If you're like me, Fantasy Baseball is one of those times of the year in which it's most difficult to marry my sabermetric view of things (value!) with a fun and rewarding game.
Problem 1: Roto or H-2-H?
The first issue is whether to choose a rotisserie or head-to-head style game. Roto leagues have a set number of teams and a set number of stats, each of which you can 'win,' scoring a number of points depending where you rank in each stat. Most points wins.
Then there's H-2-H, where there's a set number of teams that play round robin, one against the other each week throughout the season. It's the same idea as roto, but broken down into a weekly game. Instead of attempting to win each stat for the season, the goal is to win the most stats against your opponent, scoring a win. Best record wins.
The choice is obviously yours, but word of warning: if you're going with rate statistics to honor your sabermetric view of things, roto leagues can get awfully boring once you get about halfway through a season. I prefer H2H for exactly that reason.
Problem 2: Which Stats?
On one hand, counting stats are meh, but on the other hand, rate statistics aren't very fun to compete with after all your players have regressed to their mean (which is why roto leagues can get terrible after you're about halfway done).
Assuming you go H2H, the key is to choose the right counting stats and combine them with the right rate stats that don't count the same thing several times. So if you're counting doubles, triples, and homers, think twice about including SLG, which would reward the winner twice. Similarly, it's best to pick one of the better pitching metrics rather than all of them, which tend to reward the best squad rather than individual skillsets.
Personally, I like the standard counting set (singles, doubles, triples, and homers), with OBP and Net Stolen Bases. My idea is to capture essentially all offensive production without doubling up. On the pitching side, I like IP, QS, WHIP, CG, Ks, Net Saves. If you try to do more rate stats, you might make your saber side swell with pride, but you'd also risk unbalancing your game by counting the same things a bunch of times. So, go with one or the other -- but if you go with rate statistics while not doubling up, you'll have far fewer to play with (which is kinda the point).
Problem 3: Balancing Hitting/Pitching (and drafting accordingly)
There is a problem that most people don't really acknowledge, though -- and could give you a leg up on your friends. Even if you (or your commish) attempted to balance the number of stats on both sides, most leagues have a natural imbalance that favors hitting.
Take my preferred stats, for example (1B/2B/3B/HR/OBP/NS | IP/QS/WHIP/CG/K/NSV). Six on each side, if I'm counting correctly. The problem is that one of them, CG, is extremely rare, and another, QS, is a natural outcome of winning the IP battle. My instinct with CG is to ignore them, because you can't "try" to draft pitchers to win CGs on a weekly basis.
Then there's IP and QS (an imperfect but worthy stat), which play hand in hand but reward the same thing. The result is an unbalanced game that favors sluggers and innings eaters (and note that if you're ignoring CGs, or in crazy leagues, SHOs, we're down to a 6-5 stat imbalance between hitting and pitching).
Now, I could be wrong here (I'm famously terrible at fantasy baseball -- oops, did you need to know that?), but my strategy for a draft would be to go for sluggers first, going after doubles and homers, hoping to get the guys that also get walks with them. On the pitching side, I'm keeping in mind that there's really only 5 winnable stats per week and that players that eat a lot of innings are the important part. If you're crazy you could even throw out even trying for saves and stock that slot with another starter, which is what I do.
The idea, at least based on what most leagues look like, is to take advantage of an unbalanced scoring system. If you're running a league, your goal should be to limit that as much as possible. In the end, you'll be left with a game that is very close to the original idea -- mostly counting stats -- but something a purist like yourself can appreciate.