Julien's Faster than Light Blog
Jump to top
Tuesday, January 11, 2005
what does this stuff all mean?
i've had a number of people tell me they like my site but it's too technical for them to get anything out of it. i keep saying i don't wanna confuse people, so it's time to put my money where my mouth is. it's time for an explanation of the wal con pow system.
the wal con pow system
i came up with this idea one day to do an analysis that would measure a player's skills instead of his "value". that is, i wanted to find some numbers that could meaningfully tell me what *kind* of player we had, rather than how good a player. the idea was to use a statistical analysis that had a connection to reality. i developed three percentages, based on the ability to walk, the ability to hit for contact, and the ability to hit for power. together they are the wal con pow system.
my initial idea turned out to be a good one. the system has changed slightly over time, but has remained essentially the same. this website has been the result of my investigations of my idea, and applications of it. the idea works for pitchers as well as hitters, but we'll talk about hitters as we explain it, then talk about how it applies to pitchers.
the first thing i did was watch thousands of baseball games. i've been watching braves games since i was a little kid in the early 80's. then in 1987 i fell in love with the box score. i read every box score in the tallahassee democrat every day. i made lists of players, ranking them based on batting average, home runs, and rbi. i was fascinated by the numbers. i wanted to know who was the best.
then one day i came across the bill james historical baseball abstract. yes it all started with bill james. he wrote so beautifully, and so thoughtfully. it was a whole new world. i was a sabermetrician. i started calculating runs created in my head every time i found enough data. when i didn't have enough data, i estimated. i made up stats of my own. i red everything i could find on the internet. people were doing some serious thinking.
i discovered clay davenport and started reading baseball prospectus. i loved the idea of park-neutral statistics. i met clay one day and talked to him for three hours.
i played fantasy baseball. i discovered that players who are good in fantasy baseball are not necessarily good in real life. i found one thing in common with all the highest-rated players according to baseball prospectus: they hit the ball real hard. not all. almost all.
i began to pay attention to that when watching. also i was learning to tell what kind of pitch was thrown, and what kind of pitching strategies were used. i watched every start by greg maddux.
lots of things were happening at once that i didn't realize until later. maddux was a genius at not allowing home runs. but what was not emphasized as much was his ability to strike people out. this would be a theme.
i just took a break and watched randy johnson on letterman. randy's kind of wooden, but it was a good time. dave asked about steroids. he was wondering if they would allow *him* to throw 98 mph. randy said "not at your age."
speaking of strikeouts. so i began to notice that when sammy sosa was up against a good pitcher, he would strike out. later we saw the same thing with pedro martinez against alfonso soriano. 2003 alcs. but look at the 2001 world series. contact hitters get on base against mariano rivera. and beat him. luis gonzales. very exciting.
and this is what became clear to me as i worked with this blog: there are different types of players, and they have different types of skills. one player will be good against one guy, and bad against another. but with somebody else it will be the other way around.
this is not revolutionary. people have known this for a long time. but what i did is i found a way to express it numerically. previously all people could say was "he's a clutch hitter," or "he keeps the ball on the ground."
how did i do it? i broke it down to what happens in an at-bat. plate appearance. when the batter comes to the plate. pitcher throws a pitch. is it a good pitch or a bad pitch? the batter must decide whether to swing. that is the first skill. how good is the batter at deciding when to swing. how good is the pitcher at not walking people. the measure of the skill is walk percentage.
what if the batter does swing? what's he trying to do? make contact. what is the pitcher trying to do? make him miss. we take walks out of the equation, and we look at contact percentage. how often, when he was trying to make contact, did he succeed. or, for a pitcher, how good was he at making people miss.
you either do or you don't make contact. if you don't, you're out. if you do, you may be out, but you may get on base. what determines it? well, one way is to slap the ball on the ground and run. there are a few special players who are able to do this. one of them is ichiro suzuki. for the rest, you gotta hit the ball somewhere where a fielder can't get it. for any single at-bat, this is mainly luck. but what determines the chances? how hard you hit the ball. that's why we have power percentage. how hard did you hit the ball when you put the ball in play.
the key to all these measures is identifying the opportunity. what is the opportunity for a walk? a plate appearance. what is the opportunity for contact? an at-bat (including sacrifice flies.) what is the opportunity for power? when you make contact. thus we see that high-average hitters will invariably display contact *and* power. unless they're juan pierre.
i made the blog to test these ideas out. i can't believe how well they worked. you can predict any stat you want (batting average, on-base percentage, runs created, whatever) with wal, con, and pow. for speedsters, you gotta make an adjustment. but the other thing, the beautiful thing that had been bubbling up into my consciousness all these years, was the idea that different players are good against different players. different hitters are good against different pitchers. different pitchers are good against different hitters. it's not just how *good* the guy is, it's what *kind* of guy he is.
i'll show you how it works right here.
sammy sosa. hits the snot out of the ball. .108 .723 .161 wal con pow. league averages are .100 .810 .100. so he hits it hard, but he has trouble making contact. what's he gonna do against eric gagne? gagne had wal and con of .083 and .620, and his pow this year was .075. how do we combine them?
simple probability theory gives us a combined wal con pow of .090 .486 .123. sammy'll walk once in a while, but when he doesn't, he strikes out more than half the time. it makes sense: he already strikes out a lot, so who's he gonna strike out against? strikeout pitchers. but put a soft-tosser up there and he mashes.
great, but that's eric gagne. who can hit eric gagne? i'll tell you who. juan pierre. his .072 .949 .057 wal con pow matches up to a .060 .876 .043. almost as good. juan pierre will do the same thing against eric gagne that he does against any other pitcher: get on base.
there you have it. the wal con pow system in all its glory. i hope you keep reading so i can share my thoughts with you and illustrate them with numbers.
before i go i should explain the statline. aab w k hr 23 wal con pow avg. aab is adjusted at-bats, at-bats plus sacrifice flies. w is walks, to me that's a base-on-balls or hit-by-pitch. k is strikeouts. hr is home runs. 23 is doubles plus triples. then the percentages: wal con pow avg. avg is included so you can see the interaction between con and pow. and it'll give you a sense of a player's speed, if he's ichiro or pierre. peace.
Comments: Post a Comment