Thursday, June 10, 2010

The Significance of Strasburg's Debut

There are a couple of easy mistakes to make in forecasting that I often warn about:

- Never be fooled by a small sample.
- A great outcome that gets our attention is not a randomly selected sample.

I'm about to make a couple of exceptions for good reason and I will explain why.  A single game performance can be so outstanding that it is extremely unlikely that it can be accomplished by the vast majority of players.  Also, even a seemingly small sample can be large enough to tell us something in the rarest of cases.

Let me give you an example: Suppose we are standing at a golf course waiting our turn to golf on the first hole.  We are observing a player ahead of us whom we know absolutely nothing about other than what we can visually observe.  He reaches into his bag, pulls out his driver on a hole that's over 300 yards away, takes a practice swing and then confidently swings and we watch as the ball magically soars, lands on the green and rolls right into the hole for a hole in one.

Believe it or not, that sample size of just one means something and tells us quite a bit.  We can't be positive but the chances are extremely high that the golfer we've just seen did not just take the first swing of his life.  It's possible but unlikely.  He may not be a great or a pro or even a top amateur but we've at least narrowed the likely range of skill that he has by observing a single swing.

That the hole in one got our attention is dangerous... We noticed the player because he did something extraordinary the first time we saw him.  Remember Karl "Tuffy" Rhodes hitting 3 home runs on opening day a number of years back?  Better yet, how about Mark Whiten's 4 home run game?  They certainly got our attention and if we were to examine the odds of players doing what these players did, they would be low and we could easily make false conclusions about their ability.

But there is a point where someone does something that is so unlikely that it is more likely that they have unusual ability than unusual luck.  That brings me to Stephen Strasburg's major league debut.

Not only did the debut live up to the hype but it exceeded it.  Strasburg didn't just look fantastic and from a non-statistical perspective, his stuff was electric.  He struck out 14 batters of the 24 he faced but in doing so, walked a grand total of none.  That's extraordinary.  Using tools available at (a site I highly recommend and which should be a frequent stop on every reader's Internet browsing), I found a grand total of 20 games since 1968 where a pitcher struck out at least 14 batters without walking a batter.  The first few names since 1968 were Sam McDowell, Luis Tiant, Vida Blue and Frank Tanana, all of whom had had excellent careers.  In the 1980s, the names were Dwight Gooden (twice), Mark Langston and Sid Fernandez, again all pitchers who had strong careers.  Since 1990, the names are those you would expect to see: Pedro Martinez (twice), Randy Johnson (twice), Roger Clemens, Curt Schilling, John Smoltz, Johan Santana and Mike Mussina make the list.  Then there are a couple who may not have had the careers of these others but still managed to accomplish the feat in Eric Bedard and Mark Prior.  That's it since 1968.

But again, we run into the problem of singling out something that got our attention rather than a random sample of a game.  So, let's consider this from a different angle.  What if we could simulate a large number of random games made up of exactly 24 batters faced?  We could try different theoretical levels of strikeout ability and see how often our imaginary pitcher strikes out at least 14 batters.

Let's start with the average National League pitcher from 2009, who would strike out about 18% of the batters he faced.  With our ability to run high speed simulations of blocks of 24 batters faced, I ran 10 million of these and at the end of 10 million "games" the average NL pitcher from 2009 struck out at least 14 out of 24 batters in 149 of the 10 million games or about 1 in every 67,114 games.

In other words, it's possible that Stephen Strasburg is not yet at least an average National League pitcher but based on a single start, an incredibly small sample, we can already say that it's extremely likely that at least where strikeouts are concerned, that one game demonstrates that he's already above average by NL standards.

But you probably already thought that.  Let's raise the bar.  Let's give him the theoretical ability of a pitcher who strikes out about one out every 4.5 batters he faces or about the rate Josh Beckett had in 2009.  When we ran ten million games of that sort of pitcher, the 14 strikeout threshold was met or exceeded 1,491 times or about once every 6,706 games.

Let's go even further: Let's give our imaginary simulation pitcher the ability to strike out about one out of every 3.5 batters he faces or about what Tim Lincecum did in 2009.  When we ran that simulation, again ten million times, our imaginary pitcher of Tim Lincecum's approximate strikeout skill managed to achieve the 14+ strikeouts in 24 batters faced mark 21,881 times or about one out of every 457 times.

I still hesitate if not outright reject making conclusions based on a performance that gets our attention simply because of its excellence but this isn't just one night where a pitcher struck out 14 batters and walked none.  This was the top-ranked pitching prospect in the world making his first major league appearance.  That's not necessarily a random sample and we could have easily resolved to do this sort of analysis on whatever he happened to achieve in his first game.

My suspicion is that Strasburg actually has nearly proved something beyond an acceptable degree of confidence and that is that he is an even better strikeout pitcher already than I expected he would be at this early stage and my next projection revision will reflect this.  His minor league numbers this year didn't even hint at this sort of performance and I'm expecting that both he and his team were (and maybe still will be) controlling his effort enough to not over-exert himself.

One other notion that I think deserves mentioning is that we hear some reminding us that facing the Pittsburgh Pirates did not offer Strasburg an average major league opposition.  While that may be true, the Pirates may be a low-average team (worst in the majors at .238) but they are actually not that bad at making contact, at about a 79% rate, almost exactly the current major league average, which is about 79.6% so far this season.  So, Strasburg's strikeouts achievment is not significantly lessened by the Pirates having been his opposition in his first game.

Anyway, despite everything I often write about small samples and overexcitement about huge single game performances, I believe that Strasburg's first game is that rarest of exceptions.   I don't think it proves that he's going to strike out double digits in every game but it does demonstrate a significantly superior strikeout potential than I had previously expected so early in his development.  If you didn't watch his first game, I highly recommend you tune in to watch his second start.  It's not about whether he can match what he did in game one as much as you really need to see his stuff if you enjoy the game.