2016 Steamer Predictions (Latest: ZiPS, post 25)

[reillymcshane]

For those who like the predictions models, the 2016 Steamer results are in (and a lot of it ain't pretty):

 

http://www.fangraphs.com/projections.aspx?pos=all&stats=pit&type=steamer600&team=23&lg=all&players=0&sort=9,d

 

Interesting look at the starting rotation:

 

Jimmy Nelson - 7.87 K/9, 3.16 BB/9, 1.34 WHIP, 4.04 FIP, 4.26 ERA

Zach Davies - 7.07 K/9, 3.35 BB/9, 1.40 WHIP, 4.22 FIP, 4.28 ERA

Wily Peralta - 6.38 K/9, 2.94 BB/9, 1.41 WHIP, 4.35 FIP, 4.47 ERA

Matt Garza - 6.94 K/9, 2.91 BB/9, 1.38 WHIP, 4.44 FIP, 4.45 ERA

Taylor Jungmann - 7.69 K/9, 3.90 BB/9, 1.44 WHIP, 4.50 FIP, 4.66 ERA

 

I honestly don't know how accurate Steamer has been in the past, but it's fun to see some of these things.

[MadScientist]

I'm not sure I'd call those predictions 'fun'.

 

Unless I'm misreading something, they are pretty much predicting that the Brewers will have the worst starting staff in the majors. I didn't check every team, but they had the Brewers worse than everyone I did check, including the Reds, Phillies and even the Rockies.

 

Maybe they think the Brewers are fielding a AAA team and therefore are playing in Sky Sox stadium.

[reillymcshane]

I'm not sure I'd call those predictions 'fun'.

 

Unless I'm misreading something, they are pretty much predicting that the Brewers will have the worst starting staff in the majors. I didn't check every team, but they had the Brewers worse than everyone I did check, including the Reds, Phillies and even the Rockies.

 

Maybe they think the Brewers are fielding a AAA team and therefore are playing in Sky Sox stadium.

The hitters aren't much better. But think about it, many of our 'established' players are either aging, had a mediocre year(s) or are not that good. Systems like this usually project minor leaguers pretty harshly.

[pitchleague]

How do these "predictions" even work, anyway? Do they use computer simulations of 100 seasons of 2016 and average them out? Is it all done on a spreadsheet, averaging out players' 3-year marks? Are they completely 100% statistical-based or is there any human element to them?

 

I'm pretty sure that a good lot of us could come out with our own predictions and come relatively close on a good bunch of them.

 

(I'm honestly asking those questions, by the way. I'm not trying to be a jerk or anything.)

[TURBO]

Coulter listed as our #2 catcher? Anyone tell them that he isn't even a catcher anymore?

[trwi7]

More interested in seeing Cleveland's projections.

[logan82]

League average for the NL last year was 3.91ERA. None of those WHIPs look good.

[DuWayne Steurer]

So in essence, all of our starters project to be nearly identically sucky.

 

I don't see why they'd expect to see Jimmy Nelson to take a step backwards going into his age 27 season.

[pacopete4]

I don't care what the predictions say, the Brewers will not have the worst pitching staff in the MLB next season.

[reillymcshane]

How do these "predictions" even work, anyway? Do they use computer simulations of 100 seasons of 2016 and average them out? Is it all done on a spreadsheet, averaging out players' 3-year marks? Are they completely 100% statistical-based or is there any human element to them?

 

I'm pretty sure that a good lot of us could come out with our own predictions and come relatively close on a good bunch of them.

 

(I'm honestly asking those questions, by the way. I'm not trying to be a jerk or anything.)

I don't know specifically how these things work, but I believe Steamer is 100% statistical-based. Usually you have all the standard elements - past statistics, age, park adjusted factors, etc.

 

If we look at our pitchers with a track record (Nelson, Garza, Peralta) you can make a shot at how results are reached. I'd guess that they take a certain time frame - perhaps three years - weighing more recent years more heavily. They then bring certain numbers to the mean (BABIP for instance) to help determine the predicted results. Here's a look at the three players:

 

Jimmy Nelson

Predicted ERA: 4.26

Last season's ERA: 4.11

ERA the last three years: 4.21

Predicted WHIP: 1.34

Last season's WHIP: 1.29

WHIP the last three years: 1.31

BABIP last year: .285

BABIP career: .296

Predicted BABIP for 2016: .297

 

So for Nelson (as with the other guys), there may be a look at the last three years (there may be some minor league stats in the equation as well). But if you look at the predictions, and the results, it's sort of in that area of what he's actually done. The regression may be a result of Steamer predicting his BABIP will be a bit worse (.297 instead of .285 - I believe around .300 is average in the league). How they incorporate aging in is a mystery. Maybe not at all.

 

Wily Peralta

Predicted ERA: 4.47

Last season's ERA: 4.72

ERA the last three years: 4.11

Predicted WHIP: 1.41 WHIP

WHIP last year: 1.54

WHIP the last three years: 1.40

 

Why he is predicted to be worse than has three year run is a good question. My guess would be that more recent years are weighted more heavily. Last season he had a 4.72 ERA, so the predicted ERA of 4.47 sort of rests between his three year average and last season's results. In the end, the 4.47 ERA is better than last year's performance. Steamer predicts better strikeout and walk rates, plus a lower BABIP - leading to the better (but still not very good) performance. Just a guess.

 

Matt Garza

Predicted ERA: 4.45

Last season's ERA: 5.63

ERA the last three years: 4.33

Predicted WHIP: 1.38

Last season's WHIP: 1.57

WHIP the last three years: 1.32

 

Like with Peralta, numbers are skewing toward a 'three year average', but with some extra weight given to 2016.

 

Using these three guys as case studies, you can perhaps make some educated guesses on how their system works. But it is only three guys, so maybe I'm off base.

[DuWayne Steurer]

Interestingly, Steamer predicts Arcia to come out to a .683 OPS. Which (unsurprisingly) is a higher OPS (by a good margin) than they predict for Segura.

 

We have little shot at competing, so I see no reason to start Arcia's service clock at the start of the season, but honestly, I'm not surprised that the first round of comps show Arcia this high already.

[MrTPlush]

Clancy is going to be very sad when he sees Brent Suter's huge ERA and BB% spikes.

[jerichoholicninja]

Clancy is going to be very sad when he sees Brent Suter's huge ERA and BB% spikes.

 

And Braun's zero projected innings at 3B.

[DuWayne Steurer]

Clancy is going to be very sad when he sees Brent Suter's huge ERA and BB% spikes.

 

And Braun's zero projected innings at 3B.

 

Nevermind the fact that Braun said he would play anywhere *except* third base.

[umphrey]

Steamer is pretty good. Here are a couple noteworthy things about it if you're unfamiliar:

 

1. They predict FIP, and they pretty much assume FIP = ERA. FIP is mostly driven by strikeouts, walks, and HR rate. There are players that consistently overperform or underperform their FIP, and Steamer does not pick up on it. It's a big debate whether or not Jungmann is one of those guys, but all of the analytics are predicting he's not and that last year was a fluke.

 

2. It takes about 2 years of MLB time to get any certainty in Steamer. If Davies and Garza have 3.00 FIP/ERA at the all star break next year, Davies projection will improve drastically but Garza will not move hardly at all. Steamer doesn't know what to do with a guy like Davies but it knows Garza very well at this point.

 

3. To answer the "how they work" question, they look at fundamental stats as opposed to "luck" based ones and use those to build a predictive model, then adjust for age and level shift. For example, Davies averaged about 8 K/9 in the MiLB, so they think that will drop to about 7 K/9 with the increased difficulty. That goes in to a data model along with other numbers to predict FIP and ERA. Garza is in his 30s so they pretty much predict based on his last 3 years with a tilt towards the more recent data, and knock off a little bit for getting older.

 

4. Steamer goes bananas for strikeouts, and we don't have many high strikeout guys. Example: They are predicting a career year for Strasburg because he's been a 10 K/9 guy for a few years. On the other hand you have Edinson Volquez who started the WS tonight, good ERA but he hasn't reached 7 K/9 in the last 2 years so Steamer says he's pretty much Matt Garza. I don't disagree with this, Volquez is not good and I think Steamer is correctly calling him fluky.

 

It's important to remember that these predictions are baselines. 1-2 guys could improve their ERA by a full run through luck or mechanical improvements, but Steamer doesn't include luck or mechanical improvements. Conversely, 1-2 guys will probably get hurt or be terrible, but we can easily replace with a guy who will throw a 4.5 ERA so a lot of these predictions are kind of like the worst possible scenario in our case.

[umphrey]

I checked, we are probably projected at 2nd worst behind the Rockies. Twins are close and the Reds might only be above us because they are extrapolating off 95 fluky innings from Raisail Iglesius and a John Lamb prediction entirely based off K rate that won't last.

[MrTPlush]

Clancy is going to be very sad when he sees Brent Suter's huge ERA and BB% spikes.

 

And Braun's zero projected innings at 3B.

 

Nevermind the fact that Braun said he would play anywhere *except* third base.

 

So your saying SS is still a possibility? Can't maximize his value any more than that!

[DuWayne Steurer]

 

So your saying SS is still a possibility? Can't maximize his value any more than that!

 

I think (I hope......) when Braun said he would play *anywhere* that it was kind of to be taken tongue in cheek. I think he's aware that a move to 1st base is possible in the near future. In this particular context however, I think "anywhere" is exactly three positions. LF, RF, or 1B.

[1992casey]

No matter what Ryan thinks or what the Brewers think or what Brewerfan members think, Steamer only lists Braun as a right fielder. I don't think that counts as a projection. It looks like a way for the reader to view fewer players at once.

 

It's interesting that all players except for catchers are given stats for 600 plate appearances. Catcher stats are for 450. Other systems will attempt to project plate appearances.

[reillymcshane]

No matter what Ryan thinks or what the Brewers think or what Brewerfan members think, Steamer only lists Braun as a right fielder. I don't think that counts as a projection. It looks like a way for the reader to view fewer players at once.

 

It's interesting that all players except for catchers are given stats for 600 plate appearances. Catcher stats are for 450. Other systems will attempt to project plate appearances.

The top link goes to the Steamer 600 section, which (as you would expect) gives everyone 600 plate appearances. Click on the 'steamer' link and it will give you playing time predictions, so the minor league guys have little or no ABs.

 

Here's the direct link:

 

http://www.fangraphs.com/projections.aspx?pos=all&stats=bat&type=steamer&team=23&lg=all&players=0

[reillymcshane]

Steamer is pretty good. Here are a couple noteworthy things about it if you're unfamiliar:

 

1. They predict FIP, and they pretty much assume FIP = ERA. FIP is mostly driven by strikeouts, walks, and HR rate. There are players that consistently overperform or underperform their FIP, and Steamer does not pick up on it. It's a big debate whether or not Jungmann is one of those guys, but all of the analytics are predicting he's not and that last year was a fluke.

 

2. It takes about 2 years of MLB time to get any certainty in Steamer. If Davies and Garza have 3.00 FIP/ERA at the all star break next year, Davies projection will improve drastically but Garza will not move hardly at all. Steamer doesn't know what to do with a guy like Davies but it knows Garza very well at this point.

 

3. To answer the "how they work" question, they look at fundamental stats as opposed to "luck" based ones and use those to build a predictive model, then adjust for age and level shift. For example, Davies averaged about 8 K/9 in the MiLB, so they think that will drop to about 7 K/9 with the increased difficulty. That goes in to a data model along with other numbers to predict FIP and ERA. Garza is in his 30s so they pretty much predict based on his last 3 years with a tilt towards the more recent data, and knock off a little bit for getting older.

 

4. Steamer goes bananas for strikeouts, and we don't have many high strikeout guys. Example: They are predicting a career year for Strasburg because he's been a 10 K/9 guy for a few years. On the other hand you have Edinson Volquez who started the WS tonight, good ERA but he hasn't reached 7 K/9 in the last 2 years so Steamer says he's pretty much Matt Garza. I don't disagree with this, Volquez is not good and I think Steamer is correctly calling him fluky.

 

It's important to remember that these predictions are baselines. 1-2 guys could improve their ERA by a full run through luck or mechanical improvements, but Steamer doesn't include luck or mechanical improvements. Conversely, 1-2 guys will probably get hurt or be terrible, but we can easily replace with a guy who will throw a 4.5 ERA so a lot of these predictions are kind of like the worst possible scenario in our case.

Umphry, thanks for taking the time to explain things. It helps understand the situation better.

[Point Beer Is Best]

So... who are we taking with the #1 pick in the 2017 amatuer draft?

[3and2Fastball]

Didn't Steamer predict that the Royals wouldn't even make the playoffs in 2015?!!!

 

I believe that the game of MLB changes so rapidly that Steamer might be based on a outdated mode, in part.

[And That]

More interested in seeing Cleveland's projections.

I was waiting for someone to make this joke. Well done.

[ZBTMP]

http://www.fangraphs.com/blogs/2016-zips-projections-milwaukee-brewers/

 

ZIPs are out. Interesting comp on Arcia.

 

I'll take the Carter, Santana, and Jungmann lines. Weird love for Clint Coulter (relative to other minor leaguers) and distaste for Jorge Lopez.