Article: A Few Rays Of Hope In PECOTA's Depressing 2016 Twins Projections

[gunnarthor]

 

 

 

Highly inflated cluster numbers, or RISP numbers, generally even out to the mean over time.  The Twins may very well get "lucky" again... but they may also have a much better team offense, yet be on the bad side of "cluster luck" and score less runs than last year.  

I don't think anyone disagrees with this point.  I think the area of contention is how much of the Twins success is due to that.  It is often mentioned that the Twins were lucky in May last year bu the numbers don't really suggest that - they hit the snot out of the ball in May.  Their lucky month was actually August, at least runs-wise.  

 

The Twins weren't really unique clustering a lot of wins in a short period, which is what I think is part of this unstated criticisms of last years team.  Houston started the season 30-17.  The Twins best 47 game streak started about 4 weeks later and they went 27-20.  Outside of those 47 game runs, both teams were 56-59 the rest of theway.  

[bird]

 

From:

Aaron Gleeman        ✔   @AaronGleeman 

Twins moved to Target Field in 2010. Since then they've won 442 games and Baseball Prospectus' PECOTA has projected them to win 441 games.

 

 

Yeah, good thing he strung all those years together to mask how inaccurate the system was year to year. The system lacks integrity, and so do some of its defenders.

[TheLeviathan]

They were lucky when you compare run differential to wins. The point being made is that banking on another fluke month is certainly possible, but not probsble.

 

Houston should be thinking the same.

[USAFChief]

From:

Aaron Gleeman ✔ @AaronGleeman

 

Twins moved to Target Field in 2010. Since then they've won 442 games and Baseball Prospectus' PECOTA has projected them to win 441 games.

From: USAFChief

 

Since moving to Tucson, I've had one foot in a freezer during winter months, and the other foot in a bucket of boiling oil during the summer months.

 

On average, I should have been very comfortable year round.

[Kevin]

 

Ok, but I question how much 'cluster luck' is actually 'luck'. Is a manager's lineup worth nothing? Are coaches' encouragements, tips and helps valueless? Is a player's hard work meaningless?

 

The lineup card could have some impact on offensive 'cluster luck'. Knowing where to put defenders on a field could have some impact on defensive 'cluster luck'. 

 

If my premise holds true, then Molitor had a really good year. And the Twins could over achieve again next year.

"Luck is the residue of design." Branch Rickey

 

Luck can also be the explanation for what you can't explain.  And just 'cause the stats don't yet exist for explaining a phenomenon, does not mean that the explanation is luck.

[gil4]

 

2015 - projected 82 wins - off by 12

 

Something doesn't look right.  Did I sleep through the Twins winning 94 games last year?

[gocgo]

I view baseball projections kind of like all star wrestling...purely entertainment.

 

Think about how many injuries occur during the season.  No way to project that and then think about how many guys have stuff that is nagging and they try to play thru it only to see their performance suffer.

 

Project all you want boys but the beauty of baseball is the unpredictability of it.

 

One of my favorite lines from Bull Durham - "Know what the difference between hitting .250 and .300 is? It's 25 hits. 25 hits in 500 at bats is 50 points, okay? There's 6 months in a season, that's about 25 weeks. That means if you get just one extra flare a week - just one - a gorp... you get a groundball, you get a groundball with eyes... you get a dying quail, just one more dying quail a week... and you're in Yankee Stadium."

-- Crash Davis

[gunnarthor]

 

Something doesn't look right.  Did I sleep through the Twins winning 94 games last year?

Hm, not sure what happened there.  They projected 70 wins, and were off by 13.

[Wookiee of the Year]

 

Yeah, that was Gleeman's way of obscuring PECOTA's unreliability.  If you look at it year by year it's a bit different - 

2010 - projected 82 wins - off by 12
2011 - projected 83 wins - off by 20
2012 - projected 71 wins - off by 5
2013 - projected 66 wins - perfect
2014 - projected 71 wins - off by 1
2015 - projected 82 wins - off by 12

So they were pretty good at saying the Twins would suck in 2013 and 2014 (big gamble there), were in the ball park in 2012 and no where close in 2010, 2011 and 2015.  But Gleeman's tweet tries to imply that PECOTA was nearly spot on for the six years.

You say unreliability, I say even over 162 games, it's tough to determine a team's true talent level. The lesson to take from Gleeman's stat is both that the projection could easily be off EIGHT WINS in either direction, and that it could be off eight wins IN EITHER DIRECTION.

 

There's two ways a projection can be off: One is by poorly estimating a team's true talent level, the other is by estimating it correctly, but having luck get in the way of the team demonstrating that talent level (whether they overshoot or undershoot). It's true that even with perfect knowledge, the limit to a projection system is plus or minus 6.4 wins. But if the projection system accurately projects a team's true talent, then it's serving a useful purpose. How to know if it's accurately projecting a team's talent? Maybe see how it does over a period of say, 6 years rather than 1 year, and see if its misses even out.

 

Regarding "cluster luck": The thing that makes it luck is that it's not predictive. If it's a skill, you would expect it to be a repeatable skill. However, historically, teams who cluster their hits in June are no more likely to cluster their hits in July than a team that showed league average hit clustering or even league worst hit clustering. Same thing from one year to the next. If there's no rhyme or reason to whether something will continue, then it's random/luck, not a skill. Maybe Molitor's figured out the Cluster Luck Secret Sauce, but he'd be the first.

[by jiminy]

 

Interestingly, the common refrain is that the Twins were lucky in May by clustering their hits but that's not quite true.  In May the Twins were 4th in runs scored buy also 5th in OBP, 3rd in slg and 4th in wOBA.  They earned those runs.  Their "lucky" month was actually August.  6th in runs but 14th in wOBA.  

Thanks for this! I don't think I've ever seen this point before and it totally changes my perception of this oft-stated stat!

[Otto von Ballpark]

 

Interestingly, the common refrain is that the Twins were lucky in May by clustering their hits but that's not quite true.  In May the Twins were 4th in runs scored buy also 5th in OBP, 3rd in slg and 4th in wOBA.  They earned those runs.  Their "lucky" month was actually August.  6th in runs but 14th in wOBA.  

I don't know if the common refrain has only focused on one month, or on one side of the ball.  The Twins may have been lucky to escape April last year with a 10-12 record, given that their runs rank outpaced their wRC+ rank for the month, and their RA9-WAR outpaced their FIP WAR (which it did again in May).

 

I don't know that their BaseRuns performance was ever isolated to the month of May (I don't even think that split is available anywhere).

[gunnarthor]

 

I don't know if the common refrain has only focused on one month, or on one side of the ball.  The Twins may have been lucky to escape April last year with a 10-12 record, given that their runs rank outpaced their wRC+ rank for the month, and their RA9-WAR outpaced their FIP WAR (which it did again in May).

 

I don't know that their BaseRuns performance was ever isolated to the month of May (I don't even think that split is available anywhere).

David Schoenfeld at ESPN is a pretty constant critique of the Twins and often focuses on May being the lucky month.  I posted it elsewhere but I don't really think the clustering of wins means much except to a narrative.  I think you can certainly say that the Twins won more games than they were supposed to based on the RS v RA but even that should be given a grain of salt. The first week of the season was so bad it actually colored the results for the entire year.  -28 RS in the first week, +24 for the rest of the year.  Obviously, most teams will look better if you take out their worst week but that was still a really sizeable difference.  

 

In any event, I think the Twins team changed considerably over the year so I think the talent level we have is better than what the RS/RA suggests.  And that talent should improve.  That's generally why I'm optimistic about this season.  

[Mike Sixel]

 

Ok, but I question how much 'cluster luck' is actually 'luck'. Is a manager's lineup worth nothing? Are coaches' encouragements, tips and helps valueless? Is a player's hard work meaningless?

 

The lineup card could have some impact on offensive 'cluster luck'. Knowing where to put defenders on a field could have some impact on defensive 'cluster luck'. 

 

If my premise holds true, then Molitor had a really good year. And the Twins could over achieve again next year.

 

if it's real, why isn't it repeatable?

 

If a manager can help a team win 6 more games than they should, isn't that worth like 20-30MM dollars a year to a team?

 

Every time we have these discussion, it is clear some people just don't believe in math, but never put up a better way to project or predict standings. Never.

 

6/162=4%. Do you have a system that will be less inaccurate than that? Are you sure that's actually all that inaccurate? 

Edited February 19, 2016 by Mike Sixel

[gunnarthor]

 

if it's real, why isn't it repeatable?

 

If a manager can help a team win 6 more games than they should, isn't that worth like 20-30MM dollars a year to a team?

 

Every time we have these discussion, it is clear some people just don't believe in math, but never put up a better way to project or predict standings. Never.

 

6/162=4%. Do you have a system that will be less inaccurate than that? Are you sure that's actually all that inaccurate? 

I think the problem is the valuation and weight you put on your math numbers.  

Edited February 19, 2016 by gunnarthor

[Kevin]

 

if it's real, why isn't it repeatable?

 

If a manager can help a team win 6 more games than they should, isn't that worth like 20-30MM dollars a year to a team?

 

Every time we have these discussion, it is clear some people just don't believe in math, but never put up a better way to project or predict standings. Never.

 

6/162=4%. Do you have a system that will be less inaccurate than that? Are you sure that's actually all that inaccurate? 

The point, I think, of the comment was not that the projection systems are all terrible, but that assigning over- and under-performance to luck might be overlooking some factors that are not currently accounted for.

I do not think the comment was expressing a disbelief in math, only that the math might not be as complete as it could ultimately be.

 

[Mike Sixel]

 

The point, I think, of the comment was not that the projection systems are all terrible, but that assigning over- and under-performance to luck might be overlooking some factors that are not currently accounted for.

I do not think the comment was expressing a disbelief in math, only that the math might not be as complete as it could ultimately be.

 

That part of my post, about math not being real, was aimed at a lot of people, not you......

 

It is, of course, possible that it isn't luck, and something we can't account for, but then, you'd think it would be repeatable.

[Brock Beauchamp]

 

It is, of course, possible that it isn't luck, and something we can't account for, but then, you'd think it would be repeatable.

It's also possible "luck" consists of multiple data points for which we are currently unable to account and, when aligned correctly, skew results way out of proportion. We assume it's luck when it could be multiple things; some repeatable, some not.

 

It's a mistake to assume we're only missing one or two things that account for noise. It could be a handful. It could be dozens. It could be a lack of chances to exploit a particular skill (say, a baserunner taking home plate, something that drastically alters the outcome of a game).

 

Baseball is a complicated game and I'm not sure mathematics will ever be able to explain it fully because humans seem to enjoy playing the role of monkey wrench.

[Linus]

If there is cluster luck shouldn't there be cluster "unluck" that might offset it. How many times did the Twins get a bunch of hits but due to bad luck not score many runs?

[alarp33]

 

If there is cluster luck shouldn't there be cluster "unluck" that might offset it. How many times did the Twins get a bunch of hits but due to bad luck not score many runs?

 

Not sure if you are understanding how stats work.  Of course it shows both sides of the equation.  I thought that was obvious. 

 

https://thepowerrank.com/cluster-luck/

 

The Tigers were as "unlucky" as the Twins were "lucky"

[Mike Sixel]

 

It's also possible "luck" consists of multiple data points for which we are currently unable to account and, when aligned correctly, skew results way out of proportion. We assume it's luck when it could be multiple things; some repeatable, some not.

 

It's a mistake to assume we're only missing one or two things that account for noise. It could be a handful. It could be dozens. It could be a lack of chances to exploit a particular skill (say, a baserunner taking home plate, something that drastically alters the outcome of a game).

 

Baseball is a complicated game and I'm not sure mathematics will ever be able to explain it fully because humans seem to enjoy playing the role of monkey wrench.

 

Agreed, of course we can explain more and more, but that also shows us how much cannot actually be explained, and why, imo, sports has such appeal. It isn't like chess where the best player is almost always going to win......there is luck for sure.

[Brock Beauchamp]

 

Agreed, of course we can explain more and more, but that also shows us how much cannot actually be explained, and why, imo, sports has such appeal. It isn't like chess where the best player is almost always going to win......there is luck for sure.

Yep. Even if it's not luck, it may be (nearly) impossible to track.

 

Maybe Brian Dozier is *really good* at taking home plate in certain situations (let's say a left-center gapper where he tracks the ball but doesn't have that advantage in a right-center gapper). Maybe in 2015, that situation arose four times and directly led to two wins. Maybe in 2016, that situation only occurs once in a non-critical situation and leads to zero wins. That's not luck, it's just a rare skill that only arises a handful of times a season. Or maybe Dozier's ability to do something awesome is offset by Aaron Hicks getting *really* unlucky, which negates Dozier's skill-based gains and gives the team stat an appearance of "luck neutral" when, in fact, they were unlucky.

 

Metrics are awesome and I love the insight they give us into the game but it's hard to rely on them too much because, at best, they're somewhat incomplete. At worst, they're largely incomplete.