be added linearly indefinitely.

 

[smootness]

I'm not sure what you're talking about, but this is the lineup optimization tool I use when I want to mess around:

http://www.baseballmusings.com/cgi-bin/LineupAnalysis.py

It only considers OBP and SLG, which is obviously different (maybe only slightly) than plugging in rates for hits, walks, HRs, etc.

I was referring to your first statement about the simulations assuming league average power. Using that tool, which is what thewupk used, it did not assume league average power because he plugged in Gardner for every spot in the lineup.

Sure, it's not a perfect tool, but SLG is a pretty good base guide for power when BA is also given. I can't imagine selecting a different # of 2B vs. HR with the same SLG/ISO is going to radically change the numbers.

 

[smootness]

It's like you intentionally miss the intent of a comment just so you can nitpick a tiny point that isn't anywhere near the center of the topic being discussed. You consistently do the same thing with Harry's posts as well. Almost like an attempt to show how smart you are by being contrarian, but all you really show is that you completely missed the concept being conveyed.

Regardless of what I quoted, the point was essentially "WAR is WAR, regardless of how it is generated". I disagree. I then literally used "2 WAR" in all my examples to demonstrate players with equal WAR, so I never said a 3 WAR player is better than a 2 WAR player.

I think WAR made up mostly from power (Kemp's case) is more beneficial to a low-power team like the Braves than just adding his WAR to the team's total would indicate.

I also think defensive WAR is misleading because defensive WAR of individual players can't be added linearly indefinitely.

So you disagreed with his assertion that a 3 WAR player is always better than a 2 WAR player (or always helps a team more, however you want to phrase it), and now you're trying to say that you didn't say a 2 WAR player might be better than/as good as a 3 WAR player depending on what they each bring to the table?

If that's not what you meant, why would you disagree with him?

 

[Enscheff]

So you disagreed with his assertion that a 3 WAR player is always better than a 2 WAR player (or always helps a team more, however you want to phrase it), and now you're trying to say that you didn't say a 2 WAR player might be better than/as good as a 3 WAR player depending on what they each bring to the table?

If that's not what you meant, why would you disagree with him?

Actually, I said, "This is where I start to disagree with sabermetrics."

Let me paste it again with the key word bolded: This is where I start to disagree with sabermetrics.

I did not say a 2 WAR player is better than a 3 WAR player. I clearly stated that there are aspects that aren't accurately captured by WAR. Things like adding power to the team with the least power in the game. Those aspects may not make upa full 1 WAR, but they start to make up. Maybe they make a 2 WAR player better than a 2.1 WAR player. Maybe better than a 2.5 WAR player. It starts to make up for it.

Pretty much my entire post explains the concept, yet you and thew cling to the tiny (and incorrect) point that I somehow said "3 > 2", when I said no such thing.

 

[thewupk]

I'm not sure what you're talking about, but this is the lineup optimization tool I use when I want to mess around:

http://www.baseballmusings.com/cgi-bin/LineupAnalysis.py

It only considers OBP and SLG, which is obviously different (maybe only slightly) than plugging in rates for hits, walks, HRs, etc.

wRC+ already tells us most of this information since that number is a % of league average offense.

MLB averaged 4.52 runs per 9 innings. With Schoop being a WRC+ 97 hitter then he would would score (4.52 x .97) 4.38 runs per 9 innings. This does not include baserunning so Schoop's 97 WRC+ and Gardeners 97 WRC+ would score the same arount of runs based on their hitting a lone.

 

[thewupk]

Actually, I said, "This is where I start to disagree with sabermetrics."

Let me paste it again with the key word bolded: This is where I start to disagree with sabermetrics.

I did not say a 2 WAR player is better than a 3 WAR player. I clearly stated that there are aspects that aren't accurately captured by WAR. Things like adding power to the team with the least power in the game. Those aspects may not make upa full 1 WAR, but they start to make up. Maybe they make a 2 WAR player better than a 2.1 WAR player. Maybe better than a 2.5 WAR player. It starts to make up for it.

Pretty much my entire post explains the concept, yet you and thew cling to the tiny (and incorrect) point that I somehow said "3 > 2", when I said no such thing.

The real point is your assertion that power is needed is dead wrong. Good hitting is needed and that can come in many ways.

 

[thewupk]

WAR player.

I think WAR made up mostly from power (Kemp's case) is more beneficial to a low-power (but average OBP) team like the Braves than just adding his WAR to the team's total would indicate.

I also think defensive WAR is misleading because defensive WAR of individual players can't be added linearly indefinitely.

1) You can think that regarding offense but you have no proof for it. Whereas there is plenty of proof (aka math) to support my argument that Matt Kemp's 120 WRC+ was going to help the Braves tremendously whether he smashed a lot of homeruns or got on base at an insane clip.

2) The defense part of it is another argument entirely. And if you want to argue the merits of a 2 WAR offense first player against a 3 WAR defensive first player and who really helps the team more then that is fine. I understand both sides of that argument. But from a strictly offensive point of view there is no argument.

 

[Horsehide Harry]

I just don't think the high OBP low power offense translates to the real world (no matter what the math indicates or doesn't). Teams aren't building that way (even the Giants traded for Pence when they felt they were too short on power). And, as many often point out when they have nothing better to say, they've got way more resources than we do. I would expect that a team would try to do it out of desperation alone (and they may have, falling short and we just don't know).

I think the high OBP/low power club would be affected over a season as pitchers adjusted to the fact that they had no power, which would potentially impact their OBP. They would see a lot more challenge pitches which would diminish their ability to work the count (they would either swing and miss, foul off, get a hit or take a strike). A "mistake" to a guy like that results in a player on base (maybe he drives in a run, maybe not depending upon the previous situation on the bases) whereas a mistake to a HR power hitter where he hits a HR results in at least one run every time.

I think if you had a team of 3 WAR offensive players who were HR hitters and a team of 3 WAR offensive players who were not, you would find at the end of the season that the HR hitting team easily scored more runs (over 162), this is where you base the player WAR valuation for each at 3 WAR at the beginning of the season. I would find it hard to believe in any mathematical model that said differently because of the humanity involved in the game.

 

[Enscheff]

The real point is your assertion that power is needed is dead wrong. Good hitting is needed and that can come in many ways.

Again, unfortunately, you can't seem to grasp the nuance that values like wRC and oWAR are calulated against the context of the entire league, while I am talking about building a team.

So let's do an experiment centered around the stat you quoted, wRC. I picked 2 players with identical wRC values of 101, but they each achieved those values in different ways in 2016. Trumbo had a .316 OBP and a .533 SLG, while Villar had a .369 OBP and a .457 SLG.

According to you, these are identical players offensively. Afterall, they have identical wRC and very close wOBA values.

So I plugged 9 Villars into the lineup optimizer and got 5.715 runs per game. All optimized lineups were obviously the same.

Then I replaced a Villar with a Trumbo. The optimizer predictably stuck him in the #4 spot. The result: 5.741 runs per game, or 4 more runs per season.

I then continued to replace Villar's with Trumbo's, and guess what the optimal balance was? OK I'll tell you, 4 Trumbo's and 5 Villar's produced 5.771 runs per game, or 9 more runs per season than 9 Villar's.

That equates to 1-2 more wins simply from lineup balance.

So please tell me with your impressive grasp of data analysis, how replacing 4 batters with 4 other batters of exactly equal "value" managed to score a team 9 more runs per season in this simulation?

 

[Enscheff]

I just don't think the high OBP low power offense translates to the real world (no matter what the math indicates or doesn't). Teams aren't building that way (even the Giants traded for Pence when they felt they were too short on power). And, as many often point out when they have nothing better to say, they've got way more resources than we do. I would expect that a team would try to do it out of desperation alone (and they may have, falling short and we just don't know).

I think the high OBP/low power club would be affected over a season as pitchers adjusted to the fact that they had no power, which would potentially impact their OBP. They would see a lot more challenge pitches which would diminish their ability to work the count (they would either swing and miss, foul off, get a hit or take a strike). A "mistake" to a guy like that results in a player on base (maybe he drives in a run, maybe not depending upon the previous situation on the bases) whereas a mistake to a HR power hitter where he hits a HR results in at least one run every time.

I think if you had a team of 3 WAR offensive players who were HR hitters and a team of 3 WAR offensive players who were not, you would find at the end of the season that the HR hitting team easily scored more runs (over 162), this is where you base the player WAR valuation for each at 3 WAR at the beginning of the season. I would find it hard to believe in any mathematical model that said differently because of the humanity involved in the game.

You aren't wrong, see above.

 

[JohnAdcox]

Again, unfortunately, you can't seem to grasp the nuance that values like wRC and oWAR are calulated against the context of the entire league, while I am talking about building a team.

So let's do an experiment centered around the stat you quoted, wRC. I picked 2 players with identical wRC values of 101, but they each achieved those values in different ways in 2016. Trumbo had a .316 OBP and a .533 SLG, while Villar had a .369 OBP and a .457 SLG.

According to you, these are identical players offensively. Afterall, they have identical wRC and very close wOBA values.

So I plugged 9 Villars into the lineup optimizer and got 5.715 runs per game. All optimized lineups were obviously the same.

Then I replaced a Villar with a Trumbo. The optimizer predictably stuck him in the #4 spot. The result: 5.741 runs per game, or 4 more runs per season.

I then continued to replace Villar's with Trumbo's, and guess what the optimal balance was? OK I'll tell you, 4 Trumbo's and 5 Villar's produced 5.771 runs per game, or 9 more runs per season than 9 Villar's.

That equates to 1-2 more wins simply from lineup balance.

So please tell me with your impressive grasp of data analysis, how replacing 4 batters with 4 other batters of exactly equal "value" managed to score a team 9 more runs per season in this simulation?

You know, I think the reason I hang out here so much is that listening to you guys argue makes me feel just heroically well-adjusted.

 

[thewupk]

Again, unfortunately, you can't seem to grasp the nuance that values like wRC and oWAR are calulated against the context of the entire league, while I am talking about building a team.

So let's do an experiment centered around the stat you quoted, wRC. I picked 2 players with identical wRC values of 101, but they each achieved those values in different ways in 2016. Trumbo had a .316 OBP and a .533 SLG, while Villar had a .369 OBP and a .457 SLG.

According to you, these are identical players offensively. Afterall, they have identical wRC and very close wOBA values.

So I plugged 9 Villars into the lineup optimizer and got 5.715 runs per game. All optimized lineups were obviously the same.

Then I replaced a Villar with a Trumbo. The optimizer predictably stuck him in the #4 spot. The result: 5.741 runs per game, or 4 more runs per season.

I then continued to replace Villar's with Trumbo's, and guess what the optimal balance was? OK I'll tell you, 4 Trumbo's and 5 Villar's produced 5.771 runs per game, or 9 more runs per season than 9 Villar's.

That equates to 1-2 more wins simply from lineup balance.

So please tell me with your impressive grasp of data analysis, how replacing 4 batters with 4 other batters of exactly equal "value" managed to score a team 9 more runs per season in this simulation?

First off it would be nice if you actually used the right statistic. wRC and wRC+ are not the same. They both indeed had a 101 wRC as they created the same amount of runs. Villlar had a lot more PA to accomplish this however. Their wRC+ is different. Trumbo at 123 and Villar at 118. Trumbo was the better hitter so of course a team full of Trumbo's is going to be better.

 

[thewupk]

I think if you had a team of 3 WAR offensive players who were HR hitters and a team of 3 WAR offensive players who were not, you would find at the end of the season that the HR hitting team easily scored more runs (over 162), this is where you base the player WAR valuation for each at 3 WAR at the beginning of the season. I would find it hard to believe in any mathematical model that said differently because of the humanity involved in the game.

I would say that you don't know how offensive WAR works if you think that is the case. It's based off wRC and WOBA which using linear weights. Homeruns, walks, singles, etc are all assigned a point value of what they are worth to runs scored in the current years scoring environment. If two players/teams produce the same wRC+ (park adjusted number) they will have produced the same amount of runs regardless of how it is achieved (homeruns, walks, etc).

If those two teams played in the same park/league then they would score the same amount of runs.

 

[thewupk]

Another thing to note is that getting raw slg/ops numbers aren't going to be 100% accurate with this tool when trying to use it with 2 players that have the same WRC+ simply because one is park/league adjusted and the other is not.

But Belt and his 394/474 combo outscores Cano and his 350/533 combo 6.295 to 6.119. But again that's taking them at face value. Both play in pitchers parks but given neutral environments they would be the same.

 

[Horsehide Harry]

I would say that you don't know how offensive WAR works if you think that is the case. It's based off wRC and WOBA which using linear weights. Homeruns, walks, singles, etc are all assigned a point value of what they are worth to runs scored in the current years scoring environment. If two players/teams produce the same wRC+ (park adjusted number) they will have produced the same amount of runs regardless of how it is achieved (homeruns, walks, etc).

If those two teams played in the same park/league then they would score the same amount of runs.

But that's the thing it's reflective not predictive. You calculate a players WAR by what they did not what they WILL do. Now, granted, it is often a good predictor for individual performances. But not always because humanity, chance, environment all play a part.

Take Ian Desmond. His oWAR in 2014 was 3.8. His oWAR in 2015 was 1.8. His oWAR after 2016 was 3.1 but he had a position change going on at the same time.

If you were building a team for 2016 there is no way you would know what oWAR you would get out of Desmond. As a GM you could look at his roughly 4 oWAR he had each year for 2012-2014 seasons and say I hope he does that. However, there is no way to truly know if 2015 was the start of the new norm or an aberration.

 

[Enscheff]

First off it would be nice if you actually used the right statistic. wRC and wRC+ are not the same. They both indeed had a 101 wRC as they created the same amount of runs. Villlar had a lot more PA to accomplish this however. Their wRC+ is different. Trumbo at 123 and Villar at 118. Trumbo was the better hitter so of course a team full of Trumbo's is going to be better.

So how do you explain a lineup of 9 Trumbo's doing worse than a lineup of 9 Villar's (5.715 to 5.490)? A team full of Trumbo's also did MUCH worse than the optimized mix of Villar's and Trumbo's. Weird right? Almost like being balanced is better?

I knew you were going to bring up wRC vs wRC+, so I also did the same exercise with Jup (.310 OBP, .465 SLG) and Pagan (.331 OBP, .418 SLG), both with a wRC+ of 105. Here are the results:

9 Pagan's: 4.662
Optimized mix (6 Jup's and 3 Pagan's): 4.783
9 JUp's: 4.734

So again, the balanced lineups are better than either extreme. The best mix is worth about 19 more runs over the course of a full season compared to the all-Pagan lineup.

How could that be possible if they both have 105 wRC+? You insist lineup balance doesn't matter, so why are the balanced lineups scoring more runs in these simulations?

It's adorable that you are taking the time to learn about advanced stats though. A definite step in the right direction. Now you just have to figure out how to think critically and apply your new knowledge to draw logical conclusions.

 

[thewupk]

But that's the thing it's reflective not predictive. You calculate a players WAR by what they did not what they WILL do. Now, granted, it is often a good predictor for individual performances. But not always because humanity, chance, environment all play a part.

Take Ian Desmond. His oWAR in 2014 was 3.8. His oWAR in 2015 was 1.8. His oWAR after 2016 was 3.1 but he had a position change going on at the same time.

If you were building a team for 2016 there is no way you would know what oWAR you would get out of Desmond. As a GM you could look at his roughly 4 oWAR he had each year for 2012-2014 seasons and say I hope he does that. However, there is no way to truly know if 2015 was the start of the new norm or an aberration.

That is obvious. However we are taking about actual results which is what WAR represents. If a hitter has a high OBP/low SLG and produces 2 oWAR he would be equal to a guy with low OBP/high SLG regardless.

 

[thewupk]

So how do you explain a lineup of 9 Trumbo's doing worse than a lineup of 9 Villar's (5.715 to 5.490)? A team full of Trumbo's also did MUCH worse than the optimized mix of Villar's and Trumbo's. Weird right? Almost like being balanced is better?

I knew you were going to bring up wRC vs wRC+, so I also did the same exercise with Jup (.310 OBP, .465 SLG) and Pagan (.331 OBP, .418 SLG), both with a wRC+ of 105. Here are the results:

9 Pagan's: 4.662
Optimized mix (6 Jup's and 3 Pagan's): 4.783
9 JUp's: 4.734

So again, the balanced lineups are better than either extreme. The best mix is worth about 19 more runs over the course of a full season compared to the all-Pagan lineup.

How could that be possible if they both have 105 wRC+? You insist lineup balance doesn't matter, so why are the balanced lineups scoring more runs in these simulations?

It's adorable that you are taking the time to learn about advanced stats though. A definite step in the right direction. Now you just have to figure out how to think critically and apply your new knowledge to draw logical conclusions.

If you knew I was going to bring up wRC+ then why not use it first? Just using wRC shows me you have no idea what you are talking about.

And I guess you still don't understand about park factors. Pagan played in an extreme pitchers park and Jup did not. Do you really think using raw totals are going to work that way? It's like taking Arenado's raw OPS and comparing it someone else with a similar WRC+. Just isn't going to work.

 

[Enscheff]

If you knew I was going to bring up wRC+ then why not use it first? Just using wRC shows me you have no idea what you are talking about.

And I guess you still don't understand about park factors. Pagan played in an extreme pitchers park and Jup did not. Do you really think using raw totals are going to work that way? It's like taking Arenado's raw OPS and comparing it someone else with a similar WRC+. Just isn't going to work.

OK, so let's find two guys on the same team with identical wRC+ values and different OBP/SLG splits (not easy to find). Let's look at Odor (.296 OBP, .502 SLG) and Desmond (.335 OBP, .446 SLG), both with a 106 wRC+. Both play for the Texas Rangers.

9 Desmond's: 5.001
4 Odor's and 5 Desmond's: 5.042

That's almost 7 runs more per season just by having a balanced lineup.

So how is that possible if those two guys have identical wRC+ values, and played all their games against the same competition, and in the same ballparks? Do I need to sift around for another pair of players to further illustrate this point?