THE BOOK--Playing The Percentages In Baseball

I don’t object to the conclusion at all.  In fact, the idea that you can score runs without hitting HRs is pretty obvious. 

However using wOBA to define a fair tradeoff between 1B and HR and then calculating RPG feels circular to me.  wOBA is marginal run value.  What we see here is that if you drastically change the environment of hit distribution, the weights in wOBA are no longer valid.  So, what exactly is meant by “fair trade?” Perhaps you mean someone with equal cost on the open market, assuming an efficient market.  In that case, it appears you’d do better to replace your slugger with an equivalently priced hit machine.  He’s probably a better fielder, too.  The next market inefficiency?

I meant “fair trade” in terms of linear weights for the standard league (what we normally see in current baseball).  Basically, 1.7*OBP+SLG is maintained in either scenario.

You are correct that the “fair trade” would be different in Ty Cobb’s time.

So, I ran the same experiment in a team context.  I took the Phillies team batting line and then replaced the personification of this argument, Ryan Howard, with his HR-less wOBA equivalent. 

Howard is a 0.254 / 0.348 / 0.495 hitter and his replacement, with HRs replaced by 2.2 1Bs is a 0.326 / 0.407 / 0.383 hitter.  With Howard, the Phillies score 4.257 runs per Game.  With his 1B hitting replacement, they score 4.427 runs per Game.  That’s an increase of .18 RPG, which is worth 27.5 runs over the season or almost 3 wins.  Maybe I’ve done something wrong, since this is a much bigger effect than Tango found, but it seems like a more direct answer to the question of what does a lineup look like without its cleanup hitter.

So, I don’t think sluggers are undervalued.

You might have done something wrong.  In any case, the answer is going to be close.

Basically, Linear Weights (or wOBA) does a good job in describing the impact of a player, and having a slugger or not doesn’t budge the equation much, overall.

First, what I’m about to write is mathematical nit-picking, but this seems like a place where that’s ok to do (which is part of why I read and comment here).

My point was just that linear weights are the linearization of RPG.  If, for some reason, the linearization is wrong, then you can’t use those weights to measure value.  Here, you’re explicitly looking to see if those weights do apply, because that’s the question you’re trying to answer.  When doing that, using linear weights is inappropriate mathematically.  Practically, the point is made perfectly well, but it sort of bugs me when a linearization is being applied and you know it won’t be the right linearization.  The better question might be, “What homerless line would produce the same RPG as the one you started with?” That is the player you would be willing to trade the original batting line for, not the guy with identical wOBA.

Just double checked.  I must have mis-entered the Phillies’ team line with Howard.  I get 4.454 RPG w/ Howard and 4.427 without, 4.4 runs per season less with the replacement.  A small adjustment in the replacement’s line would even it up.

"Chances of Taking Extra Base” on the calculator:  what are those values based on?

By taking all of a player’s homeruns and turning them into singles it leaves the question of what opposing managers would do to defend that player. With no chance of a ball going over the fence would the defenses be position better against said hitter? Also, would they be able to sustain a high average or would the luck dragons come into play. And the hitter trading homeruns for singles better be a great base runner and the hitters in front of him better be as well.

The homerun is the only batted ball guaranteed to generate a run (unless you are Bengie Molina and miss 3b) but the idea of what a team of .300 hitters could do is still intriguing.

Larry: you are right to a point. 

In a league setting, where you are putting one homerless guy in a team of 8 typical guys, then the Linear Weights are fine.  That’s your standard.

The question is then what happens if you take that one homerless guy, who is average in that kind of league, and couple him with 8 other guys who are also homerless, and who would also be average in that typical league, but put them on the same team.

If their runs allowed remains average, then we know that there’s no impact in trading HR for singles.  If the homerless team increases the runs, then we have synergy gain.

So, it’s nowhere near as bad as your post makes it seem.

Math nits are encouraged in this blog.

RFK: I start with this:

http://www.tangotiger.net/destmob1.html

Then I make some tweaks to account for the fact that you can’t make an out on base in my Markov.

I play around with the numbers a bit until I get Markov LWTS that match (as close as I can make it) to the empirical LWTS.

I was surprised to find out that the Markov is set up to model all 9 hitters as the same.  That obviously changes the way the results ought to be interpreted.

I’m not sure that comparing a team of 9 sluggers with a team of zero sluggers is the best way to assign a value to “slugging vs. non-slugging”, any more than comparing a team of 11 NFL offensive linemen to a team of 11 NFL wide receivers, and describing the difference as “the value of speed vs. size"…

Not criticizing the Markov calculator, mind you, it’s great, but I think you really need to evaluate alternative hitter archetypes in context, to have any real confidence in the result.  Run a Markov for the Phillies lineup, and swap in Ichiro Suzuki or someone else with very few HR for Ryan Howard, and see what happens. 

That approach would also have the virtue of much more closely mirroring the sort of decisions an MLB front office has to make, which would enhance the applicability of the result, as well as its accuracy…

In the team context, the linear weights are fine.

I actually think the question is what happens if you replace Mr. BigFatCleanupHitter with Mr. HomerlessButGetsOnBaseALot.  Because that’s what the “Sluggers are undervalued” people are arguing.  They say you need Mr. BigFatCleanupHitter or you won’t score runs.  He’s indispensible.  But, of course, he’s not. 

On the other hand, there are zero MLB players who look like 0.326 / 0.407 / 0.383.  A bunch of guys with 0.280 / 0.370 / 0.460 type lines.  Only 3 of the top 20 wOBA guys have fewer than 20 HRs.  It may not actually be possible to get equivalent run production without the HRs.

Greg: you are correct.  The test that is being performed, with the Markov, is 9 hitters of average profile (.340 OBP, .410 SLG or whatever it was) against 9 hitters of powerless profile.

The 9 identical hitters of average profile will score very very close to 9 “typical” hitters of varying profile, such that their overall team OBP and SLG is also .340, .410.  So, it’s a fair test in terms of making everyone powerless (which is my point here of having no slugger).  I never said we’re comparing to 9 sluggers.

Thanks for pointing it out, for those who weren’t aware.

Probably what happens is that some observers, such as the writer of the linked post, are looking backwards. That is, they are not looking at how various theoretical equivalent hitters with different profiles (lets say R Howard vs W Boggs, just to put a face to it) can impact a team.  They are instead looking backwards at modern teams that score lots of runs, and noticing that they are much much more likely to have a couple of sluggers in the middle of the order than they are to have a couple of Boggs type hitters.

And so they conclude that the Ryan Howard batter is more valuable, in a more global sense, than his wOBA suggests, because of the actual batting profiles that exist in modern baseball. I would need to do more checking on this before dismissing it out of hand…

I agree that an interesting discussion and research can be had on synergy (and batting order by extension).  It’s too early for conclusions.

The slash line by the singles hitting Ryan Howard in post #3 reminds me a little of Phillies great Richie Ashburn’s career line.

Mr. Red - I looked up Ashburn’s career stats and they are a remarkable comparable to Howard’s 2011 stats in total value, but without the HRs. Ashburn’s modOPS (1.8*OBP + SLG but including ROE in OBP and excluding IBB) was only 6 points higher than Howard’s, 1.116 to 1.110.

So I loaded my own sim with the 2011 Phillies starting lineup stats as shown by the 8 players listed in Baseball Reference.Com plus a generic number 9 hitter based on the average number 9 hitter from the Phillies 2010 team from Retrosheet, and I ran 20000 games.  The average Runs per 9 innings was 4.32.  I then substituted Ashburn’s career line for Howard’s 2011 line leaving Ashburn in Howard’s number 4 lineup position slot and ran another 20000 games. The average Runs per 9 was 4.33 runs.  Pretty convincing evidence that you don’t need a HR hitter in the middle of the order as long as the non-HR batter has a total offensive value equal to the HR hitter.

Great job Peter!

***

By the way, Mike in his iPhone app, is calling that metric: wOPS and wOPS+.  I think that’s a great name, and goes with all the other weighted metrics (wOBA, wRC, etc).