Stats. Giggity!

[mimontero88]

When you have a save situation in extras, your win probability is less than 100%.  1 run drops to 50% (maybe a touch more if the run scoring event is anything other than a homer).  When you have a tie, that one run drops you from 50% to 0%.  Like I said, I'd really like to see the leverage charts to understand how this works.

The flaw there is the assumption that one run drops you to 50%, which only occurs in the situation where you have scored exactly one run in the top of the inning.  While this is certainly the most common outcome, scoring two or three runs are also possible outcomes, though admittedly they also lower the leverage of the situation.  However, you also have to factor in that allowing 0 runs in an extra inning tie game actually raises your win probability to 50%.  Before you put up that 0, your win probability is actually less than 50% since allowing any run is an immediate loss and you don't have the opportunity to try to score back.  You're only truly a 50% at the beginning of an inning.  At the bottom, you've already lost your opportunity to score.  In a save situation, allowing zero runs increases your win probability to 100% (though not from 50%).  I'm not sure how that all breaks down in leverage charts but it's hard for me to imagine that over a long sample size it makes more sense to use a closer in the situation where your win probability is less than 50% just to get you to 50%.

[JCA-CrystalCity]

been a while since a post here so here's one I'd love to see the numbers answer for.  Mimontero and I have been going back and forth about leverage and the relief decision last night.  Nats entered the bottom of the 9th on the road tied.  Dusty went with Romero rather than Doolittle, saving him for a save situation.  I asserted that there's no higher leverage situation than a tie on the road when the next run "walks off" so the best pitcher should be in.  Win probability goes from near 50% to 0. Mim suggested that a save situation in extras is still, most likely, a higher leverage situation.  He makes the point that, with the Nats unlikely to seize a lead >3 runs in extras, any lead situation is one where the win probability would go from some number > 50% down to < 50% (or 0) so it could be higher leverage, especially if the blown save results in a loss.  Anyone have the answer or a good discussion?

[JCA-CrystalCity]

The flaw there is the assumption that one run drops you to 50%, which only occurs in the situation where you have scored exactly one run in the top of the inning.  While this is certainly the most common outcome, scoring two or three runs are also possible outcomes, though admittedly they also lower the leverage of the situation.  However, you also have to factor in that allowing 0 runs in an extra inning tie game actually raises your win probability to 50%.  Before you put up that 0, your win probability is actually less than 50% since allowing any run is an immediate loss and you don't have the opportunity to try to score back.  You're only truly a 50% at the beginning of an inning.  At the bottom, you've already lost your opportunity to score.  In a save situation, allowing zero runs increases your win probability to 100% (though not from 50%).  I'm not sure how that all breaks down in leverage charts but it's hard for me to imagine that over a long sample size it makes more sense to use a closer for lower win probabilities.

not sure I'm on board but I've bumped the stats giggity thread to continue the discussion.   Does The Book discuss this?

[mimontero88]

not sure I'm on board but I've bumped the stats giggity thread to continue the discussion.   Does The Book discuss this?

I'm not sure if it does.  I would almost guess not because, even as a math major, I'm a bit at a loss as to how the run the calculation on this with so many different variables to consider.

[mimontero88]

http://www.tangotiger.net/welist.html

So this spreadsheet can only compare win probability added by preserving a tie vs. saving exactly a 1 run lead.  But I did the math quickly and found that saving a one-run game gives an added win probability of 15.9% whereas preserving the tie gives an added win probability of 13.4%.  So assuming you only score one run in the top of an inning, the save situation is the higher leverage situation.  In innings where you score runs, scoring exactly one run is only the result about 54.9% of the time according to a link I'll leave down below to avoid confusion.  So that muddies the waters significantly.  However, what this does tell us is that, in the most common run scoring outcome (exactly one run scored) it is better to save the closer for the save situation.

https://gregstoll.dyndns.org/~gregstoll/baseball/runsperinning.html

EDIT:  Ouch.  One issue with my methodology that I discovered is that I assumed that a top of the inning save would give the same win probability as a bottom of the inning save.  This isn't strictly true because if the save is blown in the top of the 9th, the game can be won with no chance for the other team to bat again in the bottom of the 9th.  So the 13.4% number is inflated by that assumption.  It should be lower.  Makes a stronger case for saving the closer.  To correct this number I would need data I can't find yet and would need to separate between situations where the blown save results in a tie and those where it results in a deficit.

EDIT 2:  I'm an idiot.  I can use the bottom 9 down by 1 number for the save situation.  It's not the 13.4% number that is inflated.  It's the 15.9% number that's deflated.  Thus, a successful save in the bottom of an inning of a one run game adds 19.4% to your win probability vs. 13.4% for preserving the tie.  The 1 run save is absolutely the higher leverage situation.

[varoadking]

Over the course of a long season, it's a higher leverage situation yes.  It comes down to win probability.  You're far more likely to win a game you've gotten to a save situation than you are to win a tie game in extra innings.  So you save your closer for the situation in which your win probability is much better instead of the situation where it is essentially 50/50.

You have to get to a win situation first. 

[varoadking]

http://www.tangotiger.net/welist.html

So this spreadsheet can only compare win probability added by preserving a tie vs. saving exactly a 1 run lead.  But I did the math quickly and found that saving a one-run game gives an added win probability of 15.9% whereas preserving the tie gives an added win probability of 13.4%.  So assuming you only score one run in the top of an inning, the save situation is the higher leverage situation.  In innings where you score runs, scoring exactly one run is only the result about 54.9% of the time according to a link I'll leave down below to avoid confusion.  So that muddies the waters significantly.  However, what this does tell us is that, in the most common run scoring outcome (exactly one run scored) it is better to save the closer for the save situation.

https://gregstoll.dyndns.org/~gregstoll/baseball/runsperinning.html

EDIT:  Ouch.  One issue with my methodology that I discovered is that I assumed that a top of the inning save would give the same win probability as a bottom of the inning save.  This isn't strictly true because if the save is blown in the top of the 9th, the game can be won with no chance for the other team to bat again in the bottom of the 9th.  So the 13.4% number is inflated by that assumption.  It should be lower.  Makes a stronger case for saving the closer.  To correct this number I would need data I can't find yet and would need to separate between situations where the blown save results in a tie and those where it results in a deficit.

EDIT 2:  I'm an idiot.  I can use the bottom 9 down by 1 number for the save situation.  It's not the 13.4% number that is inflated.  It's the 15.9% number that's deflated.  Thus, a successful save in the bottom of an inning of a one run game adds 19.4% to your win probability vs. 13.4% for preserving the tie.  The 1 run save is absolutely the higher leverage situation.

 

[JCA-CrystalCity]

Makes some sense when I stop to think about it.  Preserving the tie more or less brings your team back to a 50/50 shot at winning.  The WPA is basically the probability of a team scoring at least a run in an inning (in this case, a home team).  On the other hand, saving a 1 run lead more or less gives you the WPA of going from a 1 run lead to a certain win.  I would imagine  most of the other save situations add less WPA than a 1 run save.  So, it would be something like 54.9% x 19.4% + (%2 run lead) x (WPA  for 2 run save) + ....   Compare that to 13.4%.   

Really could use Mathguy to weigh in. 

[mimontero88]

Makes some sense when I stop to think about it.  Preserving the tie more or less brings your team back to a 50/50 shot at winning.  The WPA is basically the probability of a team scoring at least a run in an inning (in this case, a home team).  On the other hand, saving a 1 run lead more or less gives you the WPA of going from a 1 run lead to a certain win.  I would imagine  most of the other save situations add less WPA than a 1 run save.  So, it would be something like 54.9% x 19.4% + (%2 run lead) x (WPA  for 2 run save) + ....   Compare that to 13.4%.   

Really could use Mathguy to weigh in. 

You're exactly right on the calculation.  I just haven't been able to find numbers for those scenarios.

[bluestreak]

The Book covers in when it uses leverage index. Which I think is appropriate here.

http://www.insidethebook.com/li.shtml

If I’m reading this right up one in 9th is higher leverage than tied. But I could be reading wrong.

Nate Silver cover this in this article where he introduces his “Goose Egg” stat. The article is quite good I think.

https://fivethirtyeight.com/features/goose-egg-new-save-stat-relief-pitchers/

[mitlen]

I'll bet ya a donut, none of this is part of Dusty's thought process.

[JCA-CrystalCity]

Up 1 in the 9th will be the highest leverage, so the best case for holding off the closer in a tie in the 9th is you anticipate a 1 run lead in extras.  As Mim pointed out in the game thread (I may have moved it here), Dusty could have said, "I just don't see us scoring multiple runs" and reasoned that running Enny out there in the 9th and holding Doolittle off the 1 run lead was more likely to pay off than running Doolittle out there in a tie in the 9th and using Enny later.  But, if we score 2 runs or more, then the tie in the 9th is higher leverage than the extra inning lead.

Of course, the unstated big if in all this is if we get to extras and if we grab a lead of any kind.  Ready, aim, fire only hits the target if you get to fire it.  You had the situation in the 9th where you could have used Doolittle, but didn't get a chance to use him at all.  Maybe, as DCPatti put it, it was season managing and gauging just how good the non-Madson, non-Doolittle portion of the pen is.

[mimontero88]

Up 1 in the 9th will be the highest leverage, so the best case for holding off the closer in a tie in the 9th is you anticipate a 1 run lead in extras.  As Mim pointed out in the game thread (I may have moved it here), Dusty could have said, "I just don't see us scoring multiple runs" and reasoned that running Enny out there in the 9th and holding Doolittle off the 1 run lead was more likely to pay off than running Doolittle out there in a tie in the 9th and using Enny later.  But, if we score 2 runs or more, then the tie in the 9th is higher leverage than the extra inning lead.

Of course, the unstated big if in all this is if we get to extras and if we grab a lead of any kind.  Ready, aim, fire only hits the target if you get to fire it.  You had the situation in the 9th where you could have used Doolittle, but didn't get a chance to use him at all.  Maybe, as DCPatti put it, it was season managing and gauging just how good the non-Madson, non-Doolittle portion of the pen is.

Just doing some mental estimations, I think the answer is probably that saving the closer is the right move.  But the difference between the two scenarios seems like it's so marginal that it probably wouldn't even make a full game's worth of difference in a 162 game season.  I am shocked at how much leverage you lose with a two run lead compared to a one run lead though.  That's crazy.

EDIT:  The big if you're describing is actually fairly irrelevant.  We're talking about the best way to win the game and every single scenario where we win involves getting to the bottom of an inning with a lead so we know someone has to pitch that inning.  The question is who pitches when we're tied versus who pitches when we're ahead.

[imref]

https://fivethirtyeight.com/features/baseball-is-overrun-by-adam-dunns/

[mitlen]

https://fivethirtyeight.com/features/baseball-is-overrun-by-adam-dunns/

Thanks imref.     I also read this article from the posted site:

https://fivethirtyeight.com/features/hall-of-fame-careers-that-cooperstown-never-gave-the-time-of-day/?src=obsidebar=sb_1

[imref]

Thanks imref.     I also read this article from the posted site:

https://fivethirtyeight.com/features/hall-of-fame-careers-that-cooperstown-never-gave-the-time-of-day/?src=obsidebar=sb_1

interesting, I never would have considered Lofton a HoF'r.

[mitlen]

interesting, I never would have considered Lofton a HoF'r.

That's what got me readin' the piece.

[PowerBoater69]

From Fangraphs:Double play started, double play turned, double play finished, but what is DP? Can't be a total because it is sometimes lower than DPF.

[imref]

https://blogs.fangraphs.com/instagraphs/job-posting-washington-nationals-senior-analyst-baseball-research-development/

[Natsinpwc]

https://blogs.fangraphs.com/instagraphs/job-posting-washington-nationals-senior-analyst-baseball-research-development/

Dear Mike.

Attached is my resume.  I am retired but worked 35 years in the federal government where most managers were allergic to data.  I have honed my skills on WNFF. My specialty is latching onto small sample size performance and extrapolating that logarithmically. I like to call “game”as soon as we fall behind but to offset that can say “it’s early” when the team gets off to a bad start.  I am a team player and will treat everyone with equal disdain. I look forward to the game day buffets.

Sincerely,
Natsinpwc.

[imref]

Dear Mike.

Attached is my resume.  I am retired but worked 35 years in the federal government where most managers were allergic to data.  I have honed my skills on WNFF. My specialty is latching onto small sample size performance and extrapolating that logarithmically. I like to call “game”as soon as we fall behind but to offset that can say “it’s early” when the team gets off to a bad start.  I am a team player and will treat everyone with equal disdain. I look forward to the game day buffets.

Sincerely,
Natsinpwc.

hired!

[welch]

"Front office staff are expected to attend at least half of the Nationals home games. Meals are provided to staff during the games."

That's a perk, a benefit, not a harsh requirement.

[JCA-CrystalCity]

"Front office staff are expected to attend at least half of the Nationals home games. Meals are provided to staff during the games."

That's a perk, a benefit, not a harsh requirement.

when they say "meals provided," they don't have a requirement that you keep the meal down after the 4th baserunning mistake of the game?  That would make the job a lot more difficult.

[English Natsie]

A sad irony that this should come up after the news about MathGuy...

[imref]

Fangraphs Jay Jaffe:

"Retroactive to the 2016 season, we have swapped out the Range component of UZR for the Statcast metric Fielding Runs Prevented, which is Outs Above Average (OAA) converted to runs above average."

Talknats notes this change now gives Carter Kieboom the worst WAR in Nationals' history.