At the start of last week (assuming you count Monday as the start. I do), I wrote two pieces for FanGraphs about relative pitcher strikeouts. They covered the worst and the best such rates in baseball history. They ended up not having anything to do with the Mariners so I didn't bother linking them here. So here's to rectifying that by creating a Mariners spin off! Will it be as short lived as the original series?
You can go read the original FanGraphs pieces for the back story if you want but the gist of the idea is that I wanted to examine pitcher seasons totally within the context of the rest of the baseball league for each season. Perhaps most noticeably, strikeouts have risen over time so what seems pedestrian to us now was actually dominance back in those terrible years of history.
What came out of the all-encompassing look was finding out that Ted Wingfield existed and that Dazzy Vance really deserves to be more of a household name among households that routinely discuss the greats in baseball history. He probably does not need to be mentioned amongst families that don't because that would just get weird.
I decided to get a little more advanced for the Mariner section because I love you all. I cleaned all of your gutters and bathroom tile just the way you like them to be cleaned. I have also included z-scores (since I canvassed the entire population of pitchers) in my charts and used that to decide what I think are the worst and best relative strikeout seasons in Mariner history. It's a slightly different definition from what I did with FanGraphs, and I'm not claiming it's a better one, but I like different.
The six most Mariner Mariners' seasons (min: 500 batters faced).
| Pitcher | Year | K% | Lg K% | Lg SD | Rel K% | Z |
|---|---|---|---|---|---|---|
| Bill Swift | 1988 | 6.2% | 14.6% | 4.0% | 42.5% | 2.09 |
| Glenn Abbott | 1979 | 4.8% | 12.4% | 3.8% | 38.9% | 2.00 |
| Ryan Rowland-Smith | 2010 | 9.6% | 18.0% | 4.2% | 53.3% | 2.00 |
| Jim Colborn | 1978 | 5.2% | 12.4% | 3.8% | 41.7% | 1.91 |
| Tom House | 1978 | 5.7% | 12.4% | 3.8% | 46.2% | 1.77 |
| Carlos Silva | 2008 | 10.0% | 17.1% | 4.1% | 58.5% | 1.74 |
Poor Ryan :( I went to six solely so that I could include Carlos Silva here. You were terrible, Carlos. You were bad at pitching, unpleasant to watch and reportedly a bad teammate.
Glenn Abbott was the guy who I thought would top this list and it still gets me that Abbott was the Opening Day starter in 1979, but Bill Swift's 1988 season edged him out if you go by z-score. Abbott still has the lower relative strikeout rate, so I'm comfortable giving the title of worst strikeout season to either of them.
And because it maximizes the point, here are the seven most outlying (biggest z score) strikeout seasons in Mariner history.
| Pitcher | Year | K% | Lg K% | Lg SD | Rel K% | Z |
|---|---|---|---|---|---|---|
| Randy Johnson | 1995 | 34.0% | 16.6% | 4.2% | 204.8% | 4.10 |
| Randy Johnson | 1993 | 29.5% | 15.0% | 3.8% | 197.3% | 3.84 |
| Randy Johnson | 1997 | 34.2% | 17.1% | 4.8% | 200.5% | 3.60 |
| Randy Johnson | 1998 | 31.1% | 16.7% | 4.1% | 186.2% | 3.52 |
| Randy Johnson | 1994 | 29.4% | 16.1% | 4.1% | 182.3% | 3.27 |
| Randy Johnson | 1992 | 26.1% | 14.6% | 3.6% | 179.2% | 3.23 |
| Randy Johnson | 1991 | 25.7% | 15.2% | 3.9% | 168.6% | 2.71 |
Oh, Randy.
You might not be intuitively familiar with statistics (you should learn, even at a beginners level, it's a vastly helpful knowledge tool) and thus might not grasp how unlikely a z score over four is. With a normal distribution — of which these seasons closely approximate — a z-score of three represents the 99.9th percentile. A score of four or more (in either magnitude) is very rare. Randy Johnson was a special pitcher.
0 recs | 25 comments
HOF as a Mariner?
Or a Diamondback? Or even an Astro/Colt .45??
Ackfan - February 22, 2012
Has to be as a Diamondback.
World Series, Cy Young, and perfect game with them if I’m remembering correctly.
Cascadian Man - February 22, 2012
Expos
Eric Wedge's Mustache - February 22, 2012
Diamondback.
Randy spent eight years with the D-Bags compiling a 45.1 WAR and four CYA’s. Easily a Diamondback.
Call Jack. I'm on my way. - February 22, 2012
Yeah, looking more at the numbers, it isn't even close. Just wishful thinking I guess.
Ackfan - February 22, 2012
But.
I would still love to see him in Cooperstown with an M’s cap.
Call Jack. I'm on my way. - February 22, 2012
Please use the reply button when responding to comments.
Eyebrows - February 22, 2012 via mobile
1995. What a season.
Randy Johnson struck out 294 batters in 214.1 innings.
Tim Belcher, Chris Bosio, Salomon Torres, Andy Benes, and Dave Fleming struck out 297 batters in 532.1 innings.
Kenneth Arthur - February 22, 2012
"1995. What a season." I guess that's what they call an understatement.
Kenneth Arthur - February 22, 2012
How about "1995. It was a seson."
Ballard Erik - February 22, 2012
Whoops. Or a saison. Delicious!
Ballard Erik - February 22, 2012
D'awww.
We love you too.
Also holy shit Randy Johnson. I missed out on so many awesome players. :(
Aussie Mariner - February 22, 2012
So who has the highest z score among M's pitchers not named Randy Johnson?
Paul AB - February 22, 2012
1982 Floyd Bannister
Matthew - February 22, 2012
I never would have guessed.
Paul AB - February 22, 2012
Wow, really?
I think most people would have guessed something.
Matthew - February 22, 2012
A Randy Johnson fastball could rip your heart out and show it to you before you died.
TJDirk - February 22, 2012
"You cleaned my gutters"
Phrasing.
nathaniel dawson - February 22, 2012
Very interesting
Randy Johnson was truly amazing. But I’d caution you to not read anything into those probabilities you cite. The distribution is certainly not normal. The frequency of tail events you cite even right here makes that obvious. I’m sure the distribution is highly leptokurtic. These events are not as rare as you make them seem.
phineasd - February 22, 2012
That's actually why I said "approximates normal".
I’m curious though how you’re so sure. Did you look at the data yourself? I don’t know how rare I “made them seem” since that’s pretty dependent on each person’s interpretation, but Randy’s 1995 season is actually exactly the 99.9th percentile of the sample I used.
Matthew - February 22, 2012
It can't be normal
The z-scores are describing the distribution. The frequency of high z-scores makes the excess kurtosis evident. If you have 1000 pitchers a year in your sample, you would expect to see 1 of them more than 3 standard deviations above the average if the distribution was truly normal. The fact that there are far fewer than 1000 pitchers a year that throw meaningful innings yet you regularly see observations three and four standard deviations away from the mean makes this undeniably non-normal.
By saying they are less rare than you make them seem, I mean that you should expect to see these observations far more often than 0.1% of the time. Using z-scores to evaluate the rarity of observations in a non-normal distribution doesn’t make any sense. Your z-scores aren’t telling you how rare or unlikely your observations are, they are cluing you in to the non-normality of the underlying distribution.
phineasd - February 22, 2012
The frequency of high z-scores?
Do you know the sample size? Isn’t that a needed bit of information to decide how often something is occurring?
As for the z-score part, I took this
(Emphasis mine)
To figure that they do make some sense as a way to rank across differing means and variances. Is that incorrect? I see how it reads like I’m using them to state the rareness of the event, but I was merely trying to provide context with a somewhat related example of a normal distribution. I added and used the standardized scores only as a ranking method.
Matthew - February 23, 2012
Honestly I don't think anyone who can't find the reply button can possibly be that bright
Graham MacAree - February 23, 2012
Is this comment really necessary?
Or is it just to reinforce what a dick you are?
phineasd - February 23, 2012
Just a guess
If we have 26-30 teams and say 16-17 pitchers per team each year, we’re talking 400-500 pitchers in the sample each year. Probably makes sense to have some minimum number of innings pitched to be included as well to reduce estimation error. In order to reach 1000 pitchers in the sample each year, you would need an average of 33 pitchers from each team to throw meaningful innings.
I think using the z-scores as a ranking metric is a nice idea. You can certainly calculate a z-score if the distribution is non-normal. But it’s no longer meaningful to match up that z-score to a percentile, those percentile numbers you cite are based upon assumptions of normality. That’s why I say don’t read too much into the probabilities. It’s certainly fair to say “Randy Johnson’s 1995 strikeout rate is 4.1 standard deviations above the mean strikeout rate for all major league pitchers.” But you can’t assign a percentile or probability to that value from standard normal z-scores.
phineasd - February 23, 2012
You must Login with your SB Nation account and be a member of Lookout Landing to post a comment.