Consistency

[maddash]

Is there a stat that measures consistency in a hitter? That is, a stat that can take two players who end up with the same numbers, but differentiate them based on how they obtained those numbers - through streakyness or consistency.

Thanks

[mcgee51taguchi99]

No one is going to hit .333 by going 1 for 3 each and every day. I think you'll find that any player is "streaky." 4-4 is a streak, and 0-4 is a streak.

Therein lies the problem: clearly and concisely defining a streak. Once that is done I'm sure someone can quantify it for you.

Breaking down a player's numbers by month is generally the best you're going to get as far as analysis . . . barring time missed due to injuries of course.

[mcgee51taguchi99]

Alright my crappy mathematical vocabulary and ability is going to shine through here . . . but couldn't you, say, doe this for a hitter's avaerage by comparing a player's monthly totals using a stats equation where you compute the deviation from the mean (the season average) each month? There will be months where production is under the season average and the deviation is therefore negative, so take the absolute value (thus making the values positive), add all the deviations and divide by the total months. The lower number is the player who stays closest to his season average.

Does this terrible explanation make sense to anyone?

[maddash]

No one is going to hit .333 by going 1 for 3 each and every day. I think you'll find that any player is "streaky." 4-4 is a streak, and 0-4 is a streak.

Therein lies the problem: clearly and concisely defining a streak. Once that is done I'm sure someone can quantify it for you.

Breaking down a player's numbers by month is generally the best you're going to get as far as analysis . . . barring time missed due to injuries of course.

I agree that defining a "streak" isn't an easy think to do, but you don't really have to. I would think that someone who knows a bit about stats can simply create a scatter plot of a player's offensive performance from day to day, and then see how close that player stays to his mean. The more he strays, the streakier he his. Obviously, no player is going to stay at his average level exactly, but some will be closer than others.

[maddash]

Alright my crappy mathematical vocabulary and ability is going to shine through here . . . but couldn't you, say, doe this for a hitter's avaerage by comparing a player's monthly totals using a stats equation where you compute the deviation from the mean (the season average) each month? There will be months where production is under the season average and the deviation is therefore negative, so take the absolute value (thus making the values positive), add all the deviations and divide by the total months. The lower number is the player who stays closest to his season average.

Does this terrible explanation make sense to anyone?

Ah, deviation from the mean, I think that's what I was trying to say in my post above.

...man, I'm not looking forward to having kids and helping them through math class. And I was pretty good at math back in the day.

[New Pagodi]

Alright my crappy mathematical vocabulary and ability is going to shine through here . . . but couldn't you, say, doe this for a hitter's avaerage by comparing a player's monthly totals using a stats equation where you compute the deviation from the mean (the season average) each month? There will be months where production is under the season average and the deviation is therefore negative, so take the absolute value (thus making the values positive), add all the deviations and divide by the total months. The lower number is the player who stays closest to his season average.

Does this terrible explanation make sense to anyone?

What you seem to be suggesting is to use a concept analogous to variation - for example defining lack of consistence to be the absolute value of the one’s daily batting average minus batting average summed over the whole season. However this doesn’t seem to give the results we would expect.

For Albert Pujols, pretty much the model of consistency, this method gives the number 2.295. Here’s a plot of what’s going on (sorry about the crappiness of the plots, but it’s late)


For Jim Edmonds, the model of inconsistency, we get 2.393 which is pretty close to Albert’s number.

For Juan Encarnación we get 3.646. This is a larger number than for Edmonds or Pujols, but it is due to the discrepancy between his BA and his daily BA for the first several weeks of the season rather than what I would call inconsistency.


And for Aaron Miles we get 3.615. This is close the Encarnación’s number but for the opposite reason.


Based on these four players, it would seem that this approach seems to assign large numbers to players that have extended slumps instead of those that are truly inconsistent.

[vinsanity]

Nice plots, but I thinkt he problem is sample size there.

What if you plotted monthly averages against career averages?

It looks like you have you daily avg compared to season averages, and that being the case an 0-4 day has far less impact with 300 AB's than it does with 50.

But I can't tell sooo I don't know.

[JL21]

This book, which you might be able to find at your local library, has a huge section about player streaks.

http://www.amazon.ca/gp/product/0387988 ... e&n=916520

[New Pagodi]

Nice plots, but I thinkt he problem is sample size there.

What if you plotted monthly averages against career averages?

It looks like you have you daily avg compared to season averages, and that being the case an 0-4 day has far less impact with 300 AB's than it does with 50.

That’s a good point. Another mistake I made was to keep the whole thing analogous to variation, I should have divided the whole sum by the number of days played. But since the effect of sample size will be roughly the same for everyone, I think the general point is basically right.

The original question was

Is there a stat that measures consistency in a hitter?

This was later narrowed to can a statistic like variation be used to measure consistency. I was simply trying to say that it doesn’t seem so. Variation seems better suited for detecting players with extended hot/cold spells. As to the original question, of is there some other statistic that will work, I don’t know.

[New Pagodi]

This book, which you might be able to find at your local library, has a huge section about player streaks.

http://www.amazon.ca/gp/product/0387988 ... e&n=916520

Thanks. When I finally get someone to employ me (or a library card), I’ll be sure to read that book.