If you have the full database, just run regressions with OBP and HR added in as covariates along with their interaction term, OBP*HR. That way, you can analyze your results while controlling for one or both.

Draft Giancarlo Stanton and don't draft any Mets.

As someone who is about to start their PhD in Medical Physics in a couple of months, I am very sheepishly asking for an ELI5

Don't draft any Mets? Have you stopped following baseball? I want every Mets pitcher.

What's with people on this site having names beginning with Y and getting PhDs in Medical Physics. Like a correlation of 1. There's some stats for you.

Play in a prospects league and draft me.

Hmm... How about userID=2962 ? I hear he was a PhD candidate in Medical Physics.

A correlation of 1 in the other direction.

Your carefully selected sample consists of TWO people, abge. Well done. And it's probably because medical physics is the easiest kinda physics to do. :) Just kidding. It's not.

Yonni, what I am suggesting doesn't really work unless you actually have access to all the individual observations in the dataset, not just summary stat. You would regress your dependent variable (I'm assuming point scored or something) on your controls like age, race, OBP, HR, and whatever else, so you can analyze the predicted effect of changing one variable while holding others constant (ceteris paribus interpretation). If you don't have access to the whole dataset and a stat program like SAS, SPSS, STATA, or R (free), then it's meaningless.

Yeah, I just have the final summary not the meat that goes into it.

First off, use ZiPS if you can. Steamer is a steaming pile of crap. Second, there's not a lot you can do without any kind of range of projection. If you had, say, a 25th and 75th percentile projection in addition to the 50th they give you, you could do some cool risk/reward stuff, but otherwise you'll just have to trust and/or tweak the projections as you see fit.

Are the projections really going to be worse at protecting some stats than others? I haven't seen that. There might be more absolute difference on some of them, but the z-score approach should factor that out.

Third, draft Brad Miller.

Also, we should definitely have a webDip fantasy league again.

From what I've read, Steamer performs better than ZiPS in a lot of things which is why I use it.

And some stats are definitely predicted a lot better than others. It makes sense too. RBIs, HRs, Rs, have more randomness and smaller samples than OBP. For pitchers, things like wins and ERA are less predictable than Ks.

We should for sure have a webDip fantasy league again. I should join this year.

Sure, but those things are also going to have more variance in general. I think your only solution is to value some stats higher.

I only played webDip fantasy once a few years ago. I think I won? Anyway, we should definitely do it. Do you want to organize or should I?

The issue is that as of the 16th of this month until mid to late April I'm going to be travelling so I'm not sure if I'll have time for a draft

It's not *that* terrible to set a list and miss the draft.

That's how most drafts end up anyway. You have someone you want and someone else snags them, so you have your backup.