Hot Stove - Felix, Head vs Heart

Q.  Can you win short-term after a Felix trade?

A.  It would be easier. 

On paper -- on a Win Shares or WAR spreadsheet -- it's a piece of cake to trade Felix and add 3-5 wins overall.

You can deal a 5-WAR Felix for two 3-WAR players and then you can spend his $10-20m in the free agent market.

Hey, I'm stars & scrubs.  You stuff the family shillings into the 5 highest roster spots, and then you go about improving the more fungible roster spots.

But!  There's an exception to that.  You trade your #1 player if you can get TWO DIFFERENT Stars for him.  You definitely trade a $1.00 megastar for two different 85-cent All-Stars.   It works well in theory and it works even better in practice.

No Seattle roto league would be complete without BABVA trading a Joe Mauer for a Youkilis-Choo-Scutaro package and the league spending the next two weeks in a flame war over the veto.  :- )

Capt Jack could trade for Felix and Just. Clean. UP.  On paper, that is.


Q.  Like what kind of deal would make the M's better in 2010?

A.   Felix' value is mind-boggling.  You've got a shot at a Sizemore-Lee-Phillips deal here, even more so.

I like having one 5-WAR player at $11-13m.  I like it even better having two or three 3.5-WAR players for the same amount of total salary.

Could the M's win next year after trading Felix for (let's just say) Buchholz and Ellsbury and two other guys?   They'd probably be five games better, after spending his salary.

I wouldn't do the Adrian Gonzalez kind of trade.  I'd sort through twelve offers and take the best one that got me two non-arb players -- $1m impact players -- in places I had holes.  That's one SP and one 3B/SS.

Then I spend Felix' salary on an Adam Dunn or Jason Bay* or whoever.  Not that I like Bay in Safeco, but E=MC2 when it comes to available salary and 100 RBI.

Don't forget there's still the $25m available beyond all this.  Obviously, such a team is going to be more dynamic, and more dangerous -- on paper.


Q.  But that's on WAR paper.  On the field, you think they might be better with Felix?

A.  Yeah.

Fan loyalty is fine.  I do think you make allowances for that.  (Randy Johnson's departure still stings me bad.  My wife and I used to plan our weeks around watching him pitch.)

But more than that, there's the swagger of throwing a Randy Johnson or Roger Clemens or Felix Hernandez out there every fifth day.

The Angels can be scary.  They're not scary on the day Felix starts!  One monster HOF righthander flips the I/O switch on a ballclub's self-belief.


Q.  Is Felix worth +5 wins?

A.  As mentioned many times, I don't think Felix' "Runs Saved" capture the actual wins that he gives the ballclub.  A manager chooses when to change pitchers.

The team went what, 20-5 in Felix' last 25 starts, despite poor hitting.  That was a demonstrated +8 games right there in those four months (compared to 12-13, since they were below .500 in non-Felix games).

Don't try to sell me that the 20-5 was due to bounces of the ball.  It wasn't.  Felix Hernandez went out and muscled those 20-for-25 wins.

I'm not saying that +8 is the gospel, but I think it suggests Felix' on-field value to be more than his calculated WAR. 


Q.  So where does that leave you?

A.  We're not playing Strat-O.  It's a real clubhouse and we need a real ballclub.

The monster righty is the place you start.  Pay the man before Theo does.


Dr D



Psychologically, I tend to agree with your assessment, but I have a beef with their math too.  I believe value (if calculated correctly using a pythagorean model) accelerates with increasing divergence from average...a mega-star isn't worth the linear gap in run's worth the inverted pythagorean gap...which has a quasi-exponential lapse rate.  A true ace pitcher isn't worth 5 wins to the big #2's 3 wins.  He's worth 10 wins to the number two's 3 wins.  The same happens for the hitters (but not quite as severely because there are nine batting order slots so the payday is split more ways).


Though the 30 GM's wouldn't explain it quite as crisply as you do :- ) they all have historically bent over backwards to get that megastar.
It may be intuitive with them, but they know what a true rainmaker is worth.
I don't figure Felix as 5 wins to this ballclub.  I figure him a lot more than that.  Not sure exactly how much...


PCA credits Pedro Martinez (2000) with being worth roughly 14 wins above the zero-value margin (about 11 wins above average), whereas it credits Josh Beckett (2007) with being worth 6 wins...a still-outstanding mark, but not the kind of pitcher you attach multiplicative value to.  The ace-caliber seasons POP out of the PCA ratings...a lot more than they seem to do with WAR.


You have formulas for PCA published somewhere?
All player values are based on a 'quasi-exponential' margin?  If you've discussed that concept before, I missed it (probably in scan mode, LOL).
I don't ever remember seeing a player-value set based on a Pythag/exponential-value-over-margin paradigm.  That's awesome.
You have any predictive validity calculated for it? 

5 can read the PCA manifest...I don't know if you have time to go through the whole thing or's 75 pages (James' explanation of Win Shares is similarly lengthy...hard not to cover a lot of pages with verbage when explaining something as complicated as a full uberstat metric system)...I produced the explanation for anyone who wanted a copy.
To clarify the main idea of the system though, I use Pythagorean theory (hence...Pythagorena Comparative Analysis) to set the margin for performance, rate teams and players relative to the margin, and give significant additional value to performances that imply a forced sub-marginal performance from the opposition.  The system still uses the linear-above-margin approach for dividing credit among players (just as James does), but my marginal baseline is significantly higher than his was and the concept of supermarginal performance (adding a second value enhancement to player performances that imply the opposition is submarginal when facing a given player) places additional emphasis on superstar caliber play.  Think of the supermargin as adding the negative wins a supermarginal performance creates in the OTHER team to the team being analyzed.  I call this quasi-exponential because the supermarginal value is pythagorean, so the further above the supermargin you are, the higher you rise above your peers.
I'm not sure I discussed the supermargin in detail in the manifest...I was mostly laying out the logical foundation for my approach to sabermetrics...but I still think it's worth reading if you are actually interested in how I did my work.  PCA is a look into a little time machine for me...some of the methods I used were a little primative compared to the ideas I have today (I last published PCA after the 2004's been five years).  But when I read back over it, I am still proud of all of the careful thought I put into it.


I could see the thesis (backed by the kind of robust work you typically do) being the basis for an advanced degree.
I'm wondering, idly, whether your paradigm won't be the state-of-the-art 5 or 10 years on.  Depending on whether it's correct of course :- )
Just real quick, what other Mariners, if any, look different with this Pythag-over-margin paradigm?
In Bedard's good stretches, where has he ranked. 


I have short-hand calc methods like James does that I can whip out to get rough estimates.  I think Bedard was above the supermargin in 2007 but not since he's been a Mariner (his walk rates are too high in 2009, although that's probably caused by the injury implosion in June/July).
The 2003 Mariner outfield defensive unit was supermarginal (I don't rate the supermargin for an individual fielder since no one fielder can create sub-marginal performance form the opposition, but a whole defensive unit (Pitchers, catchers, outfielders, infielders) can, in rare cases, do it).  The 2009 Mariner outfield was roughly as good as the 2003 outfield, so it's fair to guess that the '09 outfield defense was slightly supermarginal (which is why the whole team has such a stunning DER despite Betancourt/Cedeno playing short most of the year and Lopez being only average-ish at second).  None of the Mariner bats were supermarginal this year, I don't think.  Ichiro, as good as he is, is typically scraping along the top of the standard performance really have to whip out a 170 OPS+ or something silly like that to break into the game-changing dynamics.
Aardsma probably was not supermarginal either with his walk rate, and he was the best bullpen representative.
I plan to completely recalculate PCA with some enhancements this off-season.  It won't be the full fledged idea I've been working toward - time doesn't permit me to do all of what I want to...but it will address some of the flaws with the old system and provide concrete ratings for the last five years.

Sandy - Raleigh's picture

I think you're likely correct, Matt, (to a degree), that WAR (and its ilk) are calculated linearly, when in too many cases, Baseball doesn't behave linearly.
That said, I think one of the reasons that any quasi-exponential behavior is going to be hard to track in this arena is that if in MOST cases, the uber-ace PREVENTS you from getting replacement level talent elsewhere, those sacrifices will likely limit overall wins.  Paying to get AROD - who I'm certain would've been a super-marginal SHORTSTOP at the time, hammered Texas so badly economically, they spent the rest of his tenure bailing water from the multitude of submarginal leaks.
This is one of the reasons I flinch at the idea of STARTING rebuilding plans with high dollar payouts. 
But, this, I believe, is the general problem with most of the mass populace studies in baseball.  The center will ALWAYS overwhelm the edges.  And when you're specifically talking once-in-a-generation performances, they are by definition outliers. 
That said, I believe that in baseball, the actual dollars paid out by the market to the uber-kings is very likely MORE than their actual value, (even with regard to their potential quasi-exponential impact).  They get the money, NOT because they improve results that much, (how many rings for the Yanks with AROD?).  But, they do, very measurably, impact attendance.  From a business standpoint, superstars are super.  From a winning games standpoint, I tend to believe their value is overstated by the very nature of baseball - which limits the impact ANY individual can possibly have on the final results.
Granted, SPs are a unique category. 


You're right that the Rangers had huge holes at a lot of other positions and it killed their overall wins...but an uberstat like PCA should SEE that (and it does...the Rangers' pitching staff as an entire UNIT was very close to the margin and their fielding units weren't much better).  You can track supermarginal performance easily enough.  The question is how do you decide what that supermarginal talent is worth.  My first guess was to place the supermargin at the exact opposite side of the value spectrum from the real margin and track supermarginal wins linearly up (on top of normal marginal wins) in the same way that I'd track sub-marginal wins linearly down from zero.  Maybe that's not right, but it's better to guess that than to ignore the supermargin altogether as all other metrics do.

Sandy - Raleigh's picture

Even if I completely accept your position, Matt ... my first question would be ... exactly how many players typically inhabit the super-marginal realm? 
If I use a competing stat, like FIP, Felix ends up ranked only 9th among starters in 2009, with a 3.09 FIP.  Comparing him to the elite competition, only two guys fanned 10+ per 9, (Lincecum and Verlander), while Felix was barely above 8.  There were a trio who had walk/9 below 3, (Vazquez, Carpenter and Halladay).  And a quartet of guys had HR/9 better than 0.50, (Greinke, Lincecum, Carpenter and Kershaw).
Fangraphs provides the ERA-FIP comp, (which should tend to capture the fielding aspect of a player's stats).  Well, Felix was #1, burying everyone except Carpenter), with a -0.60 E-F. 
Me?  I believe the Seattle TEAM defense was likely super-marginal in 2009, but I wouldn't be comfortable assigning super-marginal performance to ANY individual defender, (even using PCA). 


I don't assign a super-marginal score to specific defenders individually.  Units get the supermarginal bonus and they split it between all of the players involved in the same way the unit's marginal value gets split...basically the super-marginal credit gets turned into wins at the unit level and the individual fielders are just splitting a larger piece of pie.
The team defense does help Felix, but DNRA > FIP in accuracy and value IMHO and Felix (using the pitching analysis in PCA, which is essentially a slightly simpler version of DNRA where certain assumptions about the rate of XBH, GDP and rare events that aren't in the pitching record other than PBP data) are applied) did a lot on his own to prevent run scoring that isn't captured in straight DIPS.  I do not know if it would be enough for him to be well supermarginal...there are usually about 1015% of all pitchers in any given year who are (and Felix would suely be among that group), but most of them are only getting slight boosts.  The slope is steep (which I consider a good thing).

Taro's picture

Of course Felix's actual WAR in '09 was 6.9. His season was worth $31mil+ in last years market using FIP and he could get better. 5 WAR is an extremely conservative '10 estimate IMO.
At $18-20mil a year, Felix is actually quite a bargain. You'd need pretty wild return for that kind of guy.

13 in itself conservative if you assume some supermarginal performance and recognize that he was better than his FIP (accounting for the gains he created on batted ball results above the Mariners defense)

Add comment

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd><p><br>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.


  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.
  • Web page addresses and e-mail addresses turn into links automatically.