Early Keeper Musings

[harmon]

Well, consider this - Liam Hendriks and Josh Hader are two of the most dominant closers in the game right now. Hendriks BABIP has been .250ish in both '20 and '21; Hader's has been .230ish for a long time. Both have exceptional stuff and their BABIP's are likely repeatable (Hader's has for four years now). Barlow's stuff is very good, but he's not in either of these guys class.

Sure, if you can keep him for essentially free, absolutely. But if you're cutting a hitter of some quality (which your earlier post indicated), no way.

But even with regression he's still putting up Haderesque numbers.

[Ken]

But even with regression he's still putting up Haderesque numbers.

"Regression" would suggest that his numbers going forward would be the typical standard numbers - in this case a ~.292 babip. The concept of regression to the mean does not imply that an atypical statistical sample "slides" back towards the mean, rather that we should expect the mean values going forward, *not* the illusion that can created through a small sample.

We envision the slide concept as we interlace the typical sample on top of the existing small sample numbers. But that's an illusion only, the forward looking expectation in mean regression is simply the mean itself.

[Gregg]

"Regression" would suggest that his numbers going forward would be the typical standard numbers - in this case a ~.292 babip. The concept of regression to the mean does not imply that an atypical statistical sample "slides" back towards the mean, rather that we should expect the mean values going forward, *not* the illusion that can created through a small sample.

We envision the slide concept as we interlace the typical sample on top of the existing small sample numbers. But that's an illusion only, the forward looking expectation in mean regression is simply the mean itself.

I understand (or at least I think I do) what you are saying. If I just take what you have written at face value it would seem that small sample size is rendered useless. That does not seem quite right. How do you incorporate small sample size into projections? How would we even define small sample size?

[chancellor]

I understand (or at least I think I do) what you are saying. If I just take what you have written at face value it would seem that small sample size is rendered useless. That does not seem quite right. How do you incorporate small sample size into projections? How would we even define small sample size?

Question 1: Incorporating small sample size into projections - some sites have a pretty decent projection system for including minor league stats and how they'd translate. I thought Jason wrote something up on that on here, but I can't find it. I'll look at those, but also include some subjective judgement. Using Barlow as an example, do I believe his stuff and minor league stats merit a belief that he can pitch to a better than average BABIP? And if so, by how much?

Question 2: For relief pitchers, I've always thought about 100 IP is the good sample size threshold, but I'd sure like to hear from experts on that.

[Ken]

I understand (or at least I think I do) what you are saying. If I just take what you have written at face value it would seem that small sample size is rendered useless. That does not seem quite right. How do you incorporate small sample size into projections? How would we even define small sample size?

I was primarily rejecting the idea that a small sample that regresses "still" means anything for the future unknown data. Obviously there are conclusions you can still draw, depending on what type of stat you are looking at, but when looking at regression the logical fallacy occurs when using "still" with the future information.

If I flip a coin 10 times and it comes up all heads. I can say that I expect regression to the mean across the next 90 flips. That means that I expect the TOTAL average to get closer to 50%. I'm commenting on where I expect the total to move. Right now the total is 100% heads and I expect it to approach 50%.

But I still expect 50% for future coin flips (just like I did at the start).

[cavebird]

Well, consider this - Liam Hendriks and Josh Hader are two of the most dominant closers in the game right now. Hendriks BABIP has been .250ish in both '20 and '21; Hader's has been .230ish for a long time. Both have exceptional stuff and their BABIP's are likely repeatable (Hader's has for four years now). Barlow's stuff is very good, but he's not in either of these guys class.

Sure, if you can keep him for essentially free, absolutely. But if you're cutting a hitter of some quality (which your earlier post indicated), no way.

The proper move is to trade him with another keeper for a better keeper. Sure, you shouldn't keep him (Ken is completely correct here) but if you can find someone who like him.... Let's face it, closers fail constantly and player keep trading for them in the off-season. Do it.

[onejayhawk]

The proper move is to trade him with another keeper for a better keeper. Sure, you shouldn't keep him (Ken is completely correct here) but if you can find someone who like him.... Let's face it, closers fail constantly and player keep trading for them in the off-season. Do it.

That is certainly an approach to consider, especially since Hader is on the payroll.

My one hesitation is that it is a holds league, so quality set up guys have real value. Everything I am getting on Barlow is that he is a good pitcher in addition to being the closer.

The league's informal winter meetings are during Thanksgiving weekend. I should have an idea of his trade vale after that. I still need to figure out if Carson Kelly and Jake Meyers are worth their option cost.

J

[Ken]

That is certainly an approach to consider, especially since Hader is on the payroll.

My one hesitation is that it is a holds league, so quality set up guys have real value. Everything I am getting on Barlow is that he is a good pitcher in addition to being the closer.

The league's informal winter meetings are during Thanksgiving weekend. I should have an idea of his trade vale after that. I still need to figure out if Carson Kelly and Jake Meyers are worth their option cost.

J

I'm not confident that a holds league even moves the needle. How many of the top ~50 holds leaders were owned at the start of 2021? I bet the number would be extremely low relative to other categories.

Luke Jackson, Anthony Misiewicz, Richard Bleier, Tim Mayza, Mike Mayers, Jake Brentz, Garrett Whitlock, Codi Heuer, JT Chargois are not exactly household names. It's a category that you really don't need to pay for in most leagues. And the cost of a keeper spot is relatively high in your case.

[onejayhawk]

Good point. I actually own Tim Mayza and his is not an option I considered. Yet, he was more valuable this season than Barlow. The main distinctions are the role and the age.

The cost of picking up the option is middling. Think about redraft pick of #275 in a basic 20 team league, ie out of 460.

When I last posted, I had not delved into the position players at all. Some I assumed to be solid I may reconsider. Option prices vary, but all of the following are close calls at this point--Anthony Rendon, Gleyber Torres, Jo Adell (all three on last option), Jeimer Candelaria, Carson Kelly. I can drop them all and still have an offense of Ronald Acuna, Luis Robert, Cedrick Mullins, Wander Franco, Brando Lowe, Willy Adames, plus the additional players I get for the options not exercised. I am seriously considering exactly that and if so, Barlow, Lynch, etc. go too.