THE BOOK--Playing The Percentages In Baseball

Dave Cameron makes a good point about calculating pitchers’ WAR as defense-independent, since we’re including defense in nonpitchers’ WAR. What’s the best way of doing that—is it okay to figure RA as FIP*1.09?

Also, how do you calculate WAR for pitchers splitting time between the rotation and bullpen, when the only playing time information in the Lahman database includes G, GS & IP (i.e., no breakdown of innings by role)?

Easy: don’t use the Lahman database!  Baseball-Reference has the split data you want.

Yes, using FIP is the best way to do it, if you believe that his ERA already includes his fielders.

Great work by David!  I wish that the guys at BP (and other so-called analytical web sites) would do this (they are more than capable) before they give us their (sometimes) silly team previews.

I didn’t notice, but what did he use for playing time for each player?

Also, I do’t like the last column being called “value”.  It should be “FA equity” or “Salary minus value” or something like that.  At first I thought it was WAR times 4.4m plus .4 (which is the previous columnn).

Plus it is nice to know how much value a player has in terms of his salary versus his FA value, but I would like to see rather or in addition to, his real value given his service time.  That way you can evluate the efficiency of a team’s payroll right off the bat.  There probably should be two columns of “worth” - one, as if all players were FA and two, givn their service time.

I don’t disagree with you.  At the same time, this is light years ahead of what you see on other team sites.  Though Jinaz might already be there with his Reds site.  Eventually, you can see it with all the extra columns, but let’s get the new readers used to it first before we pummel them with too much information!

For playing time, he said he gave all the starters 85% of 700 PA (which is 138 games of 162).

***

The WAR and Salary Calculator blueprint I laid out is a very simple process.  I think most people are surprised it is as simple as I’ve shown it.

I think that it is as simple as I’ve shown it, and is as open as I’ve shown it, may be a reason that it WON’T be adopted.  People want to argue, and you can argue with simplicity.

Tango/#2:

I’m using the Lahman database to create Marcels. Well, no, that’s a lie: I’m just being lazy and using your PitchingMarcel2008 spreadsheet, but I wanted to run it myself to get G & GS in there. Either way, is there a quick way to determine the replacement level for a pitcher splitting time between roles, without having to do each pitcher individually?

If you’re going to be projecting the wins for an entire team like Dave has done, don’t you need to take into account the expected leverage of each of the relievers?

Aaron @6 - Dave included a leverage factor for relievers (check out Putz’s value), but he didn’t spell out exactly what it is.

It’d actually be interesting to see what the factor is by player given that a high-leverage reliever left in the Bedard trade and it’s not at all clear who’ll ‘get’ those innings.

Anthony, the average reliever I think gets 1.2 IP per game, and the average starter is at 5.9 IP per start, more or less.

So, you can use those numbers to figure out the % of innings that goes to starts and relief appearances.

For example, say you have Mulholland, 2004: 39 G, 15 GS (i.e., 24 reliefs), 123.3 IP.  How many starter innings and how many relief innings?

15*(5x) + 24x = 123.3

Solve for x=1.25

So, starter innings = 15*5*1.25 = 93
relief innings = 24*1.25 = 30

In reality: starter=87, relief=36

Maybe we find that with the swingmen, it’s more like 1.5IP per relief game.  So, instead of “5x”, we should use “4x”.

In that case:
15*(4x) + 24x = 123.3
Solve for x=1.47

starter innings = 88
relief = 35

***

So, you can generalize as:
4GSx + (G-GS)x = IP
3GSx + Gx = IP

So, for Mulholland:
3*15x + 39x = 123.333
Solve for x=1.47

Preliminary results for the Cubs (on offense) are available here:

http://spreadsheets.google.com/ccc?key=pMzjQGUNv8coDfWT31XiRQA&hl=en

I’ve color coded starters versus bench players for those who don’t know the Cubs roster by heart. I’ve triple-checked the numbers, and yet somehow none of them appear to be overpaid. I will tell you, I could not be more surprised.

Cubs are in the NL.  So, replacement is 2.0 for 3B.  In the AL, it would be 2.5.

Felix Pie is considered by Cubs fans as one of the best fielders in all of baseball.

You should have a “playing time” (or games) column.  Right now, you’ve got an implied 85%, which is ok, but you’d probably prefer something more variable per player.

Your bench players should be multiplied by 25% or something.

Outside of C/1B (and obviously Fukudome), I used Sean Smith’s defensive projections. I went in and gave Pie full credit for his defensive reputation (capped out at +15) and Ramirez’s adjustment.

As far as specific playing time adjustments - I’m going to hold off on that until I have the results of the community forecasts I’m helping tabulate for a larger Cubs site. So far I only have playing time estimates for two players (the site maintainer is doing one player at a time, so it’ll be a while before I have the numbers.)

As regards bench players - I’m using 250 plate appearances in the WAA formula, and then taking 85% of everything in the final WAR calculation. Should I leave WAA as it is and change WAR to .25, or change WAA to 700 and then take .25 of WAR?

For Fukudome I’ve got a projection of +1 in RF and -8 in center.  It’s based entirely on his speed score, as is much of Pie’s +6 since he hasn’t played all that much.

Full Cubs WAR, hitting and pitching, is now available:

http://otherfifteen.blogspot.com/2008/02/2008-cubs-preview-featuring-war.html

Also, I took the spreadsheet, ripped all the Cubs-specific elements out, and installed drop-downs to make it work for the AL and NL. But I don’t have anywhere to host it - it’s available upon request for anyone who wants an automated WAR calculator for 2008.

I have the same criticism of the otherwise great work by Colin.  That is the use of FA value for all players.  Every team will have a positive “surplus” value, but it tells us little about how good or efficient they are at salarying or evaluating players.  Teams with lots of pre-arb and pre-FA players will tend to have a large “surplus” and teams that have lots of FA will tend to not have as large a surplus, even if they are underpaying all of their FA.  Of course you have to give a team credit for being able to get lots of value from players who are still protected, but I would still like to see how good teams are at evaluating and salarying players using a formula for adjusting for a player’s service time.

For example, we see that the Cubs have 50 something million in hitting surplus and 20 something million in pitching surplus, but what does that tell us?  Nothing, unless we at least know how much surplus the average team has.  I have no idea what that number is.  So either adjust everyone’s “value” according to their service time, in which case the average surplus for a team should be around zero by definition, or value everyone as FA and then at least tell us how much less or more than the average team a particular team is surplus-wise.  I don’t really like this latter method, as I already said, because there are two separate things going on in terms of a team’s efficiency in terms of dollars per win.  One, the amount they sign their FA to (or give long-term contracts to protected players), and two, the number of protected players they have on their roster.  Surely I could take a team and just put all my best minor league players on the MLB roster and have some surplus value if I wanted to.  But my team would probably be only a little above replacement.  IOW, I could have better than average surplus value and still have a crappy team.

In the case of the Cubs, it should be pretty easy to do the sort of analysis you’re asking for, assuming I’m reading right. The Cubs only have two arbitration-eligable players, both of whom are middle relief. It’s a fair bet that Neal Cotts is going to be doing more pitching in the Pacific Coast League than the NL again this season, so it really comes down to figuring out if Michael Wuertz is making more than a middle reliever with his service time.

But if we remove the pre-arb guys making league minimum, the Cubs are looking at $14 million surplus value on hitting and $3 million on pitching. Wuertz alone is $3.6 million in surplus value, so what I’ve learned from this is that the Cubs have done a really bad job in paying for bullpen talent. Kerry Wood, Ryan Dempster and Scott Eyre are pulling down $5.34 million combined and yet somehow don’t project to 1 WAR as a group.

Anyways, I have come to realize that I’m babbling because it’s late, but you’re absolutely right that the surplus value numbers lack context - we have data for two teams right now, and if you were to pick an MLB front office to try and emulate I don’t think the Cubs and Mariners would make the top five of the list.

But it does have its uses - for instance, trying to explain why trading Felix Pie in a deal for Brian Roberts is simply horrible. (It comes up more than I’d like to admit.)

I disagree.  Replacement level players will have zero surplus value.  They will be paid the minimum.

We already know what the average MLB team surplus value is.  The average MLB team has some 33 or 34 WAR.  So, if this was an all-free agent team, they’d be paid 4.4MM x 33 above the minimum, or close to 150MM above the minimum.

The average MLB team is 90MM in payroll, or 78MM above the minimum (400K x 30 players).

That makes the average MLB team have a surplus value of roughly 70MM.

As long as a team spends less than 70MM less per season on its scouting, playing development, and minor league operations, it’s a good investment, compared to going for an all-free agent team.

Remember the equation:

Surplus Value in millions
= (WAR * 4.4) - (salary - 0.4)

By definition, a replacement level player has a WAR of zero.  So, such a player, by definition, cannot have a surplus value that is positive.

Colin, If you’d like I can take that spreadsheet shell and host it for you.  How big is it?  Perhaps it will need to be a zip.

Email me at rally monkey (numeral for the number five) at comcast dot net.

Just translate that and remove all spaces.

rally monkey (numeral for the number five)

There are five other people called rally monkey who have a comcast subscription ... man alive.

Maybe not John, I think Rallymonkey was taken, but of all the numbers I could have chosen to add to it, Five was my first choice.

Brain Downing, #5.  My firstborn cat is Brian Downing Kaat.