TT: Showalter should be leery on how he uses Johnson

[Frobby]

Just a note that the Red Sox recently promoted Koji Uehara to closer, and last night he closed for the third night in a row.

Yes, that Koji Uehjara.

[Can_of_corn]

Just a note that the Red Sox recently promoted Koji Uehara to closer, and last night he closed for the third night in a row.

Yes, that Koji Uehjara.

Koji also averages what 10 pitches an inning?

[Frobby]

Koji also averages what 10 pitches an inning?

His three outings were all 1-2-3 affairs, 2 K's in each outing. 13, 16, 15 pitches.

[Old#5fan]

Anybody know who is considered actually as the first "closer" in major league baseball? The earliest guy I remember fitting that bill was Dick Raditz of the Boston Redsox. Big, mean looking fireballer they brought in at the end of the game. The first Orioles closer I recall was Stu Miller.

[Can_of_corn]

Anybody know who is considered actually as the first "closer" in major league baseball? The earliest guy I remember fitting that bill was Dick Raditz of the Boston Redsox. Big, mean looking fireballer they brought in at the end of the game. The first Orioles closer I recall was Stu Miller.

The Senators had a guy that toward the end of Walter Johnson's career would come in and finish his starts. Firpo Marberry maybe? He ended with what is now 101 saves. Looks like he led the league five times.

[OFFNY]

The Senators had a guy that toward the end of Walter Johnson's career would come in and finish his starts. Firpo Marberry maybe? He ended with what is now 101 saves. Looks like he led the league five times.

If we can talk him out of retirement, I'm all for it.

[Can_of_corn]

If we can talk him out of retirement, I'm all for it.

He entered "second stage" retirement in 1976.

[OFFNY]

He entered "second stage" retirement in 1976.

Yeah, he might have a dead arm, which would concern me.

[Redgrape]

I'm a little late to this thread. I work as an algorithmic trader at a hedge fund, which requires me to use statistics to find predictive indictors for financial markets. I also work with a lot of MIT math and stat majors. I would say I have a fair bit of expertise in statistics.

I'm skeptical that the statistics used to show that JJ has problems starting back to back days proves that this is an issue for JJ. And let me preface this also by saying I love this website and I have a lot of respect for Tony's opinion.

There are two main issues with using the statistical comparisons in TT's article to conclude that Buck needs to make sure JJ has a day rest before pitching. The first is most obviously variance over a small sample size. I don't know the standard deviation of ERA, WHIP, etc. is for JJ but I imguessing its high enough to the point where we'd expect somewhat large variation of JJ's performance over chosen data sets that were only the result of variance, i.e. pitching after days rest vs no rest.

The second has to do with the law of multiple comparisons (http://en.wikipedia.org/wiki/Multiple_comparisons), which basically states that when you are comparing multiple statistics its very likely that one will display a high deviation from the mean and therefore statistical significance must be much higher to make conclusions off of statistics alone. A great example wikipedia gives is one below:

"Suppose the treatment is a new way of teaching writing to students, and the control is the standard way of teaching writing. Students in the two groups can be compared in terms of grammar, spelling, organization, content, and so on. As more attributes are compared, it becomes more likely that the treatment and control groups will appear to differ on at least one attribute by random chance alone."

This is not obviously the case in Tony's article. But it is likely when Tony or anyone were to look up an answer to the question "Why is Jim Johnson performing poorly?" that they would look up a lot of different statistics on JJ (multiple comparisons), stumble upon one statistical set the deviated greatly from the mean (no rest vs days rest), and then post about the large deviation. The problem again is that there is a large chance that when looking over a plethora of statistics that you'd find some statistical set with a large deviation from the mean with some logic behind the deviation.

It would be almost impossible to statistically prove that JJ benefits from more rest than other relief pitchers, but the best way to do it would be to compare this statistic over his career to the mean of other pitchers, and if it deviates significantly, it would be likely that he does benefit from more rest than other relief pitchers do. You'd also want to look at similar statistics that would have to do with rest, i.e. as Tony did with pitching for the 3rd and 4th days in a row, ERA over larger amount of pitches, etc.

This doesn't mean that JJ DOESNT benefit from a days rest more than other relief pitchers, but thats better proved by logic than the statistics used in the article.

I'd also like to apologize if this point was made in this thread ahead of me but I was in a hurry and felt inspired to write this all out .

[Rene88]

Wow, what a cool job. I am jealous, love stocks as a hobby.

[Tony-OH]

I'm a little late to this thread. I work as an algorithmic trader at a hedge fund, which requires me to use statistics to find predictive indictors for financial markets. I also work with a lot of MIT math and stat majors. I would say I have a fair bit of expertise in statistics.

I'm skeptical that the statistics used to show that JJ has problems starting back to back days proves that this is an issue for JJ. And let me preface this also by saying I love this website and I have a lot of respect for Tony's opinion.

There are two main issues with using the statistical comparisons in TT's article to conclude that Buck needs to make sure JJ has a day rest before pitching. The first is most obviously variance over a small sample size. I don't know the standard deviation of ERA, WHIP, etc. is for JJ but I imguessing its high enough to the point where we'd expect somewhat large variation of JJ's performance over chosen data sets that were only the result of variance, i.e. pitching after days rest vs no rest.

The second has to do with the law of multiple comparisons (http://en.wikipedia.org/wiki/Multiple_comparisons), which basically states that when you are comparing multiple statistics its very likely that one will display a high deviation from the mean and therefore statistical significance must be much higher to make conclusions off of statistics alone. A great example wikipedia gives is one below:

"Suppose the treatment is a new way of teaching writing to students, and the control is the standard way of teaching writing. Students in the two groups can be compared in terms of grammar, spelling, organization, content, and so on. As more attributes are compared, it becomes more likely that the treatment and control groups will appear to differ on at least one attribute by random chance alone."

This is not obviously the case in Tony's article. But it is likely when Tony or anyone were to look up an answer to the question "Why is Jim Johnson performing poorly?" that they would look up a lot of different statistics on JJ (multiple comparisons), stumble upon one statistical set the deviated greatly from the mean (no rest vs days rest), and then post about the large deviation. The problem again is that there is a large chance that when looking over a plethora of statistics that you'd find some statistical set with a large deviation from the mean with some logic behind the deviation.

It would be almost impossible to statistically prove that JJ benefits from more rest than other relief pitchers, but the best way to do it would be to compare this statistic over his career to the mean of other pitchers, and if it deviates significantly, it would be likely that he does benefit from more rest than other relief pitchers do. You'd also want to look at similar statistics that would have to do with rest, i.e. as Tony did with pitching for the 3rd and 4th days in a row, ERA over larger amount of pitches, etc.

This doesn't mean that JJ DOESNT benefit from a days rest more than other relief pitchers, but thats better proved by logic than the statistics used in the article.

I'd also like to apologize if this point was made in this thread ahead of me but I was in a hurry and felt inspired to write this all out .

Well I'll be the first to admit that you lost me about four times in your post. Not because its not well-written, but more likely because it went over my head.

Saying that, I really don't get the bolded. Statistics are there so we can see how a player has performed. They are the only way to accurately measure past performance under different circumstances. The one fault in my "statistical analysis" was pointed out by a previous poster when I used zero day rest stats do say Showalter should be leery on using Johnson anytime he had zero days rest. What I did not take into effect was that those stats also included those three day in a row stats which clearly are terrible and skewwed the overall numbers.

At the end of the day, whether you use "logic", traditional scouting, pitch fx metrics, or go by the stats Johnson has put up when working on the third day of three days in a row, it's pretty clear Johnson is less effective.

[Redgrape]

Well I'll be the first to admit that you lost me about four times in your post. Not because its not well-written, but more likely because it went over my head.

Saying that, I really don't get the bolded. Statistics are there so we can see how a player has performed. They are the only way to accurately measure past performance under different circumstances. The one fault in my "statistical analysis" was pointed out by a previous poster when I used zero day rest stats do say Showalter should be leery on using Johnson anytime he had zero days rest. What I did not take into effect was that those stats also included those three day in a row stats which clearly are terrible and skewwed the overall numbers.

At the end of the day, whether you use "logic", traditional scouting, pitch fx metrics, or go by the stats Johnson has put up when working on the third day of three days in a row, it's pretty clear Johnson is less effective.

Thanks for the response, I think my original post was way too jargony and I'll try to explain it in a simpler way.

Let me first make clear the question we are trying to answer: "Is Jim Johnson worse of a pitcher without at least a days rest?"

My first point is simply when you have a small sample size there is variance. Jim Johnson could be much worse without rest this year not because he needs rest but simply because he has gotten unlucky on days where he has one or more days of rest and gotten lucky on days where he hasn't. Rest may have very little to do with it. Luck being a main factor in results is especially true with a small sample size. The sample size issue would be very clear if he had only pitched a few innings this year (for example, would we even be using these stats if he had a 9 ERA over 3 innings of no rest relief and a 1.00 ERA over 3 innings of at least one days rest?). Variance has less of an effect over a larger sample size but we're still talking about only 47 batters faced when he has had 0 days of rest. My point is the statistics used in your article should only be small evidence in proving the argument that JJ is a much better pitcher with rest than without because there is a lot of luck factoring in.

Another point I wanted to make is that we could use different statistics to answer this question better. For example, is there any reason to think that JJ would be significantly worse without rest this year than he was in previous years? My guess is no (but I'm interested in your thoughts, maybe age plays a slight role). Let's assume for the sake of argument that this is true, that there's no reason he'd be just as bad without rest this year as compared to previous years. That being the case it makes sense not to just look at his stats this year, but also look at his stats for the previous few years if not his entire career as a reliever and compare those numbers. This would be a better indicator because the sample size is larger and therefore we should expect less variance. To improve the evidence of this argument even more, you could look at other statistics that are similar to rest vs no rest, for example his performance on his first 10 pitches vs pitches 10-20. If his ERA was significantly worse on the latter than the former, and we also see his ERA is significantly worse with no rest vs a days rest, than we have a much stronger case that rest matters a lot with him.

I did some research through ESPN stats. http://espn.go.com/mlb/player/splits/_/id/28531/type/pitching3/jim-johnson

This is for the 2010-2012. This actually shows JJ has been better with 0 days rest than with rest when pitching. Actually significantly better if were using OPS as a metric. This is still a somewhat small sample size but is 2.5x as many samples as 2013 alone. So assuming there is nothing anomalous about this year, this is actually fairly strong evidence that this year has been a "Fluke." The statistics of rest vs no rest this year could simply explained by luck/variance.

The other point I was trying to make was a much more complex one that involves statistics where there is luck involved, its a concept of using multiple comparisons in statistics. The simplest example would be to imagine you're taking 10 pitchers, and lets assume they all have identical stats up this point, maybe even that they are all identical dectuplets! You teach all these pitchers different legs kicks, but lets assume that while leg kicks make some difference in a pitchers performance, it is not that important. You have each pitcher go out and throw 30 innings using this leg kick. Because of variance, there is a very high chance one of these pitchers will perform significantly better than the others, even though in they are all theoretically similarly skilled pitchers. If we looked at these stats however it would be easy to attribute the better performance to the leg kick the best statistical pitcher was using! But in reality it made little difference. The performance increase was much more so because of variance.

This is actually very similar to what happens when you look for anomalous statistics. When you are simply trying to answer the question "Why is Jim Johnson performing badly this year?" You are likely going to look at a plethora of statistics and look at which ones stand out. Maybe you look at the difference between his curveball and his cutter. Maybe you look at pitches 1-10 vs 10-20. The problem with this methodology is that your are comparing multiple statistics that have variance. When that is the case, its actually very likely you will find something that is very anomalous. So what a stat expert would say in this case is that when you compare multiple attributes (different baseball statistics), the anomaly's must be big and the sample size must be large before you can conclusively conclude those statistics actually represent reality (i.e., you'd want to see this is the case over Johnson's whole career or a much larger sample and you'd want the ERA differences between rest and no rest to be large).

I tried to make this as simple as I could but I'm not too great of a writer in some respects so you'll have to forgive that.

But given the extra research I just did I would think Buck should use JJ frequently and not worry about rest as the best statistical evidence we have shows that he performs better over his career without rest than with rest. This would only be untrue if the data from this year is not comparable with data from previous years. An example of that would be a change in throwing motion that has more arm strain and causes more fatigue, or an injury of some sort, etc.

[Frobby]

Thought it would be interesting to revisit this thread, in which Tony pointed out that JJ had allowed runs all three times that Buck had used JJ three days in a row. Buck never did pitch JJ three days in a row for the rest of the season, so I guess he was listening.

In Tony's article (dated June 27), he noted that JJ had a 5.98 ERA when pitching on 0 days' rest. In his final numbers, JJ had a 4.58 ERA when pitching on 0 days' rest. After Tony's article, JJ only pitched on consecutive days 8 times, whereas he had done so 13 times in the first 81 games, so, his 2nd half ERA in that situation was only 2.45. In general, JJ had a lighter workload in the 2nd half of the season, mostly because the Orioles weren't going to the 9th inning with the lead as often as they did in the first half. I don't think Buck intentionally backed off on using JJ on two consecutive days, there just wasn't a save situation on back to back days as often in the second half.