Us
Meta
Reference Material
The Future Forty 3/19/09
Evaluating Defense Evaluating Pitcher Talent
Evaluating Managers
Bugs Bunny, Greatest Banned Player Ever Selected for Best American Sports Writing!
The Attrition WarDisclaimer, Copyright
The U.S.S. Mariner is in no way affiliated with, condoned or given any notice by the Seattle Mariners baseball team, who have their own website. Similarly, we have no association with the ownership group or any businesses related to the Mariners. All article text is written by the authors, all pictures are taken by the authors, who retain copyright to their works. No copying or reproduction of any content here, photographic or otherwise, is authorized. Please email us if you wish to reproduce our work.
Game 84, Mariners at Angels
The M’s are in Anaheim to take their beatings like men. The ghost of Joel Pineiro takes on the man who ate Bartolo Colon.
Tags:
Comments
146 Responses to “Game 84, Mariners at Angels”
-
Recent Posts
- The AI Has Taken Over
- ’22 40-Man Preview Extravaganza
- ALDS Game 3: Astros AT Mariners – Welcome Home
- ALDS Game 2: The Pitch and What Follows
- ALDS Game 1 – Series Preview
- Recent Events Fill Mariners Fans With Unexpected Emotions
- AL Wild Card Series Game 2: The Ray Game
- AL Wild Card Series, Game 1: Here We Go
- AL Wild Card Series Preview: The Battle of Robbie Ray
- Game 162, Tigers at Mariners
Author Links
Local M's Coverage
M's Blogosphere
M's Official Sites
Resources
Twittah
Okay, that was in poor taste.
#98 i didn’t say he wasn’t hitting better, just that he wasn’t hitting when it counted, which is true. About the same number of chances and hits as sexson with runners on, but nowhere near the number of folks driven in, because he’s not getting as many XBH in those situations.
96. About time! Baseball does not belong in Olympics! Unfair competition.
#98 Rodger’s right Beltre IS a second half player, give him a chance.
Ernie
103: And it’s a good thing he’s only a second half player. Imagine how much he’d cost if he played this good the whole year.
A bit late, but 98 is right.
Even after you guys post a long piece about how you were wrong on ibanez because, even though i’m sure you felt at the time that “all rational arguments†were against getting ibanez.
You could, perhaps, look up what we wrote about the Ibanez signing, rather than assign us opinions and quotes.
Daaaan…
You had PLENTY of chances to clarify what you meant, but you never did. All we can go by is your words, and you kept focussing on his not coming through via his RBIs. What else are we to conclude?
When everyone else is coming to the same conclusion…and it’s different from what you meant, perhaps the problem isn’t with the audience of the web site. Again, I think you need to be a lot more precise in your writing (and perhaps you need to be a more careful reader to match).
If I recall, Beltre had a two out double in today’s game. Does that count as a clutch hit?
I understand what you’re saying about the other players, but I think you’re way off concerning Beltre…
was at game tonite. was really nice to see some offense. big hits with runners on and Piñero looked exceptional in his complete game. hope this is the start of something.
Am I the new Lauren, T.C.? Sunday-worked, missed game, M’s win. Monday-looking at apartments, missed game, M’s win. Tuesday-watched whole game, M’s lose. Wedensday-watched whole game, M’s lose. Tonight-worked, missed game M’s ROLL. I’m working all day every day through Sunday. If the M’s sweep, I’ll be torn.
hi aaron!
if you like, i can only schedule you to work during M’s games. . .
Chris Snelling. Chris Snelling. Chris Snelling.
Chris Snelling. There now. it has been said.
The whole Willie “Boom Boom” thing by Rizzs was pretty funny. Still, I think that it’s got to be tough on Angel fans to hear, “and starting in center field Chone Figgins”. He’s Willie with a smidge more talent.
Daaaaan, this is a written forum, not a radio call-in show. Suggest a higher standard is called for.
Reminds me of the old joke:
How come the liberals don’t dominate talk radio?
Because they can all read and write!!!!
Not so good at math, though, the liberals.
On WFB, Daaaan is presupposing the existence of heat. He wants to keep playing Willie as long as he’s hot, but that requires that being hot is actually something, rather than just statistical variance. If it’s just statistical variance, then every day Willie plays he is likely to revert to Bad Willie – that he doesn’t is just a fluke.
well remember that there are two kinds of statistical variance. when you are talking about coin flips, regressing toward the mean is all about probability, and the second coin flip has no relation to the first.
willie bloomquist being hot is a different statistical issue. yes, eventually baseball players revert back to themselves, but two things are different. Their average performance gets better or worse throughout their careers. They get better for awhile and then they get worse. They regress toward a changing mean.
secondly, as you know, unlike the coin flip, a series of ten at bats influences the next at bat. a player may be feeling particularly strong, rested, or confident. sometimes the ball looks like a grapefruit, sometimes a pea. a player might develop a new and temporarily successful idea of what to do with a pitch. hot streaks are real, and worth considering. willie bloomquist has a greater likilehood of getting a hit tonight than in a game in september.
but I would still rather play morse and snelling…
secondly, as you know, unlike the coin flip, a series of ten at bats influences the next at bat.
Not really, no.
a player may be feeling particularly strong, rested, or confident. sometimes the ball looks like a grapefruit, sometimes a pea.
That’s not infuenced by the previous at bat, if true.
a player might develop a new and temporarily successful idea of what to do with a pitch.
Which, if true, because it’s temporary, would be as likely to end as not.
hot streaks are real, and worth considering.
Not really, no, and no.
willie bloomquist has a greater likilehood of getting a hit tonight than in a game in september.
No he doesn’t.
Isn’t this fun? David’s asserting one position, DMZ’s asserting the opposite, and I’m asserting that we shouldn’t draw a conclusion.
I am willing to entertain the possibility that WFB is a better hitter than Morse. I think Morse has the better glove at SS, though.
I agree that this is a debatable position, but would like more of a response from DMZ than no, no, and no. The alternative position to mine is that that at bats come in random order scattered around the changing mean, that the only meaningful predictor of whether willie will get a hit is what he has done over his slowly worsening career. (It is provable that streaks exist, obviously. This is a debate over whether anything besides random order contributes to them).
Is that your position, DMZ?
Argh.
We’ve argued this here before, at length, particularly w/r/t Ichiro.
I don’t care if anyone wants to argue that players can be hot. In fact, what the heck, let’s grant that assumption for a second. But what doesn’t happen is that *a hit does not influence the next at bat in any meaningful way*. Unless you want to argue the building confidence one, which… okay.
Say Bloomquist gets a hit. That hit does not make him a better hitter. He’s no more able to see the ball better. His twitch muscle fibers don’t get faster, or stronger. He may get more confident, scrappy, whatever — but that doesn’t seem to help.
The reverse is true. If he strikes out, he doesn’t become worse.
If that was the case, we would see in players much greater streakiness than we do. We don’t, so it’s either not there or it’s not important enough to show up in player performances.
See: _Curve Ball_, etc.
This really argues for the need for a FAQ that deals with this stuff that comes up all the time.
ah the confidence of one’s conclusions. all I was ever trying to do was to get you to distinguish the difference between regressing toward the mean with completely totally random events (coin flips) and regressing toward the mean when multiple factors (physical condition, experience) are changing that mean.
…
Been a while, but…
I remember having the discussion in a stats class where the professor statistically “proved” that there was no such thing as a “hot hand” (proof was in basketball). It was a pretty standard statistical analysis, similar to the one’s trotted out here and elsewhere. I know he made all of the i’s dot and t’s cross, but I don’t remember the specifics.
But, I do remember feeling very unsatisfied by his proof. I thought it took a superficial, tautological approach, but I didn’t know enough stats to disprove him.
Essentially, he took a 30% shooter, and proved that at any point in time, the next shot was only 30% likely to hit. Thus, no hot hand. I found that a bit transitive in that, if you look historically at a 30% shooter, then by definition any “next shot” will have a 70% chance of failure. But, I don’t remember what the percentages were like where you took that same player, who had hit 7-of-9, what the next shot was. Or, even if the next shot was a miss, what about the 2nd next? ie, if I’m “hot”, I might miss THIS shot, but will, over a few shots, outperform my average. (Which, of course, would eventually cancel out against the cold spell later in the game) I don’t remember the prof disproving that.
By their nature, averages will average out over time. Thus, the “law of large numbers” and all that jazz. This all relies on one shot/hit/coinflip not influencing the next. While this is true in coins, it is NOT true in shots/batting. I was never good at basketball, but even I could get “hot”, where I knew I could hit the next few shots. (of course, most of the time I was “cold”, where I knew everything would miss, but that’s another story). Over time, I was an x% shooter. And, given a large enough sample, the chance of hitting any “next shot” would tautologically approach the same x%.
I’m not arguing against statistics, since I was eventually a TA for a stats class. But, here, I would argue that we have not yet figured out how to quantify the “hot hand”, rather than saying that it does not exist. It does, and we’ve all felt it. But how to quantify it? Dunno.
Willie Bloomquist is hot right now. Or rather, has been hot for 4-5 games. He may or may not revert to “statistical Willie” tonight. We won’t know until after. At which point he’ll have a 4-game hitless streak to prove he’s now “cold”.
Some players are streaky, some steady. Over time, the streaky players will average out to their “true” ability, but this may never truly represent them at any given point in time. ie, take a guy who hits .330, .270, .325, .275 over 4 years. Is he a “.300” hitter? Stats say yes, I say “sorta, but not really.” Maybe he’s “good in odd years”, or works out harder after every bad year, but we’ve never really seen this guy hit at a .300 pace for any consistent period. So, is he really a .300 hitter? Over time, yes, but ant any given point in time, probably not.
I’m a bad golfer.
But I once played a game where my swing was really in a groove. Everything I hit was long and straight.
Did that make me a better golfer? No. Next time out I sucked as usual.
But, for that brief shining moment, I could hit the ball wherever I wanted. Maybe it was the confidence Derek hinted at in 121, maybe it was voodoo. Maybe I sold my soul to the devil like WFB. But I was hot.
How do you quantify that? Rather than say it’s not there, or that it’s been proven false, maybe we just haven’t figured out how to measure it.
So here’s my gauntlet, then, which I will now toss to the ground:
Smart people have looked at the problem and have been unable to find streakiness in hitters as a whole. Hot and cold streaks have almost always been shown to be the product of random chance, with many hitting troughs the product of injuries.
If you believe there are hot streaks, prove it. Come up with a test and run the stats. Publish your results in one of the many fine places that run original research. See what happens.
I’m entirely serious: if this is a real phenomenia, then I want to know more about it. Go forth.
Neither the “hot streak” nor the “random variation” explanation should be satisfactory to the thinking fan. Let me propose a scenario that I think reconciles the opposing views of whether such streaks exist or are random.
Let’s say a hitter makes an adjustment to improve a weakness that has been in every team’s scouting report for a while. Suddenly he’s smacking the ball while pitchers continue to deal with him according to their scouting reports. (David: “He’s hot.” Derek: “It is merely random variation.”) Eventually opposing teams figure it out, update their scouting reports, and start exploiting some other, perhaps new, hole in his swing. (David: “He’s cold.” Derek: “It was merely random variation.”)
The truth is, the batter’s performance is neither randomly variable, nor influenced by temporally proximate ABs. It is the function of distinct external influences, including the batter’s own approach at the plate and the pitchers’ response to it.
You take stats for a bunch of guys who are 10 for their last ten in basketball and see what is their shooting percentage on the 11th shot. Is it closer to 100 percent or their career average?
You’ll need a pretty big sample size to come up with any sort of conclusive result, but I believe that even the hottest NBA players tend to drop that 11th shot at exactly the rate of their career average. I think Stephen Jay Gould wrote about this in “Full House”.
Same thing in baseball – you look at the stats for a bunch of players who’ve gone 4 for 4 and see what they hit in their next game. Chances are they’re not making people forget Ted Williams.
Don’t put words in my mouth.
We’ve talked here, for instance, about Olivo/Beltre’s approach and how that gets exploited by teams, and how they don’t/do adjust to that. It’s an entirely different topic.
I would love to know more about it, too. I would love to be able to prove it. But, like I said, those stats classes were many years ago. I remember being unsatisfied at the time, but not being able to prove it.
Look at it this way – we all agree that statistically analyzing defense is tough. The metrics are still pretty rough, though they are miles ahead of where they were just a few years ago. We can look at defensive metrics proven by several smart people – only 2 out of 3 say player x is awesome, and the third says he’s awful. Are the smart people wrong? I doubt it. Rather, we’re still learning how to effectively measure it.
Same for streakiness. Just because we can’t quantify it doesn’t mean it’s not there. There are tons more statistics available for hitting than defense, which should help, but maybe there’s more noise as well. You mentioned injuries. We could add in-game situations, or sepcialist relievers. Who knows what else?
I admit, I am now at the mercy of smart people when it comes to statistics. You guys at USSM do an awesome job of bringing together statistical awareness, logical analysis, hard work, and humor and good writing. Which is why I love reading the site. And I know that, “Just because I can’t prove it doesn’t mean it’s not true” is a pretty lame excuse, but please allow that it might be true. I would love to see a proof, but until then, I’ll rely on my original queasiness over the prof’s negative proof.
Maybe there’s hot, and maybe there isn’t. I believe it exists, but has not yet been quantified. Current proof is overwhelmingly in the court of “not”. And that’s fine. But, even assuming it’s true that there is no such thing as hot, I have not yet seen the proof that doesn’t leave some questions in my mind. Which is why I can’t agree with definitive statements like, “There’s no such thing as ‘hot’.” I’m a lot happier with “Nobody has been able to prove ‘hot’, and most theories discount it.” 😉
Thanks for listening.
btw – Colm, that is exactly the kind of study I remember from stats. Big sameple size, the 11th shot and all. But what about the 12th shot? And 13th? Maybe there’s a miss or two in there, but the group, overall, is way above average? How do we create groups? Well, unfortunately, they can only be created in hindsight.
These kinds of questions were not answered in my professor’s proof. Add in innjuries, or Brian’s external influences, plus the impossibility of accurately gauging where hot/not groups start and end.
It’s pretty tough to quantify, but the “11th shot” studies seem pretty simplistic and tautological. Of course any given shot, over time, will regress to a player’s average. That’s what average means. But think n-dimensional for a minute – are there groupings of shots that are hot and cold? Never seen that proof.
Again, I’m not saying hot and cold definitely exist. Rather, that a) the existing studies don’t satisfy my analysis, but b) unfortunately, I don’t remember enough high-order statistics to prove one way or the other. That’s all.
Derek, I apologize for giving offense, as my intent is not to put words in your mouth. I used a hypothetical quote as a rhetorical device to illustrate that the scenario I outlined could very well cause the type of temporary variance in the statistical measurement of a batter’s performance which you are now dismissing as “random variation” in the case of WF”B-B”B.
Perhaps a distinction should be drawn between an individual hitter in a given situation, and “hitters as a whole.”
Westfried. If you start talking about the 12th shot, 13th shot etc., you’re changing the game a little, since you can’t use statastical analyis without a consistently chosen date set.
Based on the two points that we know have been analysed (next shot, rest of career) my guess is that if someone does statistical analysis on the next ten attempts or five games or whatever data set, the results will be pretty similar – right around career average. That is what is meant by regression to the mean.
colm and westfried,
Try this. Use the MS Excel BINOMDIST function to test the binomial probability of any BA split against the player’s “true” (lifetime?) average.
=BINOMDIST(H,AB,AVG,TRUE) returns the probability of an AVG hitter getting H hits (or fewer) in AB at-bats (to measure the probability of a slump).
=1-BINOMDIST(H-1,AB,AVG,TRUE) returns the probability of an AVG hitter getting H hits (or more) in AB at-bats (to measure the probability of a streak).
Try it using a dice or coin flip model until you’re comfortable with the function of the function, then throw in some splits. Watch out, it can be a time sink. I haven’t yet read “Curve Ball,” so I don’t know if the “smart people” have validated this approach. But it makes sense to me. I think I can show that Ichiro is significantly more streaky than Johnny Damon.
There are other factors at play in that one, which I’ve read DMZ and Dave addressing before.
Chiefly: Ichiro (along with Luis Castillo) relies on singles to a far higher extent than other players in baseball.
Singles are subject to more random variation than other types of hits (in stat speak – a larger standard deviation) hitters who rely on them have a wider range of expected outcomes, which in this case means greater monthly swings in batting average for Ichiro.
As you suggest, it pretty much validates your theory that Ichiro may be more “streaky” than hitter X, but it does not imply that this streak is a result of his getting hot or cold, merely that a certain set of outcomes falls towards one or other extreme of the expected (fairly wide) range.
#135: If the idea is that being a singles hitter can explain “streakiness” in somebody who doesn’t go “hot or cold”, that turns out not to work.
Say a hitter doesn’t get hot or cold. I figure that means that at-bat is independent from the next, and the odds of a hit are constant — let’s say 0.333.
Hitter A may get that by drawing one from [strikeout, strikeout, home run] where hitter B draws from [groundout, groundout, infield single]. It doesn’t matter. Each one is in effect flipping a three-sided coin.
“How many hits does the guy get in 1000 ABs” is a binomial distribution, identical between the two hitters. Same standard deviation, sqrt(1000(1/3)(2/3)) ~= 15.
If a hitter shows “streakiness” by a statistical test, I don’t see what else we can attribute it to other than that he for some reason really is getting hot and cold.
But the range of variation in the number of singles that players hit from season to season is wider than the range of variation for home runs or doubles.
I.e. average taken as a whole across all of a player X’s AB and outcomes is 0.330. But:
Average change of hitting a single is 0.20
Average chance of hitting a double is 0.08
Average chance of hitting a homer is 0.05
Those last two averages (doubles and homers) have smaller standard deviations than the first.
Thus player Y who also hits 0.330 but whose single/double/homer line is 0.28/0.02/0.03 will see a wider range of probable outcomes than player X.
I think it’s the same phenomenon that explains the relatively wide fluctuations in players’ batting averages compare to their OBP or slugging numbers. I know that’s one of the reasons that stat geeks like to use those two stats in player evaluation. It’s not merely that they are valuable commodities, but also that the measurements are more accurate and a better indicator of future performance.
136. Eli.
In your example hitters A and B hit for the same average, but when you break it down, they won’t have the same STD DEV, B’s will be higher. The “streakiness” that B exhibits as a result of this is not a function of his becoming hot or cold, but a function of his much greater reliance on singles.
That said if you break it down across a sample size of 1000 ABs you aren’t going to see huge variations from the mean because you’re not talking about streaks any more, you’re talking about two seasons at a go. Nobody talks about a player getting hot for two years straight.
The argument began with people refering to Player X, or Ichiro, getting hot for 20 or 30 AB, or a month – much smaller samples where the greater STD DEV play havoc with our expectations and the impression we form.
I’m very tired and suspect at least some of the above is bunk, or at least poorly expressed. I also suspect I’m the only nutter still posting on this thread. Nonetheless, I agree with myself.
“Those last two averages (doubles and homers) have smaller standard deviations than the first [singles].”
I’m not sure if we’re talking absolute or relative deviations, so let me take it back to hitters A and B. Hitter A has a 1B/2B/HR line of 0.33/0.00/0.00; the unlikely hitter B has 0.00/0.00/0.33. You’re saying STDDEV(hitter A’s singles per 1000 ABs) is greater than STDDEV(hitter B’s HRs per 1000 ABs), or did I lose something in the translation?
“In your example hitters A and B hit for the same average, but when you break it down, they won’t have the same STD DEV, B’s will be higher.”
What that has to mean is that the singles hitter is going hot and cold from one AB to another, more than the HR hitter — and maybe singles hitters do that, I don’t know either way. Because if neither went hot/cold, their #hits/1000 AB would both have the same standard deviation. You can calculate it, it’s 14.8997 (if I did it right).
Here’s another way to look at it: consider the random variable “#hits / 1 AB”. Hitters A and B have this the same, it’s 1 one chance in three and 0 the rest of the time — Bernoulli(p=1/3). So the STDDEV(#hits / 1 AB) is the same between the two hitters. If a difference comes in when we go to STDDEV(#hits / 1000 AB), where does it come from? It must be from non-independence of trials (ABs).
Ahh, you latched onto the bit I thought was bunk too. STD DEV should be the same. It probably would be the same – over 1000 AB. But now for the bit of your reasoning that I think is flawed:
Over 1000 AB things will average out. Over a month they will not. Thus a singles hitter shows more “streakiness” in a month than a patient slugger. Of course if one insists on looking at 20 AB chunks as representative samples then streakiness is going to be there when you look for it. Not one .330 hitter in baseball is going to get 6.6 hits in every 20 AB.
No I’m going to go away and try to explain to myself why I think there is a greater STD DEV for singles.
A: Because there is, ya dummy.
So far this season major leaguers are averaging a hit 26.5% of the time, with a STD DEV of 0.76%.
They hit singles 17.6% of the time, with a STD DEV of 0.93%
They hit doubles 5.7% of the time, with a STD DEV of 0.45%
They hit triples 0.5% of the time, with a STD DEV of 0.21%
And they hit homers 2.9% of the time, with a STD DEV of 0.63%
There is a much higher STD DEV for singles than for any other type of hit.
Thus a hitter like Ichiro, for whom singles represent a far higher percentage of his hits than they do for an average player (80% vs 66%), should expect to see bigger fluctuations in his monthly batting average than a player who hits for a comparable average (say Pujols) but relies more on double and homers.
No hot streaks, no special circumstances, just random variation about the mean that does not exceed what we should expect.
The data you want is deviation for a single player across different time periods. The data you list is deviation across different players, yeah?
(Also, the deviations here are biggest for singles simply because the singles rate is the biggest number here. To talk about a .333 singles hitter vs a .333 HR hitter, that would need to be corrected for.)
About the 20 AB chunk: sure, there will be more variation than in 1000 ABs. It’s a knob to turn on the binomial distribution. For any unvarying .333 hitter, whatever flavor of hits, the standard deviation in 20 ABs is .105.
The data listed is for MLB players as a whole in the 2005 season to date.
Trying to look at variations in a single player’s stats across time is difficult because you start looking at very small numbers, which would have a wide range of expected outcomes even if there were not factors to consider other than binomial distribution. The opposing pitcher, opposing defence and park all contribute to variation in hitting stats. Trying to factor out all of those variables using such small sample numbers will not allow you to arrive at any useful conclusion.
E.g. Let’s ignore all the other factors and just use binomials. Looking at Ichiro as a .336 career hitter, and assuming that Ichiro will get 704 AB this season (which is what he had last year) there is about a 92% chance that Ichiro will go 13 for his last 20 at some point this season. There is also an 88% chance that he will go 2 for 20 at some point. On neither occassion does it really imply that he is hot or cold – these are just probable outcomes that we should expect to see at some point.
To get an idea of how misleading it is to look at small numbers let’s do another illustration with our virtual Ichiro:
The chance of this Ichiro going 1 for 15 at some point in a 704 AB season is about 78%.
The chance of him going 4 for 60 is less than 0.5%.
If he goes 1 for 15 it means NOTHING. If he goes 8 for 120 it means one of his legs has fallen off.
excellent, useful discussion, very conscious of how difficult it is to prove these propositions. my final two cents is that the free throw issue is a very poor analogy, nothing like the four at bats in the next baseball games. the factors that introduce non-randomness in Willie’s next performance are many and complex. that is why we love baseball, no? in free throw shooting, the game stops, and to a large extent muscle memory takes over. sure, there are other factors, but the extent of their influence is less.