The Brewers Aren't Clutch?

[rluzinski]

The Brewer's woes with RISP have been greatly exagerated this season. When they have a game where they go 5 for 10 with RISP it's not mentioned. When they go 1 for 10 in the next game people yell, "See, they stink with RISP!" Last night, when the Brewers were doing poorly in their game w/RISP, the chants of "The Brewers are anti-clutch" was being broadcasted loud and clear. Then they quickly got 4 hits with RISP there was barely a mention. People just aren't looking to see if their perception is reality.

 

Let's look at the facts:

 AB BA Overall 4593 .255 w/RISP 1125 .239 

 

If the Brewers had gotten ONE extra hit w/RISP every EIGHT games they would have the same BA w/RISP as overall (that's 18 extra hits for the season). There is no way ANY of us can notice a hit in 8 games. Hell, I doubt any one actually notices how well we do from game to game without looking it up. It's just not a very big difference:

 

2 for 10 w/RISP <----- Worst in the League!

3 for 10 w/RISP <----- Best in the League!

 

It really comes down to perception. Many perceive the Brewers as not getting clutch hits as often as they should, so subconciously focus on those situations when they occur. While I'm not saying the Brewers have faired well in those situations (they have certainly done worse w/ RISP) the difference isn't even noticable by just watching games. The Brewers simply haven't hit for a high average in any situation.

 

Those 18 extra hits w/RISP could have swung 3 or 4 games in the Brewers favor, but it's not anything we can actualy notice through casual observation. The difference is just too small. That's why it's important to rely on stats to validate observation.

[dadofandrew]

It's entirely possible those 18 hits wouldn't have won even a single game, as most RISP situations come in games in which you score a lot of runs.

 

Also, I did this last year, same situations, same amount of runs, but opposite RISP numbers. It's a meaningless statistic.

 

I'm still amazed that OXS (of which, OPS is almost as good) is 98% accurate, and people still insist on digging up numbers a fraction as meaningful.

[DougJones43]

i agree that the brewers, overall, don't hit that bad in the "clutch." however, they seem to be incredibly inconsistent from game to game (6 for 12 one game, 1 for 13 the next game kind of thing). the overall risp batting average stat doesn't reflect this or shed much light on this.

 

i'm aware that over the course of a full season, overall numbers are generally a good indicator of success, but i'd be curious not so much about overall risp BA, but something like how many individual games have the brewers batted under .200 w/risp, and how does that stack up against the rest of the league?

[splitterpfj]

Help me out.........what is OXS?

[GenoSeligPrieb]

98% 98% 98%...

 

Just saying it in every other post doesn't make OPS 98% accurate. I can buy the notion that OPS can be a highly-accurate tool for predictive run production measures. But on a player-vs-player basis, I want to be convinced that it returns a 98% accuracy rate.....If you can categorically PROVE it's 98% right.

 

If not, you can't say it, Al...

[rluzinski]

It's entirely possible those 18 hits wouldn't have won even a single game, as most RISP situations come in games in which you score a lot of runs.

 

Well, RISP AB are probably distributed in a similiar manner that runs scored/game is. Here it is for the Brewers:

 

http://i14.photobucket.com/albums/a345/rluzinski/Crewrundist.gif

 

Since the Brewers have scored exactly 0, 1, 2, 3 or 4 runs in 56% of their games, I think it's fair to assume around half of the Brewers AB w/RISP also occured in those games. While it's possible those remaining hits would still not have given the Brewers one extra win, I think it's more likely that the remaining 9 or so hits would have resulted in 2 or 3 extra wins. The Brewers have lost 37 games by 1 or 2 runs, afterall.

[JohnBriggs12]

The Brewers offensive problems stem from lack of productivity in the 2 and 3 spots in the batting order.

 

They are 21st in OPS in the second spot (.692) and a woeful 24th in OPS in the 3rd spot (.754).

 

In fact the Brewers 7th spot has produced at a much better rate (.820) than the 3rd spot, normally the place where the best hitter on the team hits. Of course producing in the 7th spot does't lead to as many runs with the 8th and 9th hitters following.

 

Jenkins batted third through most of his bad start. Overbay has been there since early June.

 

Yost stubborness in leaving Overbay in a prime spot while he has gone almost 6 weeks with 3 total extra base hits has hurt as much as anything.

[1992casey]

OXS is on base percentage times slugging percentage.

 

Multiply it by ABs (not plate appearances) and it gives an excellent idea of the number of runs a player was responsible for over the span of those ABs.

 

Somewhere in the Statistical Analysis forum, rluzinski explains mathematically why we use ABs instead of PAs.

[rluzinski]

But on a player-vs-player basis, I want to be convinced that it returns a 98% accuracy rate.....If you can categorically PROVE it's 98% right.

 

OXS doesn't really work as well when comparing individual players. It will show you who's better but not by how much (in terms of creating runs).

 

Multiplying a player's own OBP to SLG implies the player hits himself in, which obviously isn't true. As a result, Basic Runs Created will over-estimate the marginal increase in run production a superstar gives you. That's why the newer versions of Win Shares (which uses Runs Created) places players in a league average environment first.

 

For a team it is certainly better than 95% accurate, although linear weights is a much more accurate way to estimate runs scored for a team. I'm sure Al would be more than happy to construct a spreadsheet showing that.

[BillScottCanRake]

i'm aware that over the course of a full season, overall numbers are generally a good indicator of success, but i'd be curious not so much about overall risp BA, but something like how many individual games have the brewers batted under .200 w/risp, and how does that stack up against the rest of the league?

 

I like that idea, but I think you'd be better off doing something with the number of chances in games rather than the RISPBA. Using your .200 threshold, it probably would list a lot of games where the team was 0-1, 0-2, etc, but miss a lot of games where they were 1-2. While a .500 RISPBA is nice, only having two chances in a game probably says more about how many runs the team scored than their success rate.

 

I'd like to see, for example, how many RISP/game to score 3 runs, 4 runs, 5 runs, etc.

 

I'd also like to see how many plate appearances with RISP/game to score 3 runs, 4 runs, 5 runs, etc.

 

If I had to guess (and that's all I am doing at this point), I'd say that opportunity is as important in scoring runs than is the rate at which a team converts.

[dadofandrew]

Russ may well have explained it, but all that matters is it works with AB's, not PA's.

 

Geno, if it is 98% accurate for teams, who are teams made up of...players. You can argue that it's slightly less accurate for individuals, but when you add them up, it's still 98%.

 

Splitter, OXS is better because OPS automatically weighs SLG more than OBP, because SLG is naturally higher.

 

Example:

 

Player A: .300 OBP, .400 SLG, 700 OPS, 120 OXS

Player B: .350 OBP, .350 SLG, 700 OPS, 123 OXS

 

Player B is better, but the difference is minimal, even in this example, where the difference is great. Hence, since OPS can be figured easily in your head, it's just about as good as OXS, which cannot be, for the most part.

[rluzinski]

You can argue that it's slightly less accurate for individuals, but when you add them up, it's still 98%.

 

As I tried to explain, that's simply not true. If you found the runs created for the individual players of a team and added them up, it wouldn't equal the total runs scored for a team as well as looking at the team as a whole.

[GenoSeligPrieb]

"Geno, if it is 98% accurate for teams, who are teams made up of...players. You can argue that it's slightly less accurate for individuals, but when you add them up, it's still 98%. "

 

I hear ya, Al, but as expected, I have a few issues with this explanation:

 

First of all, OBP is fueled by plate appearances and NOT at bats. As we know, walks aren't factored into at bat totals, so there's more than a 2% accuracy skew right there, right?

 

Secondly, I'll never be convinced that a formula can ever be used to 98% correctly track a player's offensive contributions if it doesn't account for bad baserunning, smart situational hitting (i.e., not grounding out when there's 1 out and the tying run on 3rd) or at least SOME nod to hitting when men are on base.

 

Also, you concede that, on a player-vs-player basis, OPS falls below the 98% accuracy rate. How far would you say, though? I'd say it drops quite a bit, even when comparing a 2004 player vs. his own 2005 output. It's that tough to factor.

 

Finally, you've got to admit that a player valuation methodology which doesn't account for fielding shouldn't be used as the ultimate statistical tool in just about every argument one makes.

 

Appreciate the push-back. Good debate, Al...

[rluzinski]

First of all, OBP is fueled by plate appearances and NOT at bats. As we know, walks aren't factored into at bat totals, so there's more than a 2% accuracy skew right there, right?

 

Walks are being factored in; it's just a result of the definition of OBP and SLG that makes you mutliply by AB. I'll paste in my explanation from the "Stats" Forum:

 

Quote:

---

S-RC = Simple or Basic Runs Created. It's Bill Jamess first crack at a run predicter. It's form was:

 

Runs = [(Hits + BB) x Total Bases] / PA......(1)

 

Since:

 

OBP ~ (Hits + BB)/PA....(2)

 

SLG = Total Bases / AB....(3)

 

Subsituting (2) and (3) into (1) gives you:

 

Runs = OBP x SLG x AB

---

[DHonks]

Statistically, is there such thing as being clutch? We've been told that there isn't. Maybe being clutch is just more a way of glorifying great players.

 

For instance, who do you want as your SS, Jeter or Neifi Perez? Which do you want batting in the 9th, down by 1?

 

So I would theorize that clutchness is perhaps simply a way of praising the guy statistically more likely to come through IN THE CLUTCH, forgetting that they are more likely to succeed in all circumstances.

 

There are some differences though. With nobody on base, would you rather have Branyan up, or Cirillo? How about bottom of the 9th, 2 outs, tying or go-ahead run is at 3rd?

 

I'm sure most would say Branyan in the first case, Cirillo in the 2nd. Why, because while Branyan is technically the better player, they have individual skills they are better at. Branyan is better at being a 1 man rally. But in the 9th, Cirillo's bat control, obp, doubles ability, etc are more likely to reach base or bring home the run. It's one of those rare situations where the strikeout is the worst possible scenario, and so in this case the better player isn't the one you want batting most likely.

 

Still, if we were to compare them to someone like Aramis Ramirez, then I'm pretty sure we'd take the vastly superior player in all circumstances. But in comparing marginal players, I'm pretty sure there are vastly different probabilities for succeeding in CLUTCH situations.

[rluzinski]

Secondly, I'll never be convinced that a formula can ever be used to 98% correctly track a player's offensive contributions if it doesn't account for bad baserunning, smart situational hitting (i.e., not grounding out when there's 1 out and the tying run on 3rd) or at least SOME nod to hitting when men are on base.

 

Looking at a whole team, all those other stats you mention don't amount for much at all:

 

http://i14.photobucket.com/albums/a345/rluzinski/teamRC.gif

 

Even the most simple runs created formula spits out a run value that's within 98% of the actual runs scored(I don't use the word accurate because it could be above or below actual runs scored, potentially). That doesn't take into account the opponent's error rate, steals, baserunning, situational hitting... nothing like that. All we need to know is how often a player gets on base (or makes an out, depending on how you look at it) and how many bases the player gets when he gets on base. That's not to say all that other stuff isn't important, just not nearly as important as some might think.

 

The table also shows that despite what Al tells you, individual runs created doesn't work nearly as well as team runs created. The individual value of runs created is only within 88% of the actual value for the Brewers.

 

There are some differences though. With nobody on base, would you rather have Branyan up, or Cirillo? How about bottom of the 9th, 2 outs, tying or go-ahead run is at 3rd?

 

You are correct, sir. While on average a K is about equal to an out, there are instances where a K can really hurt you (or when a regular out is actually worse). A runner on third and 1 out is the worst situation to have a strike out guy like Branyan up. With a runner on 1st and 2nd and one out, it's actually better to K than a normal out.

 

i agree that the brewers, overall, don't hit that bad in the "clutch." however, they seem to be incredibly inconsistent from game to game (6 for 12 one game, 1 for 13 the next game kind of thing). the overall risp batting average stat doesn't reflect this or shed much light on this.

 

What you are witnessing is simply normal statistical variance. Assuming a binomial distribution, A team that has a .250 BA and given 10 AB w/RISP per game:

 

0 hits: 6%

0 or 1 hits: 24%

0, 1 or 2 hits: 53%

 

So while the Brewerfan faithful might proclaim a 2 for 10 night w/RISP proof that the Brewer hitters aren't clutch, it really is something that naturally occurs over half the time.

[dadofandrew]

First of all, OBP is fueled by plate appearances and NOT at bats. As we know, walks aren't factored into at bat totals, so there's more than a 2% accuracy skew right there, right?

 

Maybe Bill James will give you a real reason, but I've always understood AB's are used simply because that's the way the formula works best.

 

Secondly, I'll never be convinced that a formula can ever be used to 98% correctly track a player's offensive contributions if it doesn't account for bad baserunning, smart situational hitting (i.e., not grounding out when there's 1 out and the tying run on 3rd) or at least SOME nod to hitting when men are on base.

 

You can have your own beliefs, but I sinmply think most of that stuff simply balances out. Realistically, "bad" baserunning isn't a major factor over 162. Sometimes, you'll even be rewarded for bad baserunning, as the fielder will drop the ball, it'll short-hop the catcher and it will get past them, allowing a run and an advance, etc.

 

Also, you concede that, on a player-vs-player basis, OPS falls below the 98% accuracy rate. How far would you say, though? I'd say it drops quite a bit, even when comparing a 2004 player vs. his own 2005 output. It's that tough to factor.

 

I've never seen the team total at less than 90%, so I'd hate to go less than that. I'd say luck is a bigger factor than anything else not measured by OPS, the difference between a game ending DP and a game winning base hit is inches, and for the most part, isn't something the batter controls...other than how hard he hits it.

[rluzinski]

Maybe Bill James will give you a real reason, but I've always understood AB's are used simply because that's the way the formula works best.

 

See my explanation above.

 

I've never seen the team total at less than 90%, so I'd hate to go less than that.

 

Again, look above. Runs created doesn't work well at all when looking at individual players.

 

You even looking at my posts, AL? I'm trying to help you out here! http://forum.brewerfan.net/images/smilies/smile.gif

[EdgarDiazRocks]

Also, I did this last year, same situations, same amount of runs, but opposite RISP numbers. It's a meaningless statistic.

 

I had tried to show, apparently unsuccessfully, In my '88 Dodgers thread that RISP wasn't meaningless. That in certain examples it led to things like this . . .

 

http://images.art.com/images/products/large/10103000/10103671.jpg

 

Now whether that team, that season was a "fluke" or not is certainly debateable, but as I heard from a Dallas Cowboy who won a Suberbowl on a bad call once, "Hey the ring still fits and the checks are all in the bank"

 

Also I would like to point out that the "98% accurate" debate is only applicable when relying on analysis of what a player/team "has" done not as a predictive tool for what a player "will" do.

 

PECOTA's inner workings are guarded like an Israeli ambassador in Egypt for a reason (I guess) and that is that it is right slightly more than it is wrong. But still when this year it predicted more of the same for Todd Helton it failed to see the very human response that opposing manager would have when they realized that NOBODY was hitting behind him.

 

Turns out you didn't need a Cray crunching numbers all day to figure that one out.

[dadofandrew]

Ironically, Gibson's HR came with a man on 1B, so no RISP came into play.

 

Note: Unless Davis stole 2B???

[rluzinski]

I had tried, apparently unsuccessfully, In my '88 Dodgers thread that RISP wasn't meaningless. That in certain examples it led to things like this . . .

 

I can't speak for Al, but my point isn't to say that RISP samples don't affect anything; they just are largely a result of a small samples not always matching the population they come from. The Dodgers example obviously goes against the grain of many of the SABR testiments, but you don't disprove something by finding an example where it doesn't work (not with probabilistic statistics atleast).

 

As for being able to predict with OxS, if you gave me all the individual OBP and SLG's of every player for the Brewers next year, I could estimate their collective run production to within 95% or so. Just let me know when ya got em http://forum.brewerfan.net/images/smilies/smile.gif

 

But still when this year it predicted more of the same for Todd Helton it failed to see the very human response that opposing manager would have when they realized that NOBODY was hitting behind him.

 

Todd Helton has actually hit above his career average for July and August, and is well on his way to doing the same in September. I simply don't believe that lack of "protection" resulted in Helton losing 200+ points on his OPS earlier this year. Again, that's not to say protection has NO effect, but if it does help or hinder a player, it's effect is so small that it's masked by normal statistical variance.

[GenoSeligPrieb]

"Again, look above. Runs created doesn't work well at all when looking at individual players."

 

OK, then why is it so widely used, and quoted as gospel, when comparatively debating the offensive merits of just about every player we've ever discussed?

 

Honestly, I'm not being combative, but we've all read the threads where someone asserts that Ichiro Suzuki is a terrific offensive force, but that person is told Suzuki isn't that good because his OPS is not high enough. It's just not true.

 

Scrubs like Jack Cust and Keith Ginter are hailed as stars because of OPS and in reality, they can't keep a job as a starter in the big leagues.

 

And even Vladimir Guerrero was dogged a bit in this forum because his aggressive batting behavior leads to lower OBP totals.

 

OPS is an effective gauge to assess a player's offensive worth. But we can't lose sight of the fact that so many of these high OPSs we're seeing from partial seasons can too often be attributed to favorable splits against pitchers these platooners tend to hit better.

 

An OPS figure is blind to a 75/25 split in a batter's favor. A Wes Helms can turn in an .844 OPS, but it doesn't discern that most of those at bats are the result of favorable match-ups. If Adam LaRoche were to face lefty pitchers regularly, he would be abysmally bad, but again, his OPS shows an interpretation-free .729 (which isn' event THAT good).

[EdgarDiazRocks]

Todd Helton has actually hit above his career average for July and August, and is well on his way to doing the same in September. I simply don't believe that lack of "protection" resulted in Helton losing 200+ points on his OPS earlier this year.

 

Well we disagree here. I watched Helton closely since he was one of the cog's on my Roto team and before his injury he was frustrated at the bad pitches he was seeing and probably the steriod allegation from spring training and his strikeout numbers from May and June are elevated much above career norms. As he flailed at pitches out of the zone

 

Both once he got hurt and Preston Wilson left the team and he had some time out of the lineup to re-tool (read re-hab) he reverted to norms. Ryan Sheely's stick might have had something to do with it too.

 

Looking at another hittter who was percieved to have been pitched around, Dale Murphy, especially the statisticaly freaky 1987 season, where Murphy was pitched around at a Bonds like clip shows no relation to Helton's 2005, so maybe I'm wrong, but those two moths stand out to me as Helton's problem, along with a historically bad Lefty/Righty split.

 

LHP has OWNED him this year. Interesting . . . ..

 

 

EDIT:

The Dodgers example obviously goes against the grain of many of the SABR testiments, but you don't disprove something by finding an example where it doesn't work (not with probabilistic statistics atleast).

 

I glossed over this the first time . . .

 

but that's odd I thought finding an example of when something doesn't work DOES disprove a statement (in scientific theory at least.)http://forum.brewerfan.net/images/smilies/wink.gif

 

We don't deal in a world of absolutes.

[BillScottCanRake]

An OPS figure is blind to a 75/25 split in a batter's favor. A Wes Helms can turn in an .844 OPS, but it doesn't discern that most of those at bats are the result of favorable match-ups.

 

Stats are just tools, like observations. Stats are just as easy to misuse and observations are to bias. They are useless when not used properly, just as a hammer is useless when trying to put a puzzle together.

 

A good GM and/or manager will use the Helms data and determine "Wes Helms should be used as a pinch hitter and occasional starter against lefties. When used this way, he's likely to produce. As a starter, he's less likely to be that productive." A GM/manager who's able to come to that sort of conclusion, whether it be by observations, statistical analysis or a combination thereof, will be able to maximize the talent on the roster. Rather than letting Helms rot on the bench as a valueless player, Melvin and Yost have done a good job using him in situations where he's got a good chance of succeeding. That's as good as they can do.

 

If a GM sees an 844 OPS and thinks he can plug Helms in at third for 150 games and get that sort of production, he's using a hammer to put a puzzle together.

 

I have always said that I think stats are best used in post-season self-evaluations. They can help tell you if your team was really as good as it's record, or if your team was really as bad as it's record. Which players are likely to rebound from poor years, or return to earth after above average years. That sort of thing can go a long way in helping a team maximize their talent.

[rluzinski]

OK, then why is it so widely used, and quoted as gospel, when comparatively debating the offensive merits of just about every player we've ever discussed?

 

Basic runs created is not teh greatest when attempting to calculate individual contributions in the form of actual runs added. Since a player doesn't score himself it makes no sense to multiply their own OBP to their own SLG. That has the effect of over-estimating the contribution a superstar adds to a team. You seem to jump from what I say above to "OPS isn't that good", however. It depends on what you are trying to do with it.

 

OPS (and it's stronger, older brother, OXS) is a nice, dirty tool to compare relative offensive values of two players. It's certainly not a perfect stat, but it gets you alot closer to the truth than many tradaitional stats (BA, RBIs, etc...). Better stats have been created since OPS that do a better job of comparing players, but as long as you don't use OPS as the end all, be all, it's a great tool, however.

 

Whenever you use ANY stat you have to ask:

 

1. Is the sample size large enogh to represent the population ( in this case, true ability).

 

2. Is the sample actually random (or is the player only seeing favorable matchups).

 

Also, if you are comparing players of different positions and park effects, the number doesn't have to be equal to be equal. It comes down to knowing how to use the stats and not assuming TOO much.

 

Well we disagree here.

 

I don't think we disagree as much as you might think. I just disagreed with your initial statement, which seemed to suggest ALL (or atleast significant amount) of Helton's struggles this year was a result of not being protected. There is most likely a combination of contributing factors (random variance would have a hard time explaning such an incredible swing).

 

No studies have really found any evidence of protection having a statistical effect on a player large enough to be detected. As usual, that's not to say it doesn't exist, it just appears to be smaller than random variance.

 

EDIT:

 

but that's odd I thought finding an example of when something doesn't work DOES disprove a statement (in scientific theory at least.)

 

That is 100% not true in anything that deals with probabilistic statistics. Since basically all the data we deal with in baseball has a distribution about a mean value, the actual data is rarely exactly AT the mean. The best we can do is give the probability that the value will be between such and such. A team that has a .250 batting average still has a chance to go 5 for 10 or better in a game w/RISP, but it's only about an 8% chance.

 

I recognize that that it can be frustrating when a stathead waves his hand at every piece of anectotal evidence while muttering, "small sample" or "normal variance" over and over. The fact remains, however that it's really HARD to have stats prove anything. An example or two just isn't going to cut it for us math wierdos. That doesn't mean it's not a PERFECT example of intangibles doing their thing, it's just that it's really hard to prove when a mountain of data points you in another direction.

 

We don't deal in a world of absolutes.

 

Agreed. Just probabilities! :http://forum.brewerfan.net/images/smilies/smile.gif