Stats Are Evil

[rluzinski]

Why oh why can the burden of proof never be on the guy who makes his claims based on casual observation? How come that dude never has to do any research or number crunching? Probably because there's nothing there to analyze, as he has taken the lazy position and says "I just know".

The problem with relying on such "evidence" is that a) you are not intently watching every at-bat of every game all season and b) certain events will be more salient and stand out in your mind moreso than others. Sometimes we remember players for one event and base our opinion of them on that, like a guy who threw a no-hitter, or a crappy utilityman who happened to score a couple runs during the world series and parlayed that into being overpaid for several seasons.

 

I've attempted to say this a number of times but never have been able to describe it as elequently as mothershipconnection just did.

 

Trying to make assumptions based on observation of large samples is impossible. Human bias and bad memory make it impossible. Watch 100 AB in a row and unless you take copious notes (you know, record stats) you wouldn't be able to tell me if they guy batted .250 or .300. You wouldn't even know how many HR's he hit unless you counted them. If I asked you, "Is he a patient hitter, can he hit the outside pitches, does he walk?" you would probably be hard pressed to tell me. It's 100 AB! You probably would use the first 20 AB trying to get a sense of the player and then validate those assumptions for the next 80. It's human nature.

 

Here's an example. I just assumed Hall was a very impatient hitter last year (and that part of his success early this year was a result of taking more pitches). I mean, we all remember Hall hacking to death last year at anything, right? It's obvious. Anyone who watched the games last year knew that. WRONG.

 

After I tried to back up my observation with actual facts (because that's what you DO when you want to present something to the public), I found this:

 

Pitches / PA

 

Hall, 04: 3.80

Hall, 05: 4.19

NL Ave: 3.74

 

While Hall HAS improved his P/PA this year, he was actually ABOVE league average last year. How could I have been wrong, though? This one was a no-brainer. I mean, I watched most of the games in 2004. Hall stunk.

 

It's simple. I don't like Hall and I couldn't have told you if he was a patient hitter even if I wasn't biased. The difference between 3.5 P/PA and 4.0 P/Pa is not something you can keep track of in your head.

 

I junked the whole article. I could prove nothing.

 

I'm not trying to preach here or think I'm better than anyone else. I'm just describing a limitation everyone has (unless you have photographic memory or whatever). Stats can be abused but they are the foundation to making intelligent observations.

[danzig6767]

You're never going to change the mind of the majority of folks who think some are to stat obssessed.

 

Part of it i think you're right is some think just by watching that's all they need to judge what makes a player good or bad.Some of it is people who think there are hidden things in the game that stats don't give enough representation to,their minds won't be changed on that.Lastly there are alot of people who simply don't care enough to look deeply into the many many stats out their today.

 

I fall partly into the 3rd catagory.I fully understand in such an individual sport like baseball,stats paint a pretty good picture of a players ability.With that said,beyond the basics of OBP/OPS/ERA/WHIP and maybe a few others,i just dont really care.

 

I watch baseball to be entertained and am not really interested if there is a 5.257 percent better chance of scoring if we dont bunt in a given situation.I don't care what anyones range factor is.I dont care where we rank in pitches per at bat.I dont care what Lee's VORP is,but i do care what his OPS is.I dont care how many win shares Jenkins has,but i pay attention to his OBP and SLG.I dont care much what Capuano's percentage of line drives caught are,but i will check his WHIP or K/BB ratio.

[rluzinski]

I understand that lot's of things that stat guys proclaim as fact can and should be argued. Some of the conclusions that mave been made in SABRland may have been made in haste. I don't want this thread to be about that. I'm talking about guys who are convinced that casual observation is actually BETTER than stats as a whole. I would argue that casual observation is a very limited tool for the reasons I outlined above. Any assumptions made about the nature of the game (or player) should be backed up by stats.

 

If you say a guy is a patient hitter and he actually takes an average of 2 pitches an AB, your assumption was wrong. It doesn't matter if you watched the game and SAW he was patient because he wasn't.

 

I know many arguments are centered around topics that are debatable (the merits of a steal, for instance) but more and more facts are being debated as opinion. That's what I don't get.

[scoop53714]

Stats rule and are a HUGE part of baseball and in fact make baseball the most intriguing of all major sports. However, even Bill James acknowledges that there are plenty of things within the game that are immeasurable and often misunderstood or even misleading when looking at the stats. All one can do is use stats to confirm what the eye sees. I knew Hardy would hit dispite the fact that he was hitting .180 for a long time. there wasnt a statistic to be found that would indicate that Hardy would hit and hit a lot. Low and behold, post all star break OPS .878. Find me a stat that would have predicted that.

 

I LOVE stats and grew up spinning the spinner on my ALL-STAR baseball game and recording stats for hours on end but there is a reason baseball uses scouts. The eye can be a very useful tool in determining how the stats may be misleading, overestating or understating a players real ability.

 

Plus there are very skilled stat magicians in the world that could cherry pick numbers and prove just about anything they want to prove.

 

PS....anytime there is a stat called "Line drive percentage" that doesnt have anything to do with how often a player hits a line drive, stats will be debated.

 

PS#2 ....if statistics where the end all of sports information....Las Vegas would surely have gone bankrupt decades ago. You can NOT simply punch numbers into a computer and predict with certainty how things will unfold. Once again, I am a stat lover but things go deeper than the numbers.

[rwa12]

Would tend to agree with Scoop...neither side of the house should be discounted. I enjoy watching a baseball game for what it is, I feel that I have watched enough baseball in my lifetime to form educated opinions. I do not care if my opinions clash with statistical analysis, it does not matter to me in the slightest. That is why I do not read or post in the Statistical Analysis forum. I will never discount the utility of stats, they are obviously a very important facet of the game. They do not add much to my enjoyment of the game however, and I think you can have an interesting discussion without having to prove someone EXACTLY right, or EXACTLY wrong.

[BrewCityUnit]

i love stats too but think OPS can be inflated and hitting with runners on and one or two outs should count more than hitting a double with two outs and nobody on or launching a 440 foot homerun in a 11-2 loss. My favorite stats about how good a hitter is in MY opinion are .AVG and .OBP I think .OBP can be a little flawed just like every other stat by where you hit in the line-up, reaching on errors, etc.

 

You can't weigh situational hitting and sometimes can't even describe it. That is why you just have to watch the game. Lets track Braynans at bats this weekend against Houston and rate them on 1-10 and see what kind of weekend he has. Don't get me wrong, I want him to suceed because it would help the team, I just don't have the confidence it will happen.

[scoop53714]

Quote:

---

Lets track Braynans at bats this weekend against Houston and rate them on 1-10 and see what kind of weekend he has.

---

 

I long for a stat that displays how often the player hits the ball HARD. I was excited to find out that there was a stat called "line drive percentage". I thought this was going to be GREAT. I quickly read how this stat was calculated and furious that it had NOTHING to do with how often a player can be expected to hit a linedrive. In FACT it does just the opposite. It rewards players for making the least amount of contact. The point is, all stats need to be looked at with a keen eye.

 

I wish there were a stat that gauged each AB somehow, much like the Quality Start stat, but there doesnt seem to be.

[danzig6767]

I wish there were a stat that gauged each AB somehow

 

Over 400-500 at bats though,in the vast vast majority of cases things will even out for a player.In that many at bats it would take an incredible amount of bad luck to just make a bunch more unlucky outs than other hitters with a simular amount of at bats.

 

Baseball is such a long season that generally it accurately paints a good picture of what players and teams performed well.It's pretty hard in alot of at bats or in 162 games for pretenders to shine because of mainly good fortune or just the opposite.

[BillScottCanRake]

I have a story from my childhood about observations and memory.

 

For years, I told a story about how when my brother was five years old, he was sleep walking and peed in our closet all over our brand new Christmas toys. It was pretty funny story, until my mom corrected me and told me I was the one who did the sleep walking. It wasn't until my dad confirmed mom's side of the story that I believed them, because I had told that story so often and was entirely convinced I had it right.

 

Needless to say I don't tell that story nearly as often as I used to. http://forum.brewerfan.net/images/smilies/smile.gif

 

 

i love stats too but think OPS can be inflated and hitting with runners on and one or two outs should count more than hitting a double with two outs and nobody on or launching a 440 foot homerun in a 11-2 loss

 

A few things to consider in this regard.

 

- You can't have a 9 run come from behind victory without scoring that first run.

 

Everyone remembers Bill Hall squeezing home the winning run to cap that rally from 9 down to the Reds last year. But, quick, and without looking it up, who scored the winning run? Who drove in the first two runs of that rally?

 

While those types of games are obviously few and far between, I think it underscores the fact that "garbage time" stats aren't always what they'd appear. Espeically in baseball, when you aren't fighting a clock, and you can hit all day as long as you don't record three outs.

 

- A two out double with no one on sets up a two out base hit to score a run.

 

Again, that double is useless unless the next hitter or hitters do their job, but you can't have a two out run scoring single unless you put a runner or two on base first. The driver-inner can't be clutch without the scorer getting on base first.

 

Human nature is such that everyone will remember the RBI guy, but few will remember the guy who kept the inning alive with a two out double. Nor will the guy who scored get credit for a "clutch" hit, because there isn't really a stat that tracks that sort of thing.

 

That's why I am not really impressed with stats like BA w/RISP. And it's also why I'm don't like to use my own personal observations as to who is or isn't clutch.

 

I'm not staying that stats are everything, because stats alone can tell a story that skewed, or worse. Look no futher than the pitches per plate appearance stat for Bill Hall. I wonder how his P/PA stat was due to him striking out SO often last year. You've got to see three pitches to K, and Hall did that enough last year that I think it probably inflated his P/PA total.

 

But observations are just a prone to being skewed, espeically with our memory being as selective as it can be. And I know from experience how selective memories can be!

 

That's why using statistics to support obersvations, and using stats to locate information that can be confirmed by observation, seems the best way to go.

 

For those interested in the trivia posted above:

 

- Trent Durrington scored from third

- Overbay singled home Spivey from third in the 4th and doubled home Jenkins in the 6th

 

And, yes, I had to look it up.

[huntsvillefan]

I had this problem with Nelson Cruz earlier in the year. While you were all looking at his numbers and raving, I was looking at a guy who would come to the plate with the game on the line and take 3 ugly hacks at the ball striking out with the tying and winning runs on base. He looked so unconcerned that I started to hate him.

If he had taken it to a 3-2 count fouled a couple off and holed out on a long fly, I'd have been much more impressed - BUT the result would have been exactly the same.

Once I looked at the numbers, I had to swallow my prejudices and admit that he was doing an absolutely awesome job for the team. The fact that I didn't enjoy his particular style of batting and that I had been able to feed my own prejudice by picking out a couple of key ugly at bats, just didn't come close to outweighing the overwhelming evidence that the numbers provided.

Once you have a look at how closely OPS (or OXS) correalates with actual runs scored, it becomes darn near impossible to argue against it as the most significant batting stat. The only way an argument in favor of RISP stats or other stats could be convincing, would be if they were shown to tie in even more closely with runs scored. Until that happens every other batting stat in an argument is going to be overcome with a reference to OPS.

Now that doesn't mean that personal observation is irrelevant. If, instead of being down on Cruz for looking ugly when he struck out , I could have honestly said - He hasn't seen many breaking balls and everytime he faces a guy with a good curve he doesn't produce. Then my personal observation might have been worth something and might have been an indicator that his OPS would suffer when he came up against better pitchers. But saying I thought he had a bad AB is really pretty meaningless apart from an aesthetic point of view.

Now, once we get to pitching and defense stats, I'm a lot more sceptical, simply because no-one has tied them closely to runs allowed, so it all does become much more a matter of opinion. But even then, if I want to argue against them, I have to state a reason why I find them flawed, not just state a counter opinion based on nothing but my own intuition.

[rluzinski]

I long for a stat that displays how often the player hits the ball HARD. I was excited to find out that there was a stat called "line drive percentage".

 

I assume you are refering to a hard grounder not being a line drive? Or do you mean HR's not being included? You can just add those in? K's not subtracting from it? Take em out then! Ditto with BB if you want.

 

The key to stats is understanding them and molding them for your needs. Until they have lazers measuring the trajectory and velocity of all balls hit in play, they will always be an estimate of a batter's true ability. As long as you don't lose site of that they are very usefull.

[rluzinski]

I'd like to add that batting average IS a stat that can vary based on luck alot. It would be nice to strip away that luck by knowing the nature of the balls a player hits. Just because a liner is hit right at a guy doesn't make it bad and just because a weak grounder finds a hole doesn't make it a clutch AB. It's just luck.

 

OPS is alot harder to fake. K rates and BB rates are are well. It's just important know what stats are more prone to random variation than others.

[1992casey]

huntsvillefan's post is so good that I don't know which part to highlight and quote.

 

I've valued his personal observations on these forums as he's one of a small number of us who gets to see many of the players we talk about. The fact that he's willing to check his observations against available stats tells me that his observations are that much more credible.

[scoop53714]

Quote:

---

I assume you are refering to a hard grounder not being a line drive? Or do you mean HR's not being included? You can just add those in? K's not subtracting from it? Take em out then! Ditto with BB if you want.

---

 

Not even close. My problem with the stat is this. It has nothing to do with how often a player hits a line drive. It only refers to the number of balls put in play that are considered line drives. Thats fine and dandy but it doesnt tell us ANYTHING. The stat gives the impression that guys who strike out all time time hit more line drives then guys who put the ball in play. Its total garbage.

 

I will try to keep this brief because long posts sometimes get ignored.

 

Jones ....10 ab....8k....1 line drive. 2 balls put in play, one line drive for a Line drive average of .500.

 

Anderson 10 ab 2 k 4 line drives. 8 balls put in play, 4 line drives for an average of .500.

 

Clearly Anderson hits 2 times as many line drives in 10 abs as Jones but they have the same Line Drive Average because it only counts balls put in play.

 

A Player could Fan 99 times in a row and hit one line drive and have a line drive average of 1.000. Thats stupid.

 

Jones has

[colbyjack]

huntsvillefan's post is so good that I don't know which part to highlight and quote.

 

Indeed, thanks as always huntsvillefan for your point of view.

[1992casey]

---

It only refers to the number of balls put in play that are considered line drives. Thats fine and dandy but it doesnt tell us ANYTHING. The stat gives the impression that guys who strike out all time time hit more line drives then guys who put the ball in play. Its total garbage.

---

I think what you're describing is essentially a piece of raw data. Compiling any raw data we can makes sense to me. How it's applied would be a totally different animal.

 

This might be a good topic for its own thread on the statistical analysis forum. It's already popped up a little bit over there. I see the topic of balls in play having value for future reference, but I also see it eventually getting lost in this thread.

[J Brew Crew1]

Heck I'll dive in.

 

I enjoy some of the stat related posts and I've learned a lot. It has forced me to think about a player or an issue from a different angle, which is a good thing. Some of Rluzinski posts have been very eye opening, for example. And I agree that sometimes your eyes lie to you about a player. I do form an opinion about a player and then use certain at bats to "prove" my theory. Wes Helms is having a pretty good year. I can now admit that. For 4 months every junk AB he had "proved" my theory that he was junk. Low and behold I looked at a bunch of stats and he is having a good season. I was looking at the 7 out of 10 outs he made not the 3 out of 10 hits.

 

What I hate is when someone says that we need to upgrade at position X. A good discussion ensues, then some clown says "how could you say he's bad defensively, his range factor is...." Or "he's a good hitter because his VORP is ..." There was a good discusion about contracts, age, players in the minors, amount of money that's available, free agents available and someone throws VORP at me. That's where I tune out big time. I understand VORP and Range Factor and the rest and think they are useful but not the end all. I personally think that stats do not do justice to defense. They carry the same weight with me as Gold Gloves do. Which is not much.

 

I do like OPS and OXS. What I really like is the quick 300 400 500 Ave Obp Slg rule. One glance and I can tell how good a guy is.

 

I also beleive clutch exists. RLuzinski's great article showed that you couldn't prove it with the stats. But I think that there are certain players you want to have at the plate if your year is on the line. Just like you would want a certain QB under center, or a certain golfer to make a putt. Can you prove that one person has it and one person doesn't? Nope. But like the Supreme Court said about porn, you know it when you see it. You can manufacture a stat to try to show it (eg 4th quarter comebacks in football) but they lack the rigor to be relevant. I guess that's my biggest complaint about the stat guys, rigor and relevance. If you have those, like you do with OPS or OXS, then you have something. If not, well...

 

I just reread this post and its all over the place and a little harsh. I hope not to offend anyone, that much http://forum.brewerfan.net/images/smilies/smile.gif . But as a long time lurker here I have learned a bunch and finally just wanted to throw my two cents in.

 

All the best

 

J

[rluzinski]

It only refers to the number of balls put in play that are considered line drives.

 

Scoop, that's EXACTLY what I was suggesting. I would never use the LD% as a metric to suggest who was a better player. It's just a raw stat that can be used any way you want.

 

Lyle Overbay led the entire league last year with a LD% of 28%. He had 129 hit's subjectively characterized as line drives (anything subjective dimishes it's worth of course). So if you want to know his LD% with respect to PA:

 

128/668 = 19.1%

 

My choice would be to subtract BB and HBP (shouldn't be hurt by not having a chance to hit a LD):

 

128/(668-81-2) = 21.9%

 

The stat gives the impression that guys who strike out all time time hit more line drives then guys who put the ball in play.

 

It doesn't to anyone who knows that the stat is with regard to BIPs. I wouldn't use any stat I didn't understand, however.

[1992casey]

J Brew Crew1,

 

Your post isn't harsh at all! I don't think it's all 'over the place', either.

 

I think statheads (geeks) should be perfectly willing to acknowledge the limitations of the stats they use. (I use the terms 'statheads' and 'geeks' affectionately, of course! http://forum.brewerfan.net/images/smilies/happy.gif)

 

As an example, I think Range Factor has a very limited useful value. The few times I choose to mention it, I take pains to explain why I normally don't like to use it but why it might apply in the given narrow situation.

 

I think it's generally acknowledged that OPS is a safe stat to use and really doesn't require justification. Usually, I'd submit that the argument around OPS would apply to the splits someone opts to present.

 

Clutch is a toughie. Bill James isn't ready to write it off, even though studies generally can't demonstrate its existence. I believe part of the problem is that James is seeing a flawed approach in many studies?his own and others?and would like to see them redone using different parameters.

 

I look at clutch this way. I wouldn't choose to acquire a player based on RISP stats, but I'd certainly look to maximize those numbers by balancing out my team so effective platoons, pinch hitting situations, and switches can occur.

[bjamesw22]

Some of Rluzinski posts have been very eye opening, for example.

 

and

 

I also beleive clutch exists. RLuzinski's great article showed that you couldn't prove it with the stats.

 

Ok Russ, you can stop posting under a different name now. http://forum.brewerfan.net/images/smilies/happy.gif

 

No, but really, I agree that stats are eye opening. But this is one of the statements I absolutely do not agree with :

Quote:

---

What I really like is the quick 300 400 500 Ave Obp Slg rule. One glance and I can tell how good a guy is.

---

using the stats as gospel to make determinations about players without looking into how the stat was composed by that player. OPS doesn't tell you everything. Does it generally tell you how good a player is? Yes. Does it tell you that a .875 OPS player is a better hitter than a .850 OPS player? No. Just as the eyes can decieve a fan, the stats can also decieve at first glance in the same way.

[scoop53714]

Clutch exists in my opinion. It may come done to biological reactions. Fight or flight. Release of adrenaline into the body. All factors that may help a player in a stressfull situation. Clutch may also be looked at backwards. Not completly messing up under pressure may be just as good as totally thriving under pressure. Perhaps look at a ground ball to short with 2 outs in the ninth in game 7. Is it clutch to pick the ball up and throw the runner out? Not really but an anti clutch move would be to tighten up and wing it into the 4th row. If there is anti clutch there must be clutch. The universe is made up of opposites and therefore it must exist.

[Brian the Automator]

Clutch exists, for the reason(s) scoop mentions, but there just aren't that many people with a large enough sample size to show that a large number of players are consistently "clutch".... so while clutch players are all good in theory, I don't think there it comes into play a whole lot when teams are looking at what players to acquire in trades or via free agency.

 

There certainly are things that factor into offense beside OPS, but it seems like in the majority of cases, a players' OPS will be a function of all those little things put together (excluding defense, baserunning, and leadership of course).

[1992casey]

---

Does it generally tell you how good a player is? Yes. Does it tell you that a .875 OPS player is a better hitter than a .850 OPS player? No. Just as the eyes can decieve a fan, the stats can also decieve at first glance in the same way.

---

When we see a 25 point gap in OPS, it's obviously time to look at other factors. The first thing I'd probably do is break it down into OBP and SLG. If those are relatively equal, I may decide not to care. http://forum.brewerfan.net/images/smilies/smile.gif Or I may look into other aspects of a batter's abilities that may be irrelevant if the difference between two players' OPSs were a lot greater.

 

The obvious limitation of any rate stat (average, like BA, OPS, OBP, ERA) is that it doesn't make playing time apparent. Of course, counting stats (examples being hits, homers, RBI, Ks) have that limitation, too.

[mothershipconnection]

But observations are just a prone to being skewed, espeically with our memory being as selective as it can be. And I know from experience how selective memories can be!

This is one of the contentions I have always made, as I can draw from my own experiences. When I was younger, my dad took us to a game, maybe 1988. Tom Filer was pitching for the crew that day, and I can't remember the opponent, but he absolutely had them locked down. From then on, I thought Filer was the bomb, and nobody could tell me otherwise: forget the career ERA+ of 92. To me, he was the best pitcher in the league, because I saw him one day when he was absolutely dominant. With my brother, it was Rick Dempsey. I think this was maybe 1991 or 1992 when we went down and saw a game, and it just happened to be maybe the only big offensive game Dempsey had as a Brewer. The rest of the games didn't matter though, as Dempsey became my brother's favorite based on that one game: the bad games were just a fluke,

[rluzinski]

I wish I could say that I haven't overstates the importance of certain stats or studies, but I'm sure I am. All I can say is it's the reason I've done as much studying on the subject of statistical study of baseball as I have. I want to make sure that if I write about a subject it's because I have done enough studying to ensure it has merit.