More hilarious than ever: PECOTA projects the Orioles at 71-91

[Frobby]

1 hour ago, Number5 said:

Perhaps it lies in how they choose the "comps."  On Tillman's bb-ref page, Gavin Floyd's name is nowhere to be found under similarity scores.  Interestingly, Max Scherzer is listed as the 9th most similar pitcher thru age 28 in MLB history to Tillman.  Think PECOTA would have come up with different projections for Tillman had they used Scherzer instead of Floyd? 

I don't know all the differences, but BB-ref uses career totals, PECOTA uses the most recent three years.   But they also probably weigh the different stats differently.

[25 Nuggets]

From a much more reputable source.... the bookies in Reno:

• 92.5 Cleveland
• 90.5 Boston
• 87.5 Houston
• 86.5 Toronto
• 86.5 Texas
• 85.5 Seattle
• 85.5 Detroit
• 84.5 Baltimore
• 83.5 New York
• 80.5 Kansas City
• 76.5 Los Angeles
• 75.5 Tampa Bay
• 73.5 Chicago
• 70.5 Minnesota
• 66.5 Oakland

This is quite a bit more reasonable, though I'd take the over here.

[Frobby]

1 hour ago, 25 Nuggets said:

From a much more reputable source.... the bookies in Reno:

• 92.5 Cleveland
• 90.5 Boston
• 87.5 Houston
• 86.5 Toronto
• 86.5 Texas
• 85.5 Seattle
• 85.5 Detroit
• 84.5 Baltimore
• 83.5 New York
• 80.5 Kansas City
• 76.5 Los Angeles
• 75.5 Tampa Bay
• 73.5 Chicago
• 70.5 Minnesota
• 66.5 Oakland

This is quite a bit more reasonable, though I'd take the over here.

1.   It probably is more reliable. 

2.    84.5 is very realistic.     I wouldn't want to bet that line, depending on the vig.

[weams]

10 hours ago, 25 Nuggets said:

From a much more reputable source.... the bookies in Reno:

• 92.5 Cleveland
• 90.5 Boston
• 87.5 Houston
• 86.5 Toronto
• 86.5 Texas
• 85.5 Seattle
• 85.5 Detroit
• 84.5 Baltimore
• 83.5 New York
• 80.5 Kansas City
• 76.5 Los Angeles
• 75.5 Tampa Bay
• 73.5 Chicago
• 70.5 Minnesota
• 66.5 Oakland

This is quite a bit more reasonable, though I'd take the over here.

Their actual money is at stake.  Not just some weighted statistical models. 

[Frobby]

37 minutes ago, weams said:

Their actual money is at stake.  Not just some weighted statistical models. 

I'm sure the smart money looks at the statistical models, and maybe have better ones of their own.    But anyone betting exclusively on PECOTA would be leaving common sense behind.

 

[weams]

19 minutes ago, Frobby said:

I'm sure the smart money looks at the statistical models, and maybe have better ones of their own.    But anyone betting exclusively on PECOTA would be leaving common sense behind.

 

Absolutely, but it is essential crowd sourced with Vegas. Unpopular teams are depressed in numbers and top market teams are uplifted. It is just a matter of finding the osmotic equilibrium. 

[DrungoHazewood]

On 2/9/2017 at 5:54 PM, Frobby said:

As best I understand it, PECOTA identifies the 10 most comparable players, looking at the last three seasons of data for the player in question, and then looks at the subsequent performance of those 10 players to project the future performance of the player.    For example, the most similar player to Chris Tillman as of 2016 is Gavin Floyd through 2012.    I don't know what criteria they use to determine who is most similar.   Floyd fell off a cliff in 2013, so that impacts the Tillman projection.    Interestingly, Tillman's top 10 comps include Wade Miley as of 2016 (#3 comp) and Ubaldo Jimenez as of 2013 (#8 comp).    

Tillman (2014-16): 552 IP, 1.296 WHIP, 3.99 ERA (104 ERA+), 3.2 BB/9, 6.7 K/9.

Floyd (2010-12): 549 IP, 1.295 WHIP, 4.25 ERA (101 ERA+), 2.7 BB/9, 7.3 K/9.   

That's at least how it used to work, and for all I know it still does.  Superficially that seems to be a reasonable approach, but if your top comp pulls something and tries to play through the year anyway and falls apart that probably has nothing to do with you but crashes your projections.

Marcels is at least as accurate, and all it does is a 4-3-2-1 weighting of the last four seasons with a dash of aging.

[Frobby]

10 comps may not be enough.    In any event, there are many things that happen in a season that no formula and no person can predict.    Pitchers get hurt or are healthy all year, etc.

[BohKnowsBmore]

On 2/9/2017 at 5:54 PM, Frobby said:

As best I understand it, PECOTA identifies the 10 most comparable players, looking at the last three seasons of data for the player in question, and then looks at the subsequent performance of those 10 players to project the future performance of the player.    For example, the most similar player to Chris Tillman as of 2016 is Gavin Floyd through 2012.    I don't know what criteria they use to determine who is most similar.   Floyd fell off a cliff in 2013, so that impacts the Tillman projection.    Interestingly, Tillman's top 10 comps include Wade Miley as of 2016 (#3 comp) and Ubaldo Jimenez as of 2013 (#8 comp).    

Tillman (2014-16): 552 IP, 1.296 WHIP, 3.99 ERA (104 ERA+), 3.2 BB/9, 6.7 K/9.

Floyd (2010-12): 549 IP, 1.295 WHIP, 4.25 ERA (101 ERA+), 2.7 BB/9, 7.3 K/9.   

Did you just put a few things, or are those the most important dimensions on which they compare?  

[Frobby]

28 minutes ago, BohKnowsBmore said:

Did you just put a few things, or are those the most important dimensions on which they compare?  

I don't know which stats they compare and how they weigh them. 

[weams]

8 hours ago, Frobby said:

I don't know which stats they compare and how they weigh them. 

I just know if they were betting the farm on being accurate with the Orioles, they would have lost the farm Dan Duquette's first year. And every season since. They are missing something. 

[Frobby]

Just now, weams said:

I just know if they were betting the farm on being accurate with the Orioles, they would have lost the farm Dan Duquette's first year. And every season since. They are missing something. 

It felt excusable in 2012.    After all, nobody expected them to be good.   Even in 2013, it seemed justifiable to project the O's would take a big step back, since they'd barely outscored their opponents in 2012 despite winning 93 games.    But fool them five times?    

[Frobby]

On ‎2‎/‎7‎/‎2017 at 8:51 AM, Frobby said:

 

OPS Projections:

Davis .834 (.746/.921)

Schoop .727 (.640/.813)

Machado .813 (.733/.895)

Hardy .664 (.585/.742)

Kim .773 (.648/.901)

Jones .777 (.703/.853)

Smith .759 (.671/.849)

Trumbo .779 (.696/.864)

Castillo .724 (.726/.823)

Joseph .641 (.519/.767)

Flaherty .666 (.547/.787)

Rickard .712 (.584/.843)

Mancini .760 (.636/.886)

ERA Projections:

Tillman 5.06 (6.00/3.94)

Gausman 4.34 (5.23/3.27)

Bundy 4.66 (5.85/3.36)

Miley 4.56  (5.43/3.50)

Jimenez 4.99 (6.00/3.83)

Britton 2.94 (3.66/1.87)

O'Day 3.97 (5.05/2.75)

Brach 3.96 (5.02/2.75)

Givens 4.13 (5.28/2.86)

 

I thought it might be interesting to look at the 10th percentile and 90th percentile projection for each player, so I added them above.   

[Camden_yardbird]

They had a discussion of this on MLB Network yesterday.  The general concensus among the panelists was that there is clearly something the numbers don't account for and that thing is probably Buck Showalter and team defense.

Brian Kenny was surprised to learn that the Orioles had more regular season wins over the last 5 years than any other AL team.

[Frobby]

I don't think either Buck or the team defense is the main reason PECOTA annually underrates us.   I think it's two things:

1.   We rely heavily on our bullpen.   Bullpens are notoriously volatile, and so BP annually expects ours to regress towards the mean, but they haven't -- they've been continuously excellent.    

2.   Our starting pitchers just don't seem to have the statistical profile that suggests success, despite results.    Tillman's the perfect example.    They've got him as having a 10% chance of having an ERA of 3.94 or better.    Really?   He's beaten that figure in 4 of the last 5 years, so even if you think he's been a bit lucky, how can his chances of doing it this year be 1 out of 10?    The answer is, he's NEVER had a FIP below 3.99, and his career FIP is 4.45.    There's just something about Tillman's peripherals that doesn't compute, but when you watch him, you feel you're watching a good pitcher (at least, I do).