Only three players have won an MMP and only appeared on a ballot in that year. Josh Hamilton is one, Bryce Harper is another, Phil Rizzuto is the last one. Phil Rizzuto's career has yet to be covered (we will vote on 1940-1949 in 2018). Bryce Harper's career isn't over and he looks likely to make another ballot some time. Josh Hamilton is likely to be a one-and-done MMP with only 2011 left for voting before his career is covered.

Fred Lynn, Ron Guidry and Dwight Gooden are MMP plus-one.

Player #times-on-ballot
Pujols 11
Votto 7
M Cabrera 7
Halladay 6
Utley 6
McCutchen 5
Cano 5
Beltre 5
Mauer 5
Tulowitzki 4
Wainwright 4
Holliday 4

Players with at least 3 MMP awards or at least 11 times on ballot

Barry Bonds 7 18
Henry Aaron 0 16
Willie Mays 7 16
Mike Schmidt 2 13
Roger Clemens 0 13
Rickey Henderson 1 12
Honus Wagner 4 11
Albert Pujols 3 11
Mickey Mantle 3 11
Frank Robinson 1 11
Eddie Mathews 0 11
Cal Ripken 3 8
Joe Morgan 3 7
Mike Trout 4 6
Ty Cobb 3 5

We aren't finished covering Cobb, Wagner or Trout. Pujols has 2011 left.

What's the justification for leaving Cano off the ballot entirely?

One voter doesn't include defense because it's too hard to quantify. As I mentioned in the ballot thread- that kind of defeats the purpose of participating in a project like this.

It's a year of firsts. Josh Hamilton is the first Texas Ranger to win the MMP! (Alex Rodriguez won two AL MMPs as a Ranger.) Hamilton is also the first outfielder to win the AL MMP since Albert Belle in 1998 after a decade of infield dominance.

Congratulations to Roy Halladay, this year's pitching MMP. Halladay is the first Phillie to win the award since Steve Carlton in 1980. This is Halladay's first league-specific award as well.

Congratulations to Albert Pujols on being named the NL MMP yet again. Pujols has been the NL MMP for 5 of the past 6 seasons (David Wright edged him out in 2007). That moves Pujols into some pretty illustrious company as only Honus Wagner (8), Willie Mays (8), Mike Schmidt (6) and Barry Bonds (10) have won more NL only awards.

Finally, congratulations to Felix Hernandez, the AL pitcher of the year. Hernandez is the second Mariner to win the award, joining Randy Johnson (1995).

Henry Aaron 0 16

This is the most Hank Aaron thing ever. I love it so much.

Aaron was the NL MMP once and top NL position player twice

"Mike Trout 4 6"

I definitely need to get a job that starts later in the morning so I can stay up late and watch Angels games. This is a generational talent who at age 26, may already be HoM-worthy.

Did Aaron just keep losing to Mantle or something? He NEVER made MMP? That's really odd. Of course, we haven't gone through Musial's career, where he will have year after year of competing with Williams.

Note: I saw two comments on the ballot page that I had not seen and need to respond to:

1. Chris Fluit said that I did not include defense in my 2010 ballot. I did include defense!

2. DL asked about Halladay, Longoria and Cano's RPA values:

Halladay was just outside with a +24.16 runs. Longoria was at + 21.66 runs and Cano was at +14.32 runs.

If you didn't think that I used defense, then how do you account for the fact that I have Jason Heyward at #4 on my MMP vote ballot.

He appears on no one else's ballot, let alone so high as #4 - and that is another reason that I do not take other people's evaluation of defense seriously!

I am also a bit baffled by the high rating for Cano. He was only the fourth most valuable player and 5th most valuable, if you include Sabathia, on his own team! Was it for his 'defense'? I have him as a plus on defense for 2010, but only to the tune of +.48 runs.

Note: Prior to the 2009 season I do not include defense simply because it would require me to do an extended analysis, in detail, for every player in each season. At my age, I do not think that I'll be able to do that anytime soon, or if at all.

Yikes: one more note:

My RPA's, for more than two decades have included defense for each player and the defense behind each pitcher, but for the purpose of this historical rating evaluations for a single season, I would need more than the overall defensive rating that was averaged over three years (from my prior ratings) in order to go back and separate out each player's defense in each single season.

I'm just not understanding your system. Just comparing Uggla and Cano, you have Uggla (29.71) with approximately twice as much value as Cano (14.32). But here are there triple slash lines along with playing time and OPS+:

Cano: .319/.381/.534 (696 PA, 141 OPS+)
Uggla: .287/.369/.508 (674 PA, 131 OPS+)

Now those are pretty similar lines, so I could see how an evaluation system might rate them similarly. So why the difference? By any defensive metric (or just eyeballing, for that matter), Cano was by far the superior fielder.

Just a couple other comparisons:
- by WAA, Cano (5.6) beats Uggla (2.3)
- by WPA, Cano (4.07) beats Uggla (2.71)

Yankee Stadium was a strong hitters park in 2010.
Florida played in a strong pitcher's park.

The huge difference between the two parks mean that the OPS numbers are heavily skewed against Uggla.

As for defense, yes Uggla was below average at -1.62 runs on defense, but his offense was HUGE!

More details:

Cano walked 33 times fewer than Uggla and hit into 19 GDP's while Uggla hit into only 9. OBP has a multiplier effect. It is not linear.

In addition, Cano stole 3 bases while being caught or picked off 6 times! Uggla stole 4 bases but was only caught stealing or picked off two time.

My RPA formula WORKS! WAR and all its variants are fatally flawed.

Here are the RPA two or three best defensive players, at each position, for the 2010 season:

Catcher: Yadier Molina +15.53 runs; Humberto Quintero +11.60 runs; Russell Martin +10.58 runs.
First Base: Aubrey Huff +4.22 runs: Daric Barton + 3.58 runs; Lyle Overbay +3.51 runs
Second Base: Dustin Pedroia + 3.16 runs; Aaron Hill +3.02 runs
Third Base: Casey Blake +5.71 runs; Jose Lopez +3.98 runs
Shortstop: Josh Wilson 4.87 runs; Cliff Pennington +4.62 runs; Ramon Santiago +4.20 runs
Left Field: Carl Crawford +12.28 runs; Brett Gardner +9.38 runs
Centerfield: Drew Stubbs +7.86 runs; BJ Upton + 7.72 runs; Julio Borbon +7.13 runs
Right Field: Jason Heyward +10.23 runs; Wil Venable +9.63 runs; Justin Upton +9.02 runs

You can clearly see that catcher and the three outfield positions are much more critical on defense, than the four infield positions.

The outfield is where balls that fall in, drop for doubles and triples, wheres as balls hit up the middle, on the ground can, usually, only go for a single.

First base and third base are slightly more critical than the middle infield because balls hit down the line end up as doubles.

I know that this goes counter to all you think that you know, but these are the FACTS!

Did Aaron just keep losing to Mantle or something?

I'm just guessing, but my bet is that it's a tough, tough thing to be playing in the same time and league as Willie Mays.

Some balls hit down the line that are hard enough to get by a first baseman go for doubles but some balls past the shortstop end up as gap doubles also.

OPS numbers are heavily skewed against Uggla.

OPS+ includes park effects

RPA park factors are clearly different from OPS+ although the difference between the two hitters may rely more on the internals of their performances. OPS+ park factors appears, at quick glance, to be based on the Runs Created formula, which is linear and simplistic. RPA park factors are very detailed, using all available data from the play-by-play file downloaded from Retrosheet and processed in my software.

As I stated, the internals of their performances, with the far fewer walks and far greater GDPs and caught stealing and picked off in regards to Cano, may account for a good portion of the difference. Even in taking the extra base, Uggla was thrown out twice and Cano three times. In RPA, the extra out involved in the GDP is added to Cano's plate appearances and the runner removed is subtracted from Cano's on-base numbers. The times thrown out on the base paths also removes an on-base occurrence for each runner removed. These adjustments affect the half of the RPA formula involving the 'set-up' value for the hitters that follow in the lineup.

What are the OPS+ park factors for NY and FLA in 2010? Can you let us know?

Here's the RPA park factors:

Yankees at Yankee stadium: .1423 RPA
Yankees at visitor park: .1227 RPA A drop off of almost 20 points at the visitor park!

All Visitors at Yankee Stadium: .1271 RPA
All visitors performances against the Yankees at the visitors park: .1150 RPA A drop-off of over 12 points!

The total drop off in performance, from Yankee Stadium to the visitors park, gives a park factor of .8823 to drop the performance at Yankee Stadium to league level of 1.000

Florida at Home: .1178 RPA
Florida at the away park: .1212 RPA

Visitor at Florida: .1125 RPA
All visitor performances against Florida at the visitor's park: .1218 RPA

The Florida park variant for 2010 was 1.0551 which increases the offensive performances, at Florida, to league level 1.000

In any case, when all the data was processed, Uggla was at .164 offensive RPA and Cano at .140 RPA. That's a big difference. Cano's was very good, but Uggla's was excellent.

Baseball reference park factors are found here:

https://www.baseball-reference.com/leagues/AL/2010-misc.shtml
https://www.baseball-reference.com/leagues/NL/2010-misc.shtml

Yankee Stadium is a 106/104 and Miami is a 103/103. They're almost the same.

Their formula is described here: https://www.baseball-reference.com/about/parkadjust.shtml

Ok. I went to the baseball reference site that you posted. Yikes! It was worse than I thought! No wonder the park adjustment factors are so different between RPA and baseball-reference!

It is utterly ridiculous that the Yankee stadium and Florida park adjustments that you posted were so similar. One was a big hitters park in 2010 and the other a pitchers park in 2010, but you would not know that from baseball-reference!

Baseball-reference is a wonderful source of player and team data, both current and historical, but analysis is clearly not their game!

That overly simplistic park factor makes ALL their player ratings bogus!

We have come a long way from the neanderthal era of player performance ratings. That is something that we can all take pride in, but we, obviously, have a long way to go! YIKES!

Dan Uggla's top two offensive seasons were in 2008 @ .160 RPA and 2010 @ .164 RPA.

Robinson Cano's top two offensive seasons were 2012 @ .152 RPA and 2013 @ .150 RPA.

While I have almost always had good to very good RPA annual ratings for Cano, he's been highly over-rated as to his actual value, for his entire career.

One side note and 'pet peeve': I still marvel over the 'baseball establishment' lowering the value of a fabulous hitter in David Ortiz, simply because he was a DH. In fact, while Ortiz was with the Bosox, his production at DH was often the biggest reason for their success, simply because most other teams had no understanding of the value of the DH position, by putting over-the-hill hitters and assorted 'junk' at that position. Prior to the 2016 season I told the Philadelphia SABR meeting that Mitch Moreland was no answer to the loss of David Ortiz. In fact, after I recently completed my analysis of the 2016 season, the Bosox were lucky to win their division. The Yankees were a far, far better team!

That overly simplistic park factor makes ALL their player ratings bogus!

I don't see why. They're adjusting on runs scored. You're adjusting on RPAs calculated. If RPA doesn't correlate to runs, then isn't the onus on you to show how runs scored is destroying information?

What exactly is RPA?

"OPS+ park factors appears, at quick glance, to be based on the Runs Created formula, which is linear and simplistic."
Pretty sure that A) their park factors are based on actual runs scored and B) unless something has changed of which I'm unaware, Runs Created is not a linear formula.

Runs per game is a single data point. Chance variation in single data points are enormous.

RPA uses all available data and for 162 games because it uses the home and away games as well as all the available data points for each. Each team has about 6,000 plate appearances per season and their opponents, home and away, also have 6,000 plate appearances in those 162 games. That's a lot of data to analyze and to come to a much more accurate park factor.

RPA uses single season park factor data ONLY!

I explained, in one of my first books, why using averaged multi-year park factors is a huge mistake, by using a particular season that threw the stadium park variations wildly out of line.

RPA does not work backwards from runs scored, simply because RPA was tested as a PREDICTIVE formula for run scored. It ignores the runs scored completely! It takes the internal game data, for those 162 games so as to come up with offensive production predictions of team run scoring that correlate excellently with actual team run scoring. The test is to be able to PREDICT performance. There is no predictive test for going in the other direction.

My park factors are useful because they are predictive of performance in terms of runs scored by a team over a full season. Are there exceptions, that fall outside of good predictive amounts for each team, every year, under the RPA process? Of course! That is true for any predictive system. However, the majority of team predictive runs scoring does very well with the RPA process.

In my historical ratings for pre-1990 teams, I was forced to use the unsatisfactory method of park factors based on run scoring alone. It is not something that I could have avoided, but it is not something that I would recommend as a method of analysis.

As far as runs created formula being non-linear, where does it have the individual data points, such as the single, double, triple, etc.. change value dynamically, based upon on-base production in regards to a players season totals? I see 'flat values' only. I'm not talking about individual game situational adjustments. How do you take a season line for a player with 600 AB's, under runs created, and make the values of the individual events 'flow' with the on-base production? Team scoring is most closely tied to on-base production, than any other item.

Runs per game is a single data point. Chance variation in single data points are enormous.

No, it's an average of 162 data points.

Its a single data point for home and away for home team and visitor. OK that could be construed as four points, but I demur. It is just one point and its not predictive.

In regards to Cano vs. Uggla, one more item:

In the OPS figure is onbase%, but that onbase percentage leaves out critical adjustments:

Uggla's 9 GDP's and his 2 Caught Stealing and 2 Thrown out on the base paths removed 13 base runners and added 9 extra outs.

In addition, his two IBB's should not be counted in onbase%

This results in a plate appearance figure of 589 AB's + 78 BB+HBP-IBB + 9 AB's (for the GDP out created) = 676 Plate appearances

Effective onbase% = 169 hits + 78 - 13 runners removed = 234 / 676 = .346 EOB% for Uggla.

Cano's 19 GDP's + 6 CSPO + 3 TO's removed 28 base runners and added 19 extra outs.

Cano's 14 IBB's should not be counted in his OB%.

This results in a plate appearance figure of 626 AB's + 51 BB+HBP-IBB + 19 AB's for the extra out created by the GDPs = 696 plate appearances.


Cano's effective OB% = 200 hits + 51 BB+HBP-IBB - 28 runners removed (6 cspo + 3 TO + 19 GDP) = 223/696 = .320 EOB% which is 26 points lower than Uggla's .346 EOB%.

OBP is more important than slugging%, in that scoring for others, after them in the lineup, is raised or lowered by the onbase% of those batters in front of them. Ob% has a multiplier effect and changes the value of singles, doubles, triples, etc...

"As far as runs created formula being non-linear, where does it have the individual data points, such as the single, double, triple, etc.. change value dynamically, based upon on-base production in regards to a players season totals?"
Linear just means each factor is multiplied by a weight. There's no multiplying or dividing of the various factors by each other. I'm not sure what you mean by changing the value of specific events dynamically. You clearly state that you are not changing event values based on in-game situations, so what is causing the event values to change?

"How do you take a season line for a player with 600 AB's, under runs created, and make the values of the individual events 'flow' with the on-base production?"
What do you mean by "flow" since again, you clearly state that you are not doing a WPA type in-game weighting system?

"In the OPS figure is onbase%, but that onbase percentage leaves out critical adjustments:
Uggla's 9 GDP's and his 2 Caught Stealing and 2 Thrown out on the base paths removed 13 base runners and added 9 extra outs."
I agree that OPS leaves out GDP and SB/CS data, but all the major WAR/WS systems include GDP and Baserunning in the Offensive calculation. For example, on BBRef, Cano beats Uggla 6.4-5.3 just in offensive WAR. In Offensive WS, Cano beats Uggla 24.3-22.2. In gWAR, its 5.6-4.9. In wRC+ (used for the offensive component of fWAR, its 143-135. In WPA, which takes into account the actual game situations, its 4.07-2.71. What value did Uggla bring that your system is catching, but every other system is missing? Or is this entirely that you are using drastically different park factors than every other system?

"RPA uses single season park factor data ONLY!"
You imply that single-season park factor data would have less various than multi-season park factor data. This makes no sense. On top of that Sun life's 2010 park factors are virtually the same on BBRef for 2010. It played as a mild hitters park during 2010 and the surrounding seasons.

"Dan Uggla's top two offensive seasons were in 2008 @ .160 RPA and 2010 @ .164 RPA.

Robinson Cano's top two offensive seasons were 2012 @ .152 RPA and 2013 @ .150 RPA."
Again, what is RPA? What does .160 RPA mean in baseball terms? Is it a measure of Wins? Runs? Compared to Average? Replacement? Zero? The only thing I come up with is Runs over Positional Average, but your numbers seem really low if that's what that is.

I'm trying to parse all of this. I think I would want to correct for opportunities in any GDP assessment. You can't penalize a guy if the person batting ahead of him is on base all the time. Cano batted behind Jeter, A-Rod and Teixeira that season. Uggla batted behind Gaby Sanchez, Hanley Ramirez and Jorge Cantu. What is the difference in GDP opportunities (runners on and < 2 outs)?

I also don't see how a GDP is all the batter's fault. Some of the credit goes to the defense for turning it, some of the blame goes to the baserunner. I think you're double counting a situation where the batter is really only responsible for the out they make. If the player ahead of them in the lineup had struck out before the Cano groundout why would that be better? Uggla struck out almost twice the rate as Cano. I think that's why Cano has more GDP. With runners on base Cano was able to put the ball in play and Uggla was not. In what world does a strikeout say that a player is a better batter than the guy who put the ball in play?

Where do you give Cano credit for those 14 IBBs? That is an out avoided and a baserunner with the potential to score runs. I would assume that figures into RPA. How do you break down the "non-intentional" intentional walk where the batter is just pitched around? That would kill Barry Bonds' late career productivity if you ignored those.

RPA means: Runs per Plate Appearance.

RPA uses the full season 600 AB's dynamically to determine the value of individual hits. A single, e.g., is .29 runs, but that value changes within the 600 AB's by the effective onbase% for half the AB's. A hitter has two responsibilities: 1. to the runners on base and 2. to the batters following. It is the value of the hits for the batters following in the lineup that change depending on the on base% of this hitter. That is why the dynamic change in value only affects 1/2 of the data. As for situational park effects for individual situations, that is another matter. I dealt with that about 20+ years ago, rating all the individual situations for all 30 parks, over several seasons, and have been thinking about doing an individual player situational rating for some time. It's another big job but I've got a lot of other 'balls to juggle' at this time.

No ballpark is on an island, even the ballpark that is fully enclosed. Everything is relative! Teams do not play all 162 games in the same park. If scoring is higher at the other parks, the park variant will make the enclosed park a pitchers park and vice versa. In any case, even in an enclosed park, temperature, barometric pressure, air conditioning and even external sun/darkness and fan attendance changing backgrounds and noise affect the play on the field.

Weather is huge! Parks often play hugely differently year-to-year, depending on the weather. In one of my first books I showed how lucky the NY Mets were when the were unsuccessful in getting any interest in trading Howard Johnson, after HoJO's supposedly terrible 1998 season, when he hit .230.

HoJo was #10 in the MVP in 1997 and #5 in 1999, so what happened in 1998? In 1998 all the eastern USA was cold in the Spring and even into the summer. Nowhere was this more evident whenever the Mets were at home. Shea Stradium was terribly affected. Once I adjusted the park variant on a year-to-year basis, HoJo's RPA ratings were a straight line improvement from 1997 to 1998 to 1999. HoJo did not have a bad year in 1998. And the Mets were very lucky that no team knew HoJo's real 1998 value.

Sometines the wind blows in. Sometimes the wind blows out. Sometimes it rains, but not enough to postpone the game. Some days it is blistering hot. Some days there is a high sky, that affects defensive ability. Some days the ball 'flies'. Some days it just does not fly.

Weather varies greatly from year to year.

HoJo was #10 in the MVP in 1997 and #5 in 1999, so what happened in 1998? In 1998 all the eastern USA was cold in the Spring and even into the summer. Nowhere was this more evident whenever the Mets were at home. Shea Stradium was terribly affected. Once I adjusted the park variant on a year-to-year basis, HoJo's RPA ratings were a straight line improvement from 1997 to 1998 to 1999. HoJo did not have a bad year in 1998. And the Mets were very lucky that no team knew HoJo's real 1998 value.

Howard Johnson retired in 1995. But I can see how in your system he could have produced value given the weather patterns in the late 1990s.

Barry Bonds being penalized by RPA??? Hahahaha... If anything, the RPA makes Barry Bonds even a greater hitter! Even with ignoring the IBBs!!

In Barry Bonds last season, 2007, he had an incredible .204 RPA!!! That's a Hall-of-fame rating! It is one of the facts that I used to proclaim that Barry Bonds was being blacklisted. There isn't a team on this planet that wouldn't salivate to have that kind of offensive power in their lineup! There was only one single player in all of MLB who had a higher RPA that 2007 season: Alex Rodriguez, and that was only by a single point at .205 RPA!!!

If I included IBB's, what we we do with the 8th place hitters who would be given IBB's, so as to get to the pitcher? That would really give an unfortunate and untrue value to their performances! An IBB is just a plate appearance that did not take place.


As for gdp opportunities vs. strikeouts, I think that your point iis well taken and something that I've thought about for some time, but do not have a clear answer. Strikeouts are a lot better than hitting into GDPS, which are the ultimate rally killer, but there is also the question of moving runners along when putting the ball in play. That is why I've long wanted to do the situational ratings, but I have always pulled back, not only due to lack of time, but also about their accuracy simply because situational events, by definition, are small subsets of data with, therefore, huge swings year-to-year that are inherent every time you split up events into smaller and smaller bits of data. Like when this hitter is 6 for 24 when playing at night or when facing the Mets or on Tuesdays, etc...

Thanks for the correction! I meant 1987, 1988, 1989 for HoJo!

Wow! 1987 is 30 years ago! I've been doing this so very long! Yikes!

I think your comparison of the lineup in front of Cano vs. Uggla doesn't hold water.

Here's the RPA's and the HR's for each. The reason that I include HR's is because that has a huge effect on the onbase%. There is no on onbase after a home run, including the base runners on, when the homer was hit. The HR LOWERS the effective onbase % for all the hitters following and, therefore, lowers the value of all hits after the HR. The HR has huge value for the runners already on base and that is where RPA gives it its due credit.

Here's the 6 players you listed, for the 2010 season:

For the Yankees:

Jeter: A horrible .106 RPA with 10 Homers
ARod: A pedestrian (for ARod) .137 RPA with 30 HR's
Teixeira: A good but pedestrian .139 RPA with 33 HR's

For Florida:

Gaby Sanchez: .132 RPA with 17 HRs
Hanley Ramirez: A terrific .160 RPA with 21 HRs
Jorge Cantu: a horrible .109 RPA (but still better than Jeter) with 10 HRs

I would contend, from those data points that Uggla probably had more opportunities to hit into GDPs, but partially avoided them by hitting the ball in the air and by striking out. Hitting the ball in the air is the main way to avoid GDPs.

"I would contend, from those data points that Uggla probably had more opportunities to hit into GDPs, but partially avoided them by hitting the ball in the air and by striking out. Hitting the ball in the air is the main way to avoid GDPs."
None of these data points actually tell us who came up with a runner on first with less than 2 outs more often.
Jeter: 136 Singles, 63 BBs plus 4 IBBs and 9 HBPs = 212 times on 1B
ARod: 80 Singles, 59, 1, 3= 143
Tex: 85, 93,6,13 = 197
Total: 552

Gaby Sanchez: 97,57,2,5 = 161
Hanley: 112,64,12,7 = 195
Cantu: 80,29,1,6 = 116
Total: 472

Now, I don't know how many of these occurred with less than 2 outs, plus some of these innings ended prior to Cano/Uggla batting and other guys also batted in front of them at various points in the season. But without analyzing play-by-play data, I'd say its very likely that Cano had significantly more GDP opportunities than Uggla in 2010.

A further note on effective onbase%:

Not only does HR's clear the bases but triples also clear the bases, while putting the hitter onbase. Effective Onbase% (EOB%) in RPA takes that into account. Also a double also puts the hitter on base, but also removes all base runners from 2nd and 3rd base and sometimes even from first base. Likewise a single will remove a runner from third base, and often a runner from 2nd base. RPA does its best to account for those removals of runners as well as the addition of the hitter to the base path in regards to singles, doubles and triples in the EOB%.

Carl: Cantu's numbers are a partial season numbers. He came over to the Marlins in-season. I'll do a Cantu+Wes Helms combo ( Helms was a horrible.102 RPA). They combined for 628 AB's and were the primary third basemen.

IBB is a subset of BB's. You can't add them together. That would be double-counting. You must subtract IBB from BB's.

Gaby Sanchez, 97,55,5, +4 reached on an error (ROE), -19 hrs -14 GDP -2 caught stealing, picked off and thrown out (CSPOTO). = 126
Hanley: 112, 52,7 + 7 ROE, -21 hrs - 14 GDP - 12 CSPOTO = 131
Cantu+Helms: 99,47,10 +7 ROE -14 hrs -18 GDP -3 CSPOTO = 128

Total = 385

Jeter: 136, 59,9, + 6 ROE -10 hrs, -22 GDP, -7 CSPOTO = 171
ARod: 80,58,3, +4 ROE -30 hrs, -7 GDP - 5 CSPOTO = 103
Teix: 85,87,13, +6 ROE -33 hrs - 15 GDP -4 CSPOTO = 139

Total = 413

Yes, a very small 28 more opportunities in Cano's 696 Plate appearances and 601 AB's, which can easily be accounted for by the fact that Yankee Stadium was a hitter friendly park and those 28 extra opportunities accounted for an increase of how many actual GDPs for Cano? Cao had 19 GDPs in those 413 opportunities. A rate of 1 GDP for each 21.74 opportunities. Ergo. it only accounts for no more than a little over 1 extra GDP as compared to Uggla!! And that single GDP can easily be accounted for by the fact that Yankee Stadium was a hitters park!

Uggla was a better hitter than Cano in 2010.

So how do you calculate RPA?

If grounding into a double play is really something that should be penalized with a double out for the batter then that batter should bunt _every single time_ there is a runner on first base and fewer than 2 outs just to avoid the DP.

I have no idea what question you are trying to answer with effective on base percentage.

Good catch on the IBBS. GDPs from a batter 2/3 slots in front of Cano/Uggla would be irrelevant to double play opportunities for Cano/Uggla since they are extremely likely to have occurred in the previous inning. I could see subtracting those for only the batter immediately in front of the batter in question; maybe half credit for 2 batters in front and none for 3 batters in front. Also, HRs a couple batters before the hitter in question should not be fully counted either.

That said, BBRef gives Cano -2 DP runs and Uggla an even 0 DP runs for 2010, so clearly Uggla was better at avoiding DPs than Cano, but probably not by as much as the 10 DPs difference would indicate. As DL has pointed out though, GDPs are definitely not worth a full double counting of the outs.

"Uggla was a better hitter than Cano in 2010."
Clearly every other method out there (of which I'm aware anyway) disagrees with you. That doesn't mean you are wrong, but I'd need to know your methodology to judge for myself.

"I have no idea what question you are trying to answer with effective on base percentage."
Ditto. It seems on the surface to be adding complication for complication's sake. Though, like RPA I don't know how its calculated, so maybe it will make more sense once he explains it. It certainly seems as though he is using an extreme park effect for Sun Life though.

BBRef gives Cano -2 DP runs and Uggla an even 0 DP runs for 2010

This stat is adjusted for opportunities, Cano was 2 runs below average and Uggla was average. RPA is saying there is a huge difference between the two players due to park effects and double plays. BBREF is saying there is essentially no difference between the two players due to park effects or double plays.

There is no difference in run scoring between pitching around a batter and intentionally walking him. I wouldn't take IBB out of any assessment. That player earned the walk. There are base-out situations that make a walk worth less than average but that applies equally to intentional and non-intentional walks in those situations.

"BBREF is saying there is essentially no difference between the two players due to park effects or double plays."
I think Fangraphs and Bill James are saying the same thing only they don't break it out as neatly as BBRef.

"I wouldn't take IBB out of any assessment."
I think he was saying that IBBs are already included in BB on BBRef so its double-counting to add them again. I think that's a true statement, but I haven't verified it.

"RPA uses the full season 600 AB's dynamically to determine the value of individual hits. A single, e.g., is .29 runs, but that value changes within the 600 AB's by the effective onbase% for half the AB's. A hitter has two responsibilities: 1. to the runners on base and 2. to the batters following. It is the value of the hits for the batters following in the lineup that change depending on the on base% of this hitter. That is why the dynamic change in value only affects 1/2 of the data. As for situational park effects for individual situations, that is another matter. I dealt with that about 20+ years ago, rating all the individual situations for all 30 parks, over several seasons, and have been thinking about doing an individual player situational rating for some time. It's another big job but I've got a lot of other 'balls to juggle' at this time."

So you use one set of coefficients for half the batter's at bats and a different set for the other half? How do you decide which half of his stats are applied to which set of coefficients?

Cano's effective OB% = 200 hits + 51 BB+HBP-IBB

If IBB are in BB, then just don't include an IBB term in the equation. (H+BB+HBP)/(AB+BB+HBP)

Sorry, I thought you were talking about my back of envelope calc of GDP opportunities. He is subtracting them out of Effective OBP as well, which I agree is wrong.

I did a little searching with the play index sorting by rDP (double play baseruns). The highest season ever is +6 and there are 17 seasons >=5. The worst season is -6 and there are 5 seasons <= -5. When you adjust for opportunities this just doesn't matter. The spread from the best to worst is 12 runs and nearly everyone is within +/-3 runs of average. Double plays are turned by the defense, not caused by the batter. Including them in an offensive analysis the way it seems to be done in RPA is _adding_ error into the calculation.

Caiman's (Mike Gimbel's) explanation of RPA can be found at the following URL. www.baseballthinkfactory.org/btf/scholars/visiting/articles/RPA_explanation.htm

By the way:

Florida had 80 wins in 2019. My RPA predicted, based on the data alone, without seeing the final record, that Florida should have won about 79.5 wins in 2010.

The Yankees had 95 wins, and the RPA predicted just over 93 wins, based on the season data alone, in 2010.

While not all results, or even most, should be expected to come that accurate, I find it remarkable how close my method came! And it could not have come anywhere close, if I had not used the park factors in RPA.

I should note that I balance out the results, offense + defense + pitching over the entire league so that the league win/loss value is as close to .500 as possible. That 'balancing' affects every player and every team with the same factor. Therefore, any changes kept every team's wins the same in relations to all other teams.

But ultimately, RPA seems to be based in Runs. I can see some difference in Park Factors cropping up, but going from a mild hitters park based on straight runs to an extreme pitchers park based on RPA is a big difference. If RPA is actually predictive of team runs scored, it seems to me that these park factors should be much more similar.

"Florida had 80 wins in 2019. My RPA predicted, based on the data alone, without seeing the final record, that Florida should have won about 79.5 wins in 2010.

The Yankees had 95 wins, and the RPA predicted just over 93 wins, based on the season data alone, in 2010.

While not all results, or even most, should be expected to come that accurate, I find it remarkable how close my method came! And it could not have come anywhere close, if I had not used the park factors in RPA."

Assuming you are applying similar park factors to both offense and defense, it shouldn't make to0 big of a difference in terms of predicting wins (I'm assuming by pyth method) whether the park factor used was accurate. For example if a team in a neutral park scored 700 runs and allowed 600, pyth predicts 92.3 wins. If you applied 120 park factor to both sides, it still rounds to 92.3 wins. If you change to 90 park factor, it still rounds to 92.3 wins. 90 to 100 to 120 is a pretty extreme park factor swing and it really doesn't affect win prediction much at all.

Carl: you are not doing win prediction. You are starting from the actual runs scored and are working backwards. The 'trick' is to work from the data so as to predict run scoring. You started, above, with the assumption of a 'neutral' park, which is an assumption that I would dispute, based on the method used. The runs scored less the runs allowed is a small data set that allows for a huge chance variation in results. I am not interested in applying park factors to the runs scored, as a result. The RPA, as stated, develops the park factors from the thousands of AB's at each stadium, for both the home team and away team and at the away park for the home team and the away team from that original home park, as shown above. in comment #24.

Also: I use a different park factor on defense. Every position, in each park, plays different from other parks. I compare the results of every batted ball, through each defensive position, in the same manner as the park effects, except it measures, e.g. the particular shortstop for the Yankees results vs. the opponent team's shortstop IN THE SAME GAMES, at the home stadium and the away stadium. As such, the results for Jason Heyward are for his performance against the RF's performing against his team in the same games ONLY. You cannot compare defense at the old stadiums, such as Fenway, with a stadium like Coors Field, without doing the position adjustment for the stadium involved. It is one of the fabulous beauties of the game of baseball that the individual fields are so different, one from the other. The RPA accounts for that difference, although there is at least one adjustment that needs to be made (but I do not have the time!) because not only is there a position adjustment, but there is a 'learning' problem for visiting players, that give an advantage to the home defensive players. Some parks, when I did that study years ago, like old Shea stadium, gave almost an advantage to the visitor due to the 'cookie cutter' design of that park. The Colorado defender, coming from a difficult defensive park, had an advantage over the Mets defenders AT SHEA STADIUM, simply because the Colorado defenders were going from a difficult park to an 'easy' park to play defense in! Parks with extreme differences, be it Colorado and its elevation, or parks with 'weird' shapes and special or bad backgrounds, bad infield conditions, or any other particular difference from what the visiting player expects, gives the home defender a big advantage.

An example of a player who was best shown for his defensive prowess, in RPA, was Ozzie Smith.

When I read, in on of the Sabermetric publication, that "we all know how great Ozzie is, but we can't prove it in the data". When I did my defensive study, in the method shown, Ozzie (OF COURSE!) ended up as a phenomenal defensive shortstop, even at Ozzie's old age year, near the end of his career, that I used for the study.

Another example was the unfortunate shifting of Cal Ripken from SS to 3B. My RPA showed that Cal Ripken was still an excellent shortstop and moving him to 3B destroyed both his defensive and offensive advantage for the Orioles! The shift was a big mistake.

So you're developing a park factor for each individual event? ie a park factor for singles, doubles, groundouts to SS, etc.?

Another example is Roberto Alomar. My defensive RPA ratings for Alomar showed him to be an average to slightly below average defensive 2B.

He was playing on the carpet in Toronto and it distorted the impression of his defensive prowess.

I'm a New Yorker and my co-workers and friends were often huge Mets fans. When the Mets acquired Alomar, they couldn't believe me when I said that Almoar was not a great defensive second baseman. That belief of Alomar's defensive prowess quickly changed after they got to see him play every day and he showed he was not the spectacular defensive player they had been led to expect.

Yeah BBRef WAR has him just slightly above average overall, but below average during his time in Toronto.

So you're developing a park factor for each individual event? ie a park factor for singles, doubles, groundouts to SS, etc.?

Carl: No No No! Where do you get that idea that I developed a park effect for each individual event?

How do you get your park effect then?

A couple of examples to explain the defensive RPA ratings:

Carl Crawford's +12.28 runs in LF was based upon these numbers:

The opponent LF @TB allowed an average of .148 runs (per RPA) for every ball hit into LF.
Carl Crawford allowed a value of .132 runs per the 236 balls (143 outs, 59 singles and 34 doubles) hit into LF, while he was playing LF @TB. That's an advantage of .016 runs per every ball hit into LF @TB

The opponent LF @the opponent's park allowed an average of .155 runs per ball hit into LF.
Carl Crawford allowed a value of .121 runs per the 250 balls (158 outs, 63 singles, 29 doubles) hit into LF, while playing LF at the opponent's park. That's an advantage of .034 runs per every ball hit into LF.

That's a total of 486 balls hit into Crawford's defensive area. Even with 486 data points, it is possible to get lucky, just as with some years hitters get lucky, but RPA treats defense just as it treats offense for a batter with 486 balls put in play. A credit for the hitter is a debit for the defender and vice versa. It should all average out over a full season, but doesn't always, as we all know.

Dustin Pedroia missed half the season in 2019 but still managed to lead the MLB in defensive value at 2B.

The opponent 2B, at Fenway, had an RPA defensive value of .065. Notice the huge difference in that value from the value of each ball hit into LF in Crawford's example. It is much harder to be outstanding on defense in the infield because the overall value for each ball that is hit on the ground is so low.

Pedroia's RPA on 135 (112 outs and 23 singles) balls hit to 2B area, @ Fenway, was .049, an advantage of .016 runs per ball for Pedroia.

The opponent 2B, at the opponent's park, had an RPA value of .074.

Pedroia's RPA on 77 (61 outs, 15.5 singles, .5 doubles - the .5 results from balls hit into the area between Pedroia and the first basemen, where their is shared responsibility) balls, for an RPA of .061, an advantage of .013 for Pedroia. Since this only involves 212 balls put into play, over that half season, it is likely that the differences, per ball, home and away would narrow a bit for Pedroia over the full season.

Note for the above:

The opponent's RPA, at the particular park, is based upon the full season data for all the opponent players at that position, at that park.

2 Things
1) A player has an RPA on offense and defense for every park they played in for a given year and that's all added together for their season total?
2) If a ball is hit to LF for example, the LF and the hitter are the only players credited or blamed for what happens?

Pedroia's RPA on 135 (112 outs and 23 singles) balls hit to 2B area

Pedroia certainly turned some double plays and had more outs per ball hit to him. I think overall when you add things up a the team level your system looks to be reasonable but I think your accounting for double plays is putting credit in the wrong player buckets which is causing some distortions. Outfielders almost never turn double plays.

"The opponent 2B, at Fenway, had an RPA defensive value of .065. Notice the huge difference in that value from the value of each ball hit into LF in Crawford's example. It is much harder to be outstanding on defense in the infield because the overall value for each ball that is hit on the ground is so low."
Are you saying that a good defensive LF is more valuable than a good defensive 2B? Or are you saying that a good LF stands out more because the other LF general aren't great defenders?

Also, I confirmed on BBRef that OPS+ and ERA+ adjustments use single-season park factors and not multi-season. Any difference in Sun life park factors for 2010 lies in the actual calculation of caiman's method vs BBRef's run-based method. I did not find out which park factor is used for WAR calculations, but it stands to reason that it would be the same as OPS+/ERA+.

Found it; they do use 3-year Park Factors in WAR. Not sure why the disconnect between WAR and OPS+/ERA+.

I read through that article MrC posted in #57 that gives a bit more detail into caiman's method. Thanks by the way. A few thoughts on just the offense side of the equation. In selecting a version of WAR to use for HoM (and by extension, these MMP votes), I wanted to get at the player's value as an individual. That's not to say that the team lineup dynamic isn't important, just that its more useful in evaluating the GM (how the team is constructed) and the Manager (how that lineup is deployed). I feel that a linear weights system does a better job of isolating an individual offensive player's contribution than does a non linear system such as BaseRuns or Runs Created. These systems do a much better job of evaluating the entirety of a team's offense than a linear weights system. The individual player doesn't really have control over what situations his teammates present him with or what his teammates do with the situations with which he presents them (beyond distracting the pitcher with his peskiness on the bases and taking extra bases as baserunner).
From what I've read on RPA, it seems like on offense, it falls more into the BaseRuns/RC category (although it takes a much different route to get there) in that it tries to account for the dynamic nature of an offense; ie the 2 basic components of getting on base and then advancing/scoring the runners on base.
In summary, I guess what I'm saying is that it depends what question(s) you are trying to answer.
1) Measures like rbat, rbaser, rdp (BBRef) or wOBA, wSB, wRC+ (Fangraphs) that use linear weights to determine offensive value tell us that Cano was probably (remember, whatever system we are using is only generating an estimate of value) a better offensive player in isolation than Uggla in 2010. I'd like to see what Fangraphs or BBRef think an offensive confidence interval would look like for these 2 players, but I'm comfortable enough with the measurements in linear weights to say that its more likely Cano was the better offensive player. When adding defense into the equation, the differences become large enough that I am close to 100% confident Cano was the better overall player.
2) RPA is telling us (IMO) that Fredi Gonzalez probably did a better job of deploying his offensive resources (or at least Uggla and those around him in the order) than Joe Girardi did with Cano and those around him.

Defensive ratings, like all the RPA ratings are relative. LF is where the worst OF'ers usually play. Therefore, Carl Crawford had an advantage that is reflected in the runs prevented by him. I am not rating a player's defensive talent! I am rating that talent as it compares to others at that position, at the same park on the opposing team. Nothing more.

However, all three OF positions have a huge difference from the infield positions in the value of balls hit through the position. It does not matter which way the ball is hit. If it is hit in the air, the ball has a much higher value in terms of run production.

I'll have to pour through my software to answer your question about the defensive GDP. My software is huge, with many, many thousands of lines. It has been a long time (several years) since I wrote the defensive portion. I assume that I gave a ball hit to Pedroia, the same way that I handled offensive GDP's, accounting for the full result, although I have no current recollection. At this point, I can only say that I assume that I did. Wish that I could be clearer.

One more thing: I do not care about errors. If the batter hit a grounder, and the infielder threw the ball away, with the batter ending up on 2nd base, my software credits it as a double against the fielder. It is where the batter ends up that matters, not what the official scorer recorded. As such, my software runs through every play, in the downloaded Retrosheet database, to determine the actual result. The same goes for the OFer. If the OF botches the ball, even if it were officially called a single, but the batter went all the way around the bases, it is scored by RPA as a home run against the OFer.

By the way:

I assume that the Yankee intention is to play Stanton in LF.

He's not a good defensive RF, while Judge appears to be a good RF.

Judge was outstanding with a +6.55 runs in 2017.

Here's Stanton's bad defensive runs situation for the last three years for RF:

2015: -3.25 runs
2016: -7.49 runs
2017: -2.94 runs

He'll be greatly benefited by playing LF, but if he can't cut it there, despite having weaker defensive players at that position, then he's a DH in the making.

It would be a mistake to move Judge from RF. I assume that they are NOT contemplating doing that.

Just some notes:

Byron Buxton was an incredible +14.84 runs on defense in CF in 2017.

One of the best, if not the best player available as a free agent this year is Lorenzo Cain.

Cain's offensive RPA, in 2017 was a very strong .149.
Here's Cain's defensive numbers for the last 5 seasons:

2013: +7.04 runs
2014: +8.26 runs
2015: +8.23 runs
2016: +9.35 runs
2017: +5.52 runs

DRS sees them roughly even. Problem is, they have one of the best defensive LFs in baseball who probably moves to CF. Unless Stanton or Judge becomes the regular DH, their OF defense probably takes a hit. Not anywhere near enough of a hit to offset the offensive gains though.

Note: This is from the 2011 MMP Discussion thread, but it related to some 2010 issues, so I'm pasting it here as well.

"The age adjustment is NOT for evaluating past performance. As stated, my RPA ratings, each year, for players at each position, are for the coming season, after evaluating the player's past two seasons. As such I add or subtract points for the age of the player."
So you're voting on players for 2011 MMP based on your projections for the 2011 season? We know what happened in 2011 already. We don't need projections.
This would explain the Dan Uggla being rated better than Robinson Cano issue from 2010. He's rating 2010 based on 2008 & 2009 plus an age adjustment. Based on a combo of 2008 & 2009, I would rate Uggla better than Cano as well (even without using his park factors and minimized 2B defense). Problem is that we were voting on what they did in 2010.

Problems with defense and positional adjustments aside, surely MMP voters are required to vote on what the players did in the year for which we are voting!