THE BOOK--Playing The Percentages In Baseball

Wednesday, February 08, 2012

New PECOTA

This was quite a surprising claim from Colin:

These values are on a very different scale, since due to the lack of an intercept the values have to sum to one for the first regression and to three for the second regression, but they’re also very different in a more meaningful sense; recasting the first year to 1 (which is practically already done for us), we get weights of 1/.92/.90.

As you know, Marcel uses 5/4/3 for hitters (meaning 1, .8, .6) and 3/2/1 for pitchers (1, .67, .5).  (I think it was 3/2/1… can’t confirm right now.)

I personally use .9994^daysAgo for hitters and .9990^daysAgo for pitchers, which has the effect of being 1, .8, .64, .512 (and so on, each 80% of the previous) for hitters, and 1, .7, .49 (and so on, each 70% of the previous) for pitchers.

Tests from other research makes me think that it should be even more aggressive, so maybe 1, .7, .5 for hitters and 1, .5, .25 for pitchers.  But, I haven’t researched that, so, I’ll just leave it there for now.

Colin has gone way to the other side, essentially going with a .9998 or .9999^daysAgo kind of model.

Now, I agree with the framework for his testing, that you should and must include the PA component when establishing the weights.  Frankly, this is an important step.  When I did it for Marcel, I basically forced everyone in the system to have at least 300 PA, so that I didn’t have to worry about this portion too much (I should have worried a little about this, at least).  Indeed, if you give everyone at least 500 PA for each of the three years, this step becomes basically unimportant (no worries at all).  That’s because the weighting of each year (the PA of year 1 divided by the PA of years 1 + 2 + 3) will be the same for each wOBA of year 1, 2, and 3.

So, getting back to Colin’s important point: he’s saying that if you introduce the PA weighting component, we see that every year is important.  I find this very hard to believe.  I mean, it’s an exciting finding if true, and I’d like to see more research on this for sure.  My guess at the moment is that there’s a selection bias issue, with guys of limited number of years, or for young guys. 

Basically, does Colin’s finding apply across-the-board, or is it really limited to a subset of the population?  I’d bet on the latter, and I’d bet that the Marcel 5/4/3 would still hold for players who are regulars.  In any case, it’s an exciting prospect to consider.

***

A correction to Colin’s note here:

The third, and perhaps most important, takeaway has to do with regression to the mean. We can add a simplistic version of regression to the mean to our forecasting model by adding a TAv_REG of .260 (the league average) with a PA_REG of 1200. (The PA_REG comes from the Marcels; it’s included here mostly for the purposes of illustration. The regression component in PECOTA is a more rigorous model based on random binomial variance—again, the purpose here is only to illustrate the concepts.

Consider a player with 650 PAs in three straight seasons, or 1950 total PA. Using the Marcel weighting of 1/.8/.6, that comes out to 1560 effective PA— in other words, throwing out 20 percent of a player’s PAs during that time period. That means 56 percent of a player’s forecast comes from his own performance, and 44 percent comes from the regression to the mean component. Using weights of 1/.92/.90 yields 1833 effective PA, throwing out only about six percent. Using the same regression component, that’s 60 percent of a player’s forecast coming from his own production and only 40 percent coming from regression to the mean. (And if you follow from the conclusions above and start using more years to forecast a player as well, even less regression to the mean is necessary.)

There’s a calculation error in there.  Marcel uses 5/4/3/2 model, with the 5/4/3 being the weights for years T, T-1, T-2, and the 2 being the weight for regression toward the mean (using 600 PA as the seasonal number).  So, if you had say 700 PA in year T, 400 in year T-1, and 500 in year T-2, you get these effective weights:
year T: 700 x 5
year T-1: 400 x 4
year T-2: 500 x 3
regression: 600 x 2

That 600x2 is the same for everyone.  Colin’s calculation error is that rather than using 5/4/3, he used 1/.8/.6.  The net effect is that he showing a far bigger regression amount than Marcel is actually doing.

Expos breathing

Fun stuff.

Batman, the webslinger?

It sure seems early to do a reboot, considering that SpiderMan 2 is one of the best comic-book movies ever, and SpiderMan 1 was pretty good as well.  Tobey McGuire was also a great choice as actor.

And by the looks of the trailer, it seems to follow the Batman model: focus on the dead parents (which were barely mentioned in the movies or comics when I was growing up), focus on the science, and get grittier.

But, I’ll get past it, because once I set all that aside, this movie looks promising!

Tuesday, February 07, 2012

How much is winning a Cy Young worth to a young pitcher?

Oh, about 4MM$ a year for the next 3-4 years.  Kershaw signed two years, to forego his first two arb years.  How much would he have gotten, had he had the same performance, but not won the Cy?

I think Matt would be in a better position to answer that.  However, if we look at the big four (Felix, Verlander, JJ, Weaver), we see a pretty typical pattern for one year deals, and what we should consider as the upper limit: 3-4MM$ the first year, 7-8MM$ the second year, 13-14MM$ the third year, and 20MM$ if it gets there for a 4th year.

As Dave reminds us, young guys sign away their early years as well.  Lincecum got 2/23 for his first two years, but he came off TWO Cy Youngs. 

Cole Hamels signed 3/20.5, which is a discount had he gone year-to-year like the Big 4 did (25MM$ or so for their first 3 years).

Kershaw just signed 2/19, and one would think that if he signed for a third year, he’d have gotten another 15MM$ or so to come in at 3/34.  Signing multi-year also means you are getting a discount, so, that probably means he’d have expected to get say 38MM$ had he gone year-to-year, compared to the 25MM$ the Big 4 got.  That’s a 13MM$ bonus for Kershaw over the Big 4 for those 3 years.  Or 15MM$ bonus over Hamels.

Basically, a Cy Young really takes the sting out of arbitration, and accelerates a pitcher’s service clock by one year.  What Lincecum did for his year 1 and year 2 arb deal (of 2/23) makes more sense to think of it for his year 2 and year 3 arb deal.  (The Big 4 for example were paid about 21MM$ going year-to-year.) Same thing with Kershaw, who signed for 2/19, which is very much in line with thinking that his Cy Young accelerated his service time, since it matches to the Big 4’s year 2 and year 3.

When to purposefully lose the lead

According to the incomparable Brian Burke, the Pats should have given up the go-ahead TD at the two-minute warning.

The smartest play of all would’ve been for Belichick to have allowed the touchdown even earlier. The Patriots certainly could have done so on the play prior to Bradshaw’s touchdown run, when he was stopped for a one-yard gain, forcing New England to burn its second timeout. In fact, they probably should have allowed a touchdown as early as the two-minute warning. That’s the point at which the Win Probability of receiving a kickoff down by four or six points (0.23) exceeds the Win Probability of trying to stop the Giants from bleeding the clock dry (0.2). The Patriots would have had almost two minutes, two timeouts, and all four downs available to get a touchdown and steal the win.

Basically, every time out has a certain win value, every second lost has a certain win value, every yard lost has a certain win value.  And Brian is saying that the Pats would have maximized their chances of winning by allowing the TD to happen at the two minute warning.

This is exactly what win expectancy charts (and to a lesser extent, run or point expectancy charts) should be used for.

The will of the people?

As many of you know, the Unconstitutionality of Prop 8 (banning gay marriage) in California was upheld today by a 3-person panel of the 9th Circuit.

A leading proponent of Prop 8 said this:

“We are not surprised that this Hollywood-orchestrated attack on marriage – tried in San Francisco – turned out this way. But we are confident that the expressed will of the American people in favor of marriage will be upheld at the Supreme Court,” he said.

California voters passed Proposition 8 with 52 percent of the vote in November 2008, five months after the state Supreme Court legalized same-sex marriage by striking down a pair of laws that had limited marriage to a man and a woman.

Putting aside the (important) issue of a majority being able to dictate the rights or lack thereof of a minority, it really rankles me when any group uses the “will or mandate of the people” argument when 50-something percent vote for against something or someone. I mean, isn’t 52/48 essentially a split? That is almost as far from “a mandate” or “will of the people” as you can get!

And the person quoted above says, “the American people.” Of course this was a California vote, not a national one. However, let’s talk about “Americans” since this guy did in fact say, “the will of the American people.”

According to this Gallup poll:

http://www.gallup.com/poll/147662/first-time-majority-americans-favor-legal-gay-marriage.aspx

53% versus 45% of Americans favors gay marriage!  So this guy, in addition to incorrectly (and irrelevantly, since Prop 8 is a state issue) talking about the “will of the American people,” is full of crap as far as his facts are concerned, at least according to the poll I referenced.

For Your Soul

Poz has an interesting game.  It’s clear that if you pick baseball, you should pick a pitcher.  That’s because, for one game, a pitcher has far more impact than any other player.  That makes the choice quite simple: Pedro Martinez, 1999 or 2000.  Or does it?  After all, Pedro did not pitch the full season, and do you want to go into a winner-take-all game with a chance that this guy won’t pitch?  So, Dwight Gooden 1985 might be the better choice.

For hockey, goalies don’t have the same one-game impact like pitchers do.  Gretzky is the obvious choice, the Devil will counter with the Broad Street Bullies, naturally.  But again, it’s a player in a team sport, and as we know they are just one part of the game.  Basketball would be better of course, but still not enough.

So, you have no choice but to go with a one-on-one situation.  And that means choosing Tiger at his peak.

Golfers “playing through”

When do you let golfers play through?  How much do you let it escalate?  And check out the ESPN comments (sort by MOST LIKED), as there are tremendously funny comments in there.  ESPN readers really outdid themselves on that one (too early though?).

Bronx Parking

I didn’t realize how few people actually parked when going to a Yankees game.

Monday, February 06, 2012

Pitch Diversity Index

I love the idea, off the bat.

But, I don’t like the question.  The peripherals already includes the diversity.  A better thing to ask is if a pitcher with more diversity performs better than someone with less diversity, if you control for the fact they have the same fastball speed.  (We will end up with a selection bias at the low-speed end, but not the high-speed.  Well, we might, because we don’t have a measure for strike zone location, which, really, is what pitching is all about.)

Anyway, would love to see more work on this.

When is a life entity considered a person?

Position-switchers

ESPN has the list of story lines.

Interview with Howard Baldwin

Great stuff from Timo.

Could this ad run in the U.S.?

(See below.)

Read More

Sunday, February 05, 2012

Forecaster’s Challenge: 2012?

I’ve been running the Scouting Report and the Community Survey balloting because I firmly believe in their value.  The scouting report gives “shape” to a player’s fielding description, while I’ll put up the Community up against any one’s playing time forecasts.

But for the Forecaster’s Challenge… well, I’m wondering if I’ve said everything that needs to be said about that.  Am I doing it simply to keep doing it, or am I learning something new.  I don’t think I’m learning something new any longer.

What say you?

Super Simple Baseball Game

Great job from Rally:

http://www.baseballprojection.com/ssbbg.html

Flash mob for a SuperBowl commercial

Really cool stuff.

(See below.)

Read More

Is Nate Silver alot more certain than he lets on?

I’ve been following Nate’s mean forecasts for the five primaries so far.  So far, he’s made 25 predictions over those 5 primaries (and obviously, they are interdependent).  His worst forecast result was Santorum in Iowa, where he gave him a mean forecast of 19.1, and he ended up at 24.6, for a difference of 5.5 points.  His average error over those 25 forecasts is 2.34 points, with one standard deviation being 2.71 points.

However, his posted uncertainty level is much higher than that.  Let’s take Mitt in Iowa as an example.  He gave him a mean forecast of 24.5, with a range of 13 to 32 (a range of 19 points).  In another article, he notes that his range is the 5th and 95th percentiles.  Those levels are reached at the +/-1.645 standard deviations (or a range of 3.29 standard deviations).  This means that one standard deviation for Romney is 5.8 points.

So, I calculated it for all 25 forecasts, and one standard deviation averaged 4.6 points as Nate’s uncertainty level.  However, as I noted earlier, the actual observed standard deviation was 2.71 points.  This means that Nate’s uncertainty level is 4.6/2.71 too wide, or 1.7 times too wide.

Now, either he made a calculation error of his historical data (making the width of his uncertainty level almost twice what it should have been), or this year, things simply worked out alot closer to the mean than expected, just by luck (after all, we only have 25 data points).

Here’s the data for those who want to take a crack at it:

Read More

Saturday, February 04, 2012

Complete PITCHf/x classification

Great stuff available at BrooksBaseball.

It’s interesting to me when we classify things as “Fastball”, when obviously a Moyer fastball and a Strasburg fastball moves differently.  First, there’s the difference in speed (16mph difference), which means that gravity has more time to impact a Moyer fastball (you can see this in the charts a bit further down).  The end result is that a Moyer fastball sinks more than a Strasburg sinker.

But, then there’s also those pitchers that throw alot of fastball AND sinkers, and so, they will try to differentiate them enough so they can get value from the two pitches.  For example, Strasburg doesn’t throw a sinker alot, but when he does, the movement is a couple of inches different from his fastball.  It’s also two mph slower.  Moyer on the other hand throws his sinker and fastball at the same speed, but it has significantly different movement on the pitches: 7 inches on the horizontal and 5 inches on the vertical.

And it’s not just Moyer.  CC also throws fastballs and sinkers, and the speed difference is only 1 mph, the horizontal is 6 inches and the vertical is 3.  Felix throws his sinker and fastball at the same speed, with 7 inches of difference on the horizontal and 3 on the vertical.

Verlander is like Strasburg, in that he rarely throws a sinker, and he gets 3 inches on the horizontal, and 4 on the vertical, at the same speed (so, a bit more movement differentiation than Strasburg, but not as much as CC/Felix).  And maybe the reason is that his fastball already gets so much movement to begin with.  Verlander’s fastball on the horizontal moves as much as Felix’s sinker on the horizontal.  So, he really doesn’t have much room to maneuver there.  One might think that Verlander can improve his repertoire by following a Felix/CC model of limiting movement on the fastball, to differentiate it more to the sinker.  But, it’s hard to argue with the success of Verlander/Strasburg to think that they can actually pitch better.

Anyway, all to say that since a Verlander/Strasburg fastball really lies half-way between a Felix/CC fastball/sinker, that it’s not really helpful to simply classify their pitches as “fastball”.  It’s really a Verlander-fastball and a CC-sinker and a CC-fastball.

I think it’s perfectly fine that when you treat the pitcher as his own universe, that we stick with the standard classifications.  But, when we combine pitchers, it’s may be more helpful to distinguish them based on their movement and speed, rather than how they clustered for a particular pitcher.

Bill James Baseball IQ app

I downloaded Bill James Baseball IQ onto my iphone (I don’t think it is available on droid phones, but I’m not sure). Here is the web site for the app on Acta Sports:

http://www.actasports.com/titles/bill_james_baseball_iq_app/

It is pretty cool. You can read a description and see some screen captures on the above site, but basically it allows you to see heat maps and color maps of batters and pitchers (in all combinations, counts, situations, etc.) for K zone, batted balls, pitch type, etc.

Best of all, the app is free! Seems to me that they could have charged for this one, but I know nothing about the best way to make money from apps. It also seems like they could use these graphics more often on TV broadcasts.

Anyway, give it a try and see what you think…