Puig

For the record, this is starting to sound like I don't pay attention to statistics.  Nothing could be farther from the truth.  We actually partnered with Baseball Prospectus and TrackMan which should mean something.  I am all for information, the more information the better.  

At the same time I believe there are things that happen that have nothing to do with statistics.  Some of these "beliefs" are based on common sense.  It is my "belief" the batting order plays a part in nearly every game played.  Even more so without the DH.

Does a National League pitcher know when the opposing pitcher is on deck? Does that ever have a bearing on how he pitches to the 8 hitter?  Sure there are situations where this is more important than others.  For sure some pitchers can hit much better than others.

Ralph Houk, the Yankees manager in 1961 put Maris in front of Mantle so he might get better pitches to hit.  That decision has been documented.   This was before we had the sophisticated statistics we have now days.  This decision was simply based on common sense.  That being pitchers are not going to pitch around Maris to get to Mantle.  We might have statistics that tell us it didn't make any difference.  But how do we know for a fact what would have happened if those 600 AB's hitting ahead of Mantle didn't happen? How do we take the other 4 HRs and make sense out of that.  There are lots of guys who have hit 4 HRs in a week, but never come close to 61.

Maybe we should research how many times Maris hit a HR that year when the on deck hitter was the tying or winning run. I just can't wrap my mind around a pitcher not knowing where he is in the batting order.  Maybe sometimes in amateur baseball where there is not as much information about the hitters, but not in the Big Leagues. For sure the manager is going to know and makes decisions accordingly.

So I'm not discounting the value or even accuracy of statistics.  However, they can only tell us what did happen, they can't tell us what would have happened or what will happen with 100% accuracy.

Kind of feels like I'm in a court room.  I'm all for information.  I'm also for beliefs. I'm sure JH and others understand the math involved much better than I do.  

Also, I was a big Roger Maris fan.  I always felt like he was slighted for his accomplishments.  In fact, his son is a very good friend.

PG- As more information has come about, it has been discovered that much of the "common sense" that has been prevalent in the game for so long is, in fact, not sensical at all. For example, to continue on the topic of lineup optimization, it has been proven that the third hitter in the lineup should not be the best hitter in the lineup. The second hitter should be. The difference in production over the course of a season is very small, perhaps a few runs or so, but it is nonetheless relevant. "Common sense" has been proven incorrect.

In the case of Maris and Mantle, and many other instances where lineup protection has been cited as a reason for hitters being placed in the lineup where they are place, the "common sense" that goes into the thinking is also not sensical.

As a pitcher, why would I give Roger Maris a better chance to hit the ball knowing that Mickey Mantle is on deck? Why wouldn't I just try to get Roger Maris out the same way as I would in any other at-bat? I'll never, ever understand why a pitcher (or a manager) should approach a specific at-bat worrying about who is going to hit next. Changing a game plan due to the next hitter is effectively lessening the optimal chance of retiring the hitter in the batter's box. Mickey Mantle was a great hitter - one of the best ever. That doesn't have anything to do with Roger Maris. Statistics, fortunately or unfortunately, support that claim as well.

Roger Maris was a very good hitter, but it wasn't because Mickey Mantle hit behind him. I don't mean to be confrontational, PG, but your beliefs are not supported by evidence. Ralph Houk may have rationalized his lineup decisions based on that same belief, but that doesn't mean it made Maris and/or Mantle a better hitter.

I've mentioned it a few times on this board, and I'd like to mention it again. There is a book entitled "The Book: Playing The Percentages In Baseball" that touches on just about every "common sense" thought in the game. Some of the contributors to the book used to contribute to Baseball Prospectus, which is now one of your partners (wonderful idea, by the way. Seriously.). It's a somewhat cut-and-dry book that doesn't have any type of storyline, but it does perhaps the best job explaining some of the things I try to explain here. 

Isn't this what every announcer says? That the pitcher has to bear down and get this guy out so the 'elite' hitter doesn't come up with anyone on base. And if true, wouldn't that have the opposite effect of protection as traditionally thought and hitting in front of an elite hitter actually make you look worse?

You're conflating different meanings of the word statistics. Baseball "statistics" are a numerical record of what occurred in the course of one or more baseball events (whether those events are PA, Games, seasons, or whatever). We can use mathematical "statistics" to analyze that record and assess the significance of those events.

Using those mathematical methods to analyze the records of MLB, I can state with a high degree of confidence that protection doesn't exist. The fact that Maris hit 55 of his HR with Mantle on deck (he hit at least 2 of his batting 3rd HR with other people on deck) is a fact. Whether that fact has meaning in analyzing your hypothesis of lineup protection is a question of mathematical statistics. Given that I've shown that Maris could easily have that arrangement of his HR by chance alone, that particular piece of the baseball "statistical" record does not really support your hypothesis. We could look at a ton of other factors (opposing pitchers, ballparks, day/night games, etc) that would all have an impact anod further illustrate that there are entirely too many factors at play for when he hit those HR to be able to draw a reasonable conclusion that Mantle's "protection" was the causal factor.

There are very large numbers of "statistical" events in the baseball record. Given that, it's inherently the case that really rare things will happen some of the time. Perfect games, hitting for the cycle, first AB HR, Joey Votto pop-ups. Being human, we like to assign reasons for these things, to find the pattern that explains them, even when a lot of the time there isn't one. Randy Johnson threw a perfect game and so did Philip Humber. If all I knew about them was how uncommon perfect games were, I might be inclined to think that meant they were both fantastic pitchers. I can, however, use statistics to realize that a (potentially large) chunk of any perfect game is luck, and that Philip Humber was just lucky (really, mind-bogglingly lucky) and not particularly good as MLB pitchers go while Randy Johnson was just garden variety perfect game lucky and an insanely good pitcher.

Finally, on the topic of common sense, lots of people believe all sorts of things thanks to "common sense". To paraphrase Lionel Caffey Jr, it doesn't matter what you believe, it matters what you can prove. In all seriousness, http://rationalwiki.org/wiki/Common_sense

This is getting crazy.  I "believe" I can get this hitter out doing this.  I "believe" I need to pitch to this guy because Barry Bonds is on deck.  I "believe" I will walk Barry Bonds with the bases loaded because I "believe" that gives us the best chance to win the game. I "believe" I will look for a fastball.  What "proof" do I base these decisions on? What statistic do I use?  Is there proof that walking Barry Bonds with the bases loaded is the right or wrong decision?

When something has already happened, how do you 100% prove it would have happened anyway no matter what took place at the time? You can only prove the things that did happen.  You cannot prove what would have happened.  The information (statistical) has great importance.  It cannot completely cover everything.  

PG- No one is saying that there is any stat that exists that can cover "everything." But certainly there are predictive statistics that enable people to measure possibilities in the future. Denying the existence of- or not using- the statistics would not be beneficial to an organization, knowing that other organizations use data successfully. 

Further using your example, walking Barry Bonds with the bases loaded was never a good decision. Ever. There was never a point in Barry Bonds' career when the chance of him driving in more than one run without recording an out was 100%, so giving him a 100% chance of reaching base and the Giants scoring a run was a bad decision. 

There is a lot of statistical proof of all of the items being discussed in this thread, much of which was also shared here. I'd love to further help disseminate information but to be honest, I'm not sure what else I can provide to explain that a lot of "beliefs" are untrue.

JH, if it's late in a game and you have a multi-run lead, the proper question for a probability analysis is not the odds of his producing multiple runs, but whether the Giants would be more likely to get multiple runs with him walking and the next guy definitely batting, or with you taking your shot with Bonds and then possibly still also facing the next guy.  There was a period in Bonds' career where the odds definitely favored the intentional walk.  Even when the next guy was Jeff Kent!

Incorrect. Barry Bonds was walked intentionally with the bases loaded on 5/28/98. When the run was scored on his intentional walk, the Giants had a 10% less chance of winning the game than before the intentional walk (http://www.baseball-reference....N199805280.shtml#wpa). The RE24 (http://www.tangotiger.net/re24.html) chart remains the same, but obviously a run is added in actuality because in the particular event being discussed, a run was scored in the situation at hand. Barry Bonds should NEVER have been intentionally walked with the bases load. Ever.

Again, I don't mean to sound argumentative, but I don't understand why people keep arguing with the numbers. It's black and white and it's not difficult to understand.

From Wikipedia

In the history of Major League Baseball, six players have been issued intentional walks with the bases loaded (thus giving the batting team an automatic run). This is similar to the intentional safety in American football, which also concedes a certain amount of points to the opposing team in order to avoid a situation in which the opposing team might score more points than the safety itself. This is only done in the rarest of cases, typically when the pitching team is leading by four runs or less late in the game and a particularly feared hitter is at the plate. The six players given such passes are Abner Dalrymple (1881), Nap Lajoie (1901), Del Bissonette (1928), Bill Nicholson (1944), Barry Bonds (1998), and Josh Hamilton (2008). In all six cases, the pitching team went on to win the game.^[4][5]

If this is correct, walking a hitter with the bases loaded has worked everytime. Small sampling, but still 6 for 6! Those are the numbers!

"Further using your example, walking Barry Bonds with the bases loaded was never a good decision. Ever. There was never a point in Barry Bonds' career when the chance of him driving in more than one run without recording an out was 100%, so giving him a 100% chance of reaching base and the Giants scoring a run was a bad decision."

JH,

There doesn't have to be anywhere near 100% chance of Bonds driving in more than one run for this to make sense. If I'm up by 2 or more runs, I may well be better off taking 100% chance of the Giants scoring 1 run by walking Bonds, rather than pitching to him and taking xx% chance of the Giants scoring 2 or more runs and tying or winning the game.  If "xx%" is 90% this is certainly true, and probably true at 50%.  I don't know how low this percentage would have to be to pitch to him, but that's the manager's call.

Every time a pitcher has thrown 12 perfect innings in a game, his team has lost! The numbers don't lie, throwing 12 perfect innings in a game is bad for your team!

Actually, this is an intriguing response. Smitty- correct me if I'm wrong, but I think this has to do with the economic theory of marginal propensity to consume…in this case- accepting the present value of something in return for the future potential results. I think that this particular scenario might prove difficult to quantify the true value of the decision-making process, but in theory, I don't think you're wrong.

It is precisely this reasoning (along with confirmation bias and an inability/unwillingness to understand probability) that makes it possible to play poker profitably (for some of us).

It's also related to why Mitt Romney didn't have a concession speech ready on election night in 2012, or why anyone thought McCain had a real chance to win on election night in 2008.

At a rough approximation the on-deck hitters' OBP has to be lower than Bonds chance of hitting at least a single, which wasn't the case in the circumstances given (it's actually a lot more complicated than that, but the WPA calculation quoted above very likely captures all the permutations reasonably well).

There's also this, http://www.stat.duke.edu/~jerry/Papers/brj04.doc.  If it's generally not a good idea to walk bonds intentionally in less highly leveraged situations, doing so in this one isn't going to be better.

In 1998, Bonds had on base percentage of .438 overall. But when he did not walk he had an OBP of only .309 on 167 hits, 8 HBP's, 6 SF, on 566 plate appearances excluding walks. So if he didn't walk, he had a 31% chance he would at least tie the game with a hit or HBP. This intentional walk took place with two outs in the ninth inning. If he was going to make an out in 69% of his plate appearances when he did not walk then why not pitch to him. Not only that but the HBP and some of the singles would only tie the game. If he made an out and he absolutely was not walked then he would fail to get any run in 69% of the time. Don't forget his off the chart OBP were that way because of all the walks he was given. Now I understand that his OBP would go up because sometimes he walked when a pitcher was NOT trying to walk him but I bet that wasn't many times as one would think. But the whole analogy is NOT walking him here and pitching to him, so it seems that some 69% of the time he would make the last out of the game. Maybe JH or someone here can point out if my math thinking is flawed.

Wouldn't it also matter if Bonds was hitting a lifetime .600 against whoever stood on the mound or .150?

"Specific batter vs pitcher data is probably the worst use of statistics in the entire sport.” - Dave Cameron

http://books.google.com/books?...ge&q&f=false

http://www.fangraphs.com/blogs...er-match-up-numbers/

Not really.  The sample sizes for the vast majority of pitcher/batter matchups are far too small to generalize from, and there's essentially no reason to think Bonds would hit like anything other than Bonds vs any particular pitcher (allowing for platoon differences and whatever the ability of that pitcher actually is, of course).

FWIW, coming into the PA in question, he'd faced Gregg Olson twice and walked both times.

Also, to clarify three bagger's post, pitching to Bonds is safer because Bonds can only tie the game for sure with a 2B or better (it's 8-6 when he comes to bat with the bases loaded), though a lot of singles will probably do so as well, and it's probably going to take a HR to win it (his odds of tripling there are negligible).  After walking Bonds, Mayne will tie the game by safely reaching base in any possible way, and will win the game with any 2B or better. So Mayne's only slightly less likely to win the game outright than Bonds is, but he will tie the game much more often (and those ties retain the possibility that the next hitter will win it) with the ultimate result that walking Bonds is a net loser for the DBacks.

Bonds had more career plate appearances against Greg Maddux than any other pitcher, with 157. 

His highest career batting average against a pitcher he faced more than ten times was .667, in 17 plate appearances against Carl Pavano. Using the same arbitrary point for plate appearances, He was 0-14 vs. Chris Peters, 1-20 vs. Ray King, 3-41 against Mike Bielicki. 

For his career, Barry Bonds hit .303/.458/.626 against RHPs, in 8459 plate appearances. He hit .289/.417/.569 against LHPs, in 4147 plate appearances.

None of the matchup totals against specific pitchers are large enough samples of data to form a conclusive predictive measure for results of a particular plate appearance, and his career platoon does not indicate an extremity that would lead me to believe a bases loaded IBB would behoove the team. 

Three responses on the same topic in 7 minutes, all using different data to prove the same point. Go evidence!

I think you are right that in this case 69% of the time he would make the last out of the game.  You didn't say anything about the guy batting behind him (here we go again...) but if I had an 80% chance the next batter would make the last out, wouldn't I rather pitch to him instead? 

jacjacatk said this above: 

Also, to clarify three bagger's post, pitching to Bonds is safer because Bonds can only tie the game for sure with a 2B or better (it's 8-6 when he comes to bat with the bases loaded), though a lot of singles will probably do so as well, and it's probably going to take a HR to win it (his odds of tripling there are negligible).  After walking Bonds, Mayne will tie the game by safely reaching base in any possible way, and will win the game with any 2B or better. So Mayne's only slightly less likely to win the game outright than Bonds is, but he will tie the game much more often (and those ties retain the possibility that the next hitter will win it) with the ultimate result that walking Bonds is a net loser for the DBacks.

As I mentioned upthread, the following batter needs (at a minimum) to reach base safely via any method less often than Bonds' is likely to at least get a hit. Brent Mayne was the ondeck hitter, and his career OBP at the time was about 30 points higher than Bonds' batting average (and that slightly understates the gap for various reasons).

Other than a pitcher (who'd get PH for and so isn't going to happen), there's almost no one who'll fit the bill in this case.

No reason for discussion here. Someone has a stat reference on the internet to dispel what might have been the thinking during the season or the AB in 1961 or 1998, when the stats internet articles didn't exist.

Gonna be watching Bochy to make sure he is on the internet before he makes those double switches.

The math for the Bonds decision was easy enough to do at the time, no internet needed.

Batting Maris in front of Mantle really doesn't make any difference, no matter how it was justified at the time or now.

Strat-o-matic baseball, coincidentally, was invented in 1961. Bill James once suggested that every big league manager should be required to play 1000 (I think that was the number) games of SOM before being given the job, because understanding the math behind the decisions they have to make is fundamental to the job, and knowing how to manage the players as people isn't all that useful if you don't have the requisite knowledge to make good decisions about deploying the human talent you have to work with.

Like I posted, no reason for discussion here.

Why not? Questioning a manager's decision-making has been happening since the beginning of the game. Statistical evidence indicates that several decisions cited on this site were incorrect decisions. Why is that something that shouldn't be discussed?

It seems as though the implication you're making is that us "nerds" just don't get it, but maybe I'm misreading it.

Not sure what the point of posting something to a discussion board is if you don't think it's worth discussing.

The statistical evidence being used to question the Houk decision in 1961 and the possible decisions of pitchers in 1961 is based on statistical analysis in 2014. Same with Bonds in 1998.  To me, it is the ultimate in Monday Morning. The question is how to apply these to in game instant decisions. Do we, in 2004,  pitch to Bonds or walk Bonds when his OBP is off the charts as is his slugging percentage and BA. It is one way to look in the rear view mirror and judge.

It isn't one extreme or another, in my view. The effort to suggest PG, for instance, is all wrong for being alive in 1961(Hope that is right PG)and relating how the game was and Yankees and Maris were viewed from the eyes of 1961 is part of baseball.  PG isn't all wrong, assuming my memory is correct,  and I question if retro stats are so dogmatic to be  right.

To illustrate my point, my post on whether Bonds was hitting .600 or .150 got immediate rejection with internet references.  I have enough experience to know players can go 20-30 in one streak and 0-20 in another.  Within those streaks, I might want to make a baseball judgment for that player at that point in time and that AB, against that pitcher, no matter that a stat guy wrote an article on the internet which says the human judgment element in baseball should be turned over to a computer.

infielddad- I see the discussion from the complete opposite view. Now that we are capable of having such analysis at our fingertips, using retrospective examples of situations that could arise in the future help make our overall knowledge of the game, and the decisions that occur within the game, that much better. Why not use the capabilities to our benefit? What frustrates me is seeing people not willing to take these things into consideration. I'm not here saying anything negative about Roger Maris or Mickey Mantle or Ralph Houk or Buck Showalter or Barry Bonds, and I'm pretty sure jacjacatk isn't either. I'm ABSOLUTELY not questioning Jerry Ford's - or anyone else's - knowledge of the game. I'm trying to illustrate, using relatable examples, that the game can be optimized in better ways with the use of data. 

One question is what is the statistical evidence that the game is played better, in situational situations, based on the statistical look back.

Stated in a different manner, on MLB TV I am listening to nearly identical analysis which I have read and heard for many years. This guy looks like,,,,,,, reminds me of........., projectible body, going to get bigger and stronger....

How is any of the narrative from 4pm, EDT to now, any different in projecting? When dp stats work or not, iin analyzing a baseball situation where they were applied, not looked at in the rear view mirror..  I love reading the Bill James stuff. I also love what Brian Sabean does, for instance.  Baseball has an amazing human element, emphasized today by the death of Don Zimmer. If the statistical elements make the human judgments better, that is all the better.  This discussion, to  me at least, is using them to say past judgments were wrong. Maybe I am misunderstanding but that is what I am reading.

No one's questioning Houk's decision, just whether his stated rationale had any impact on Maris' 1961 season, or whether Maris' 1961 season is good evidence of protection as defined by PGStaff and others (it's not).

Showalter's decision was questioned at the time in 1998 (online even, and I suspect you could find me questioning it with some digging). Heck, IIRC correctly, Showalter's own coaches questioned the decision at the time. In any event, the information necessary to analyze the decision was available then, though it's more trivially available now, but certainly any MLB staff would have had at their disposal the means to realize it was a bad decision if they had wanted to employ that information.

If you're making "baseball judgments" based on whether a guy's in a 20-30 streak or an 0-20 one, you're making bad decisions that are not evidence based. It's confirmation bias waiting to happen.  A guy who's 0-20 who gets a hit was due, and one who doesn't is just still in a slump. A guy's who 20-30 and gets a hit is just red hot, and one who doesn't was bound to cool off some time.

If you could actually predict what players were going to do in their next X PAs based on how they had done in their last Y PAs (where X and Y are numbers like 20 or 30), it would be trivially easy to make yourself rich with that information. And MLB teams would have a lot more variable lineup construction than they actually do.

Like I said, no reason for discussion here!!!!!!

So, if the Yankees had them on the roster and placed Phil Linz or Tom Tresh in the Mantle slot behind Maris, would Maris have hit 61? If the Yankees put a guy who hit 10HRs behind Maris rather than Mantle, who battled with 54 until his late season injury, Maris would still have hit 61?  So much for the old baseball axiom that a guy picks up his teammate. From now on, stats and computers pick them up, not their teammate?

So the 2 points in your eyes is, some of us think we can look back use statistical analysis and say, hey, this did nothing. You think we can’t look back and say it was wrong and that Ralph Houk knows better than we do?

There's a substantial body of statistical evidence that posting about something on a discussion board is a poor way to stifle discussion about it.

I am pretty old but I think my earlier post is I love reading and learning about Bill James and his information AND I love reading and learning of the Brian Sabean approach. They are not the same and for a number of years, I questioned the Sabean approach. Over the last 7-8 years, he has made a huge believer of this guy.

And yes, I don't think you can take current data and put it in the context, pressure and environment of 1961 with Houk. To do so, to me, diminishes Houk, Mantle, Maris and baseball being a team game. Just my view!

"If you're making "baseball judgments" based on whether a guy's in a 20-30 streak or an 0-20 one, you're making bad decisions that are not evidence based. It's confirmation bias waiting to happen.  A guy who's 0-20 who gets a hit was due, and one who doesn't is just still in a slump. A guy's who 20-30 and gets a hit is just red hot, and one who doesn't was bound to cool off some time."

This statement is one that I just don't agree with, and is perhaps the problem I'm having with this thread.  I am a big believer in data, but at the same time I've played enough baseball to know that hitters get hot and hitters get cold.  There are periods of time when you see the ball, when your mechanics feel right, and hitting is easy.  It doesn't matter who's pitching or what the situation is.  We used to call it being in the zone.  There is no doubt in my mind that a hot hitter is more likely to get a hit than a hitter in a slump.  Sure, hot streaks and slumps begin and end, but there's no denying a hot Josh Hamilton is an unbelievable hitter, while a cold Josh Hamilton is a near certain swinging strike out.

The problem with data as it's being used here is that it can smooth out the peaks and valleys.  Averages are just that.  They don't account for the variability within the data.  They don't account for the variances in performance around the mean.  I am not seeing how they account for streaks and slumps.  Perhaps people are doing Monte Carlo or other analysis to factor in these variances, but if so I must have missed it.  I suppose if you don't believe in streaks and slumps than this wouldn't be necessary, but then I would say this is bad decision making.

Smitty- The issue with small sample sizes has to do with the predictability of the results within the sample. A streak is just that - a small sample with no predictive value. Sure, streaks exist and players feel hot and cold. No doubt about it. But it's impossible to predict when those hot and cold streaks will start and/or end. Your Josh Hamilton example is a perfect one to illustrate the volatility and arbitrary nature of streak lengths. Relying on unpredictable data for measurement of future actions, in and of itself, is bad decision making. 

I'd add that this is the same fallacy that leads to non-professional investors selling at market lows and buying at market highs. You can't time the market, and you can't predict hot and cold streaks with any certainty. See Gambler's Fallacy.

In fact, you can probably make a lot of money by betting on regression to the mean. Charlie Blackmon put up a 1034 OPS in April this year, and I'm betting you could have made good money in Vegas betting against him coming anywhere near that in May (when he actually put up a 718). He's much better than even money to end up closer to his current career .774 than to his current .856 by the end of this season

In regards to results based streaks. You can't measure streaks based on results (which is why citing 0-20 or 10-20 is totally useless) in which the inputs are pitcher skill, batter skill, defense skill and luck. You can't detect a true hot or cold streak. Who's to say the batter is really seeing the ball and just roping line drive after line drive but he happens to hit it right at the fielder and goes 0-5. Or let's say the batter really is seeing the ball better but runs into Clayton Kershaw or RA Dickey, or just some random schlub pitching the game of his life and just doesn't do anything. Or maybe he comes to the park sick and miserable with a monster headache, doesn't want to play and goes 5-5 with 3 dribblers down the line and a few bloops off the end of the bat.

So basically what i'm saying is, hot and cold streaks probably do exist in some form, BUT WE DO NOT HAVE THE CAPABILITY OF SPOTTING THEM and anyone who claims they do is lying or a fool.