Predictability in Sport: Let's Talk Stats

[Chris_Moderato]

This debate is never-ending. It's also almost always contentious, well-rounded, and interesting.

As a springboard to discussion, check out this Radio Lab segment (http://www.wnyc.org/shows/radiolab/) about predictability in sports. It's the episode titled "Stochasticity". The segment begins at the 22:00 minute mark and is only a few minutes long. They discuss basketball, but the argument can be applied to baseball as well.

Which side do you come down on, and why? Is there a way to accurately predict a player's performance over the course of a season, or do intangibles and immeasurable variables mean "anything can happen"? Do you believe in clutch hitting? What statistics are important to you? Why? Is it possible for stat-heads and traditionalists to co-exist?

[Dreamscape]

Really depends on your definition of accurate.

No, projections won't tell you exact numbers. They give you an impression of what is likely to happen based on the player's history, the position output, league history, ballpark factors, etc. The best projections systems (I tend to lean on PECOTA from Baseball Prospectus) do a very good job at giving you the most likely out of outputs, while also projecting an array of seasons that could be had and the percentile those seasons will occur.

But there is no way to accurately project a season because, yes, anything can happen. That's why we have the "luck" stats like BABIP or FIP. They tell us what our eyes might and that's how lucky the player has been. You can't project luck. Sure, you can add a bit into your projection system, but it's not quantifiable from a projection standpoint. You can't say before the season, "I think Kelly Johnson's BABIP will suffer." Just doesn't work.

As for clutch hitting, I take the wordslayer approach which I believe he borrows from Bill James. There is such thing as clutch at-bats, but not clutch hitters. What defines clutch? The ability to step up in tough situations? Well, most great hitters don't change their approach whether it's the first game or Game Seven. Jason Bay had a great article a few years ago when he was Pirate talking about how baseball and football were different. Football players need to get amped up. It makes them play better. Baseball players need to be far more even keeled. In any given year, a player's RISP can rise and fall. It's not based on being clutch one year or the next, it's based largely on sample size and the different production per year. Even if the player had two identical years, his RISP could be vastly different because of small little variables.

As a rule, hitters will hit better over their career with RISP much like the league will do so every year. This is mainly because of sacrifices coming into play along with pitchers being distracted and throwing their effectiveness out of whack with slide steps and rushing the ball to the plate.

Sure, there are the rarites that can't be explained. Players who do significantly better in RISP over their career than when no runners are on and vice versa. But these examples are not the rule, but the exceptions that proves the rule.

As for favorite stats...for hitters, I tend to focus on OPS first. Simple number, valuable number. EQA, VORP, and OPS+ are also ones I look at when quickly evaluating a player. Again, easy numbers that tell me more than AVG, OBP, SLG, or any counting numbers. If doing studies on players, I will look at batted ball stats (where is he hitting the ball, is he hitting grounders or flyouts, infield flyouts, etc.), how he is being pitched and any marginal differences from years past, ratio stats like BB%, HR/FB%, and so forth. I did this with KJ in a recent blog if anyone wants to read (should be in my sig).

For pitchers, I hardly will ever look at ERA, though ERA+ grabs my attention. With pitchers, I'm much more interested in ratio stats to begin with and comparing them to previous years. HR/9, K/9, BB/9...these are my lifeblood along with WHIP, FIP, or DERA (defensive ERA which is essentially the same as FIP). If I'm studying more, I will look at strikes, balls, pitch selection, etc.

Statheads and traditionalists do currently co-exist. I do think both have their place. Both love the game, just differently. I enjoy looking at the numbers game, trying to project. Sometimes, I'm right and sometimes I'm wrong. But it's a learning exercise.

...way too long of an entry.

[wordslayer©]

Good thoughts thus far

I love all the new stats. I wish I were smart enough to understand them.

The game has become much more of a science and that is a good thing. Like Dream said, there isn't any way that you can predict with any certainty what is going to happen, but you can certainly see trends and, if you look at the stats closely, you can, on average, make some very good assumptions sometimes.

Things I look at:

1) Hot women
2) Exotic cars

But since this is baseball related, here are the things I look at:

With a young hitter, I look at their on base percentage. Do they know how to work a count? Are they patient? If a hitter has these qualities early on, they will usually keep them for the majority of their career, and if they don't have them early on, then they probably won't ever get them.

If they don't have power, will they develop power? Overall, as a hitter ages, he will lose speed and gain power. If he shows some power at an early age, and if it isn't a ballpark effect, then he will probably add to his power as he ages. In my opinion, for young hitters, OBP is the gold standard.

Hitters aren't really that hard to assess in relation to pitchers, and this is why I hate seeing pitchers taken early in a major league draft. With a pitcher, there are WAY too many variables. It is much less risky to draft a young hitter than it is a pitcher, and why the hell so many teams take young pitchers in a major league draft is something that I will never understand.

As for pitchers.....the main thing I look at is strike outs. That's the platform you build from. And once you look at those, then you look at your ratio stats.

A pitcher, IF he has strike outs, has part of the biggest battle won. Let's look at Tim Hudson for a minute.

Personally, I wouldn't touch him with a ten foot pole, and the reason is, I don't like his strike out rates. As of now, today, his last season, I have no doubt that he can still be a decent pitcher. He walks very few and his WHIP is good. But his strike out rates are just not good enough for me to take a flyer on him, especially with him coming off of arm surgery and his age. IF Tim Hudson, in his last couple of seasons, was showing me a good strike out to innings pitched ratio, then I'd feel totally different about him.

If a 35 year old pitcher is having good strike out rates, then yeah, let's sign him to a four year deal, but if he isn't, then I'm ready to sell the farm on him. It's just too risky. Things can change too quickly. There is a reason that baseball history is littered with older pitchers who strike people out, but there ain't very many 35 year old ones that hung around who couldn't strike people out.

On a tangent now, but this is the reason that I was opposed to signing Lowe. It's way too risky. If he loses a mile an hour or two on his fastball, then he will get shelled. Pitchers like him have NO wiggle room. Strike out pitchers do. Still pisses me off that we signed him to a long term deal. If you are paying THAT kind of money, you don't put it up to risk like that.

The strike outs are the main reason that I love Vasquez. I'd sign him to a long term deal in a second, IF he was the same age as Lowe and IF he had the K/9 rates that he does now.

Gotten way off base here, but yeah, in essence, I love the advanced stats, and if you know how to interpret them, they can be as good as predicting the weather.

[absintheofmalaise]

I'm pretty much in line with both Dream and Word on this subject. I also really like using Linear Weights to help see what a player has contributed offensively. B-Ref has 2 great stats based on lwts, BtRuns and Bt Wins. I only use WPA to see how a player has done in a particular situation.
I will say that I really don't pay attention to defensive stats at all. They are so subjective, and I'm talking UZR and +/- here, that they are only good for a quick judge. The other D-stats aren't worth anything more than lining the bottom of a bird cage.
Stats and eyes easily co-exist. You really need both to make an accurate judgment about a player. Especially the eyes on defense. Quite a few teams, I'm still unsure of the Braves approach, combine both. It's when you rely too much on one or the other that you get in trouble.

[bravos4evr]

I thought I read some report last year that pointed out that while there appears to not really be such a thing as "clutch" there does appear to be certain players who arre "not clutch". There are some players who apparently really faulter in clutch situations.... I find that interesting.

On the subject at hand, I love the new stats and I think they are very good at predicting the long term production of a player. Not down to any sort of exact numbers , but if you look at a guys career OPS and it's .850 at 2b, you know he is a valuable player and as long as his year by year numbers haven't been regressing, you know he is likely to hit somewhere between his career year and his bad year. Pretty obvious stuff IMO!

[bravos4evr]

Hmmm, my mathematical little brain had an interesting thought. I wonder what the difference in performance would be if batters never saw a pitcher twice. Like if there were 162 different starting pitchers faced in a season. Then a different 162 the next season ( consider the same for relievers).

In other words, what I'm really asking is I wonder what the scouting report repeat viewing of pitchers effect is!