Top SPs are drafted below value. Definitely. It’s not all pitchers in general as it starts stabilizing in the middle rounds.

And yes on #2.

I got Jemile Weeks in a few leagues b/c of a similar gut feeling

The most insight that I gathered from that is
1) SPs are underdrafted (as you stated, due to strategy)
2) Comparing averge draft pick to actual projected value, you can skip a Rickie Weeks and 10 rounds and get the same value from Jemile Weeks

I’d like to be able to factor in low-medium-high projections, but I don’t think that would get anywhere valuable. That’s the problem with others’ rankings — they are probably based on projections / last year’s numbers, and a gut feeling that they could break out. I’d rather just rely on numbers.

man-oh-man .
how boring it would be to have the job that requires you to
compile pitch-type stats for every freakin’ pitcher .

glad that i don’t have to do it .

have you consulted this site for your risky pitchers’ report ?

http://brooksbaseball.net/

I’m curious how you consider the predictions give you a 10% edge vs. your competition? I am not trying to be a jerk here – honestly trying to understand where you’re coming from and, selfishly, planning to use that edge myself muhahahhahhaha…

Also, I think you’re in the minority regarding how people use predictions during a draft. I think a large majority of owners (at least, serious owners who good fantasy players need to consern themselves about) use some kind of prediction vs. expected targets comparison during their drafts to help guide them toward players whose skill sets they need (i.e. do I draft Jayson Werth or Cameron Maybin, for example).

It’s taken me a couple of years of playing around with predictions to get to this point but I agree with your final assessment. Predictions are so scattered and categories are so tight that using them as a drafting guide is actually a disadvantage.

That being the case, should we tout and encourage other owners to use predictive tools as a Bible?

Sorry….my evil side coming out. It happens.

But just b/c projections have large margin for errors doesn’t make them useless. Even if all the work I do just creates a 10% edge vs. my competition, that’s still helping me towards winning the league.

I just wouldn’t sweat trying to hit, say, a HR target when the reality is that almost all the teams will be separated by a small enough margin that everyone could finish many places above/below their projection – hence, my point that preseason standings are pretty much useless aside from getting a gauge at your team’s strengths/weaknesses.

Pat’s Steaks – you think I’m from Philly?

I use ZiPs and Steamer for the rates and Fantistics + Rotochamp for playing time estimates. I make some edits to playing time estimates as I see fit – the only edits I made to the rates from ZiPs and Steamer was to increase the ERA/WHIP for relievers turning to starters (tends to be +1.00 in ERA and I think 0.10 in WHIP).

This offseason I’m going to test the various free stat projection services – when I do that, I’ll test using my preseason playing time estimates so I can test how my ZiPS/Steamer concoction does compared to using a single source.

Thanks again for the kind words -
Rudy

Looking forward to the results from 2012 as well to see how similar it is to the 2011 results. Also, I changed my stat and playing time sources so hoping that helps.

After reading through this the question that comes to mind for me is:

What level of stat variance between predicted stats and actual stats has to exist for them to be worth using to predict?

If I told you “I can guarantee the stats you’ll receive in my prediction set are 80% accurate” you’d probably agree that sounds like a prediction off which you can work. For example, lets say I used my algebraic super duper formula to predict Albert Pujols averages 37 HR. That means I am guaranteeing he’ll be between 31 and 45 HR. While it’s probably correct, that wide a range is essentially useless as a predictive tool.

In my 10 team 5×5 league it took 277 HR to gain 8 points last year. Tops was 315, bottom was 227. These ranges are reasonably similar to the previous few seasons.

Working backwards, I use my prediction set to guide me to draft a team which predicts out to 277 home runs. That’s my start point. If I get to a predicted 277 home runs with 80% confidence that means my team could hit anywhere between 222 and 333 home runs!

That level of confidence would leave me predicted to finish first, last or anywhere in between. That level of confidence is therefore useless. For the record, that level of confidence leaves me first, last or anywhere in between in every counting category but Saves (likely because one or two owners punt it).

Even if my predictions are 90% accurate that would leave me anywhere from 249 to 305 homers which would leave my point total from 2nd to 9th. That level of confidence is essentially the same – a distinction without a difference.

The way I see it, in order for predictions to actually have meaning in the standings, I need 98% accuracy. If I hit my predictions with that level of accuracy, I’d leave my draft knowing I would be between 6-8 points in Home Runs. As insane as it sounds, 98% is the level of confidence needed for predictions become useful as drafting guides.

I guess the point I’m belaboring here is that the margins between 3rd place and 8th place in most categories are so tight that predictions under 98% accuracy could leave you anywhere within the range. We are interested in being able to predict how many points we’ll get per category (in roto, at least. H2H is different) In my opinion, such an impossibly high level of confidence required to meet that task renders using predictions essentially useless as drafting guides (and really, that’s what we’re using them for).

I think fantasy baseballers use them as a crutch. They either make them up on their own, use some formula to create them or find some publication containing predictions with which they agree. The owners then shape their drafts around these predicitons. This explains the phenomenon which takes place of the end of every draft – the part where half of the owners claim their predictions show them in first place.

To sum it up, until we can get to 98%+ accurate on predictive stats we might as well not use them. And since that’s not happening any time soon….

Seriously, extremely insightful piece. You are truly the yin to Grey’s yang. Looks like lots of studying for you back in the day – you weren’t just hangin’ at Pat’s Steaks every night, correct?

Questions: I’m pretty sure you base your projections on ZiPs and Steamer, but how do you determine your final projection line for each player? And have you analyzed these projections to see how closely they predict actual player performance?

Thanks for all the hard work. You make the world a better place for many reasons, man.

Take care,
Mike D.

It makes me giggle a bit to see ESPN do so poorly.

I wanted to throw out a variable. Since the leagues are hosted by ESPN, and therefore use their default rankings, there was bound to be more then a handful of teams that missed the draft, ended up with a pile of ESPN pre-rank garbage and through trades and FA pick-ups created a more Grey/Rudy-esque team. Assuming the Grey/Rudy team then performed OK, might it be true that it actually added a bit of positive correlation to the ESPN ranks? Just a thought. There’s a ridiculous amount of variables to consider when you do these kinds of analysis, but I’ve got to say, you seemed to account for many of them. Nice work, keep them coming.

Can’t wait for this years results.