The RotoWire Blog has been retired.

These archives exist as a way for people to continue to view the content that had been posted on the blog over the years.

Articles will no longer be posted here, but you can view new fantasy articles from our writers on the main site.

Charging the Audit, Avoiding the Pitfalls

A reminder - Chris Liss and I are moving this to the blog – something we should have done years ago. No more waiting for four replies to publish, no more editing in the times of publication so painstakingly – just the ongoing conversation. And it's no longer behind a paywall.

Chris (8:22 pm PT, Sunday, February 22)

I wrote up my "Beat Chris Liss" league here. I was reasonably happy with it, though I had the handicap of going against a few guys who listen to the show a lot and for God knows what reason were swayed by my arguments for some of "my players." It's the same thing that happens in the Vegas football draft when most people are drafting off the newly minted RotoWire magazine rankings (which I made), and I have to draft a whole different set of players than I do in other leagues. For example, I like Zack Greinke more than most, but he went at 3.3 (27 overall), and Bryce Harper was gone before I picked in the third round too.

So while I wanted to be aggressive in the context of my overall strategy which I detail in the post, I ended up value taking more than I usually do. One player I took who I didn't plan to was Billy Hamilton in Round 4. Drafting out of the 11 spot I knew I'd be stuck at the 3/4 turn with players I didn't love, but instead of playing it safe with ADP or whatever Steamer liked (probably David Ortiz - Steamer loves him every year), I swung for the (inside-the-park) home run. I also drafted Machado (at pick 134). But I got Ortiz at 107, Freddie Freeman at 59 and Jayson Werth at 155, who were just guys who seemed to fall too far.

As for avoiding busts, I agree that's certainly as important as finding sleepers, but my comment wasn't so much we should help people on that - though we should - but more that on the whole there are more busts than profitable guys. If you look at the preseason top-10, top-20, top-100 or whatever, that asset class, taken as a group, will almost necessarily lose money. That's because lots of players from outside the preseason top-100 displace the guys in it. But it seems a disproportionate number of preseason articles are focused on guys people think will bounce back or break out. I think one reason for that is people don't like it when you rip the players on their teams, so you get more pushback. It's like betting on the "Do Not Pass" line at a crowded craps table.

Another reason is sleepers and breakouts are mostly low-risk predictions. If I think Kolten Wong or Machado will have a huge breakout, and they don't, maybe you wasted a mid-round pick on one of them, but we both knew it was, if not a long shot, at least less than a 50/50 proposition. But if I say avoid Anthony Rendon, and he simply matches (or comes close to) the numbers he had last year - which is more likely - I look like an idiot.

Third, if I pick an aging mid-level player like Chase Utley to be a bust, no one really cares. So to do a legitimate bust article, you have to go after players from the early rounds, and those are necessarily good ones against whom most do not want to bet. But had you picked Prince Fielder, Chris Davis, Carlos Gonzalez, Bryce Harper, Ryan Braun, Cliff Lee, Yu Darvish, Hanley Ramirez, David Wright, Justin Verlander, Joey Votto or Jay Bruce, you would have saved people who took your advice a lot of grief. And you can see from that long list, you had a good chance of connecting on your highly-ranked bust list. (Unfortunately, mine was Andrew McCutchen, Adam Jones, Felix Hernandez and Adam Wainwright last year.)

But in that spirit, here are a few players I won't have on too many teams unless they drop well below their ADPs: Ian Desmond, Michael Brantley, Corey Dickerson, Dee Gordon, Cole Hamels, Jason Kipnis, Yoenis Cespedes and Evan Gattis. But based on my track record from last year, you might want to bump all those guys up five spots.

Jeff (12:35 pm PT, Sunday, February 22)

I've been running smack dab into the ADP Vortex in my current NFBC Draft Champions League. At least four times now I've missed out on my outlier players who didn't make it back to me. I've fallen prey to "needing" a commodity rather than just picking the guy I like best. I'm in the 5-slot in a 15-team league, so there's no double-up possibilities either, which is something for which I should have accounted when making the picks. One of the things I really respect from the NFBC crowd as a whole is their willingness to stick their necks out on "their" guys. Kris Bryant went in the 10th round in LABR, but in this draft he went at 7.11, as one example.

So that stops now. I'm taking my guys and not crossing my fingers, hoping they come back. It's going to cost me a little at the scarce positions and probably with a starting pitcher or two, but it's a lot harder to find the hitters I like anyhow. One example - Manny Machado's ADP has been falling to around 150 in the NFBC, but I have him projected to be around 96 after going through my "audit" and just acted on that, taking him at 8.11, or No. 116 overall. It's well ahead of ADP but I don't care - the central question now, at least at this point in the draft, is: "Am I going to be upset if I don't get Player X?" If the answer is "yes," I take him.

I believe that's close to your "hit winners" construction - perhaps not best defined via formula or ranking, but under a "players I like" idea. If I'm not taking my guys because I'm worried I don't have them properly ranked or might be reaches, why am I in this business anyhow?

There's one other concept I wanted to ask you about, however, that we also discussed on the show. And that's the notion that the best advice we can give is not finding sleepers, but avoiding busts. And by "busts" we're not referring to the obvious guys, but the genuinely good players that for whatever reason we think will fall short. I believe our instincts have been finely honed from the accumulated experiences of our leagues, even if we can't always articulate our reasons for our wariness.

Do you have a related maxim for avoiding busts? "Don't hit double-faults?" And were you able to hit a lot of winners in "Beat Chris Liss?"

Chris (2:35 pm PT, Friday, February 20)

Two statisticians go hunting, one shoots five feet to the right of the deer, the other five feet to the left and say "Got him!"

It's funny because I was starting a blog post on a similar topic but my work got interrupted by the insane two-day League of Leagues draft we did this week. Essentially, it's hard to do (or even consult) projections and not wind up with blind spots in precisely the areas you point out.

As Baseball Prospectus' Rob Arthur recently noted well-known projections systems do well on average, but they all miss the breakouts and collapses - and that's almost necessarily so, as they're built to regress everything to the mean. If you're starting with three-year averages - or any other formulaic baseline - the same thing will likely happen to you.

Generally three-year weighted averages modified by age, park factors, projected playing time and health probably do a decent job, but if the player is 2014 Michael Brantley or Corey Kluber, it's not going to help you much. And when you look at fantasy teams that win their leagues, they usually have a couple Klubers, Brantleys, Anthony Rendons or Dee Gordons. It's one thing if you're investing in stocks and content to get a solid return on investment each year, beating the market averages, but quite another if you're aiming to win your 14-team league - or even more - cash in the NFBC overall.

The other trap I fall into - and I wish I could avoid it - is getting sucked into the gravitational pull of Average Draft Position (ADP). Ideally you want to be aware of ADP so you don't take a player in Round X who is likely to be available in Round X+1, but the price you pay for that knowledge is steep, as it's almost impossible to use ADP only for that purpose and not have it seep into your beliefs about how players should be ranked. The "wisdom of crowds" - the aggregate rankings of everyone that create the market price for these players - has to be better than your idiosyncratic guesses, right?

While that's likely to be true in cases where everyone was reasonably informed and independently came up with their own rankings first, it's not the case where people are being influenced by each other's rankings, and especially where the ADP itself becomes a source of that influence, drawing divergent opinions toward it like a black hole. In that case, you're often better off sticking to your outlier opinion which takes into account hunches based on experience. If pressed, you might acknowledge any particular hunch wherein you diverge sharply from both the market price and the more "objectively" derived systems is a less likely, if plausible, outcome. But sometimes the willingness to entertain those scenarios is the key to winning competitive league.

When I did the FSTA Draft in January, my hunches were all I had as I hadn't even looked at an ADP list and had done virtually no research. And I love the way that team turned out. But as I approach my NFBC "Beat Chris Liss" draft tonight, having seen projections, researched most of the player pool and looked over ADP lists, I'm having a harder time remembering who I originally liked and what impressions I've absorbed from reading other peoples' work. When you get confused, it's tempting to turn to a reference point like ADP or some projection system and just tell yourself you'll diverge from it as necessary. But now that it's in front of you, it's easy to imagine a good argument for and against anyone on the list, given his current placement - after all, there's a reason Player X has an ADP of 45, rather than 25 or 65.

To combat this, I'll probably just use our IPad App with your projections, knowing full well it comes with both three-year weighted average biases as well as yours, alternatively sort the players by NFBC ADP (which our app does) and knowing both are not objective or perfect, give myself wide latitude to "reach" based on my hunches during the draft. If I'm really stuck without a hunch at a particular juncture, I might glance at the Steamer list to see if there's someone its formula really likes in that spot, but that's only if I have no lean whatsoever.

Maybe my answer was beyond the scope of your question "Do I have any holes in my game?" though. The short version is yes, but my tendency to imagine plausible scenarios to be more likely than they actually are is more of a feature than a bug, and I think my game is best when I'm trying to hit winners even if it means a few more mistakes.

Jeff (6:45 pm PT, Thursday, February 19)

I've discussed the importance of auditing your rankings, both on-air with you during our Sirius/XM show and in a subsequent blog. The idea is that projections and rankings often look great in the abstract until you test them out in a draft and you find that there are a number of players wildly overrated and/or buried. I've done that, and adjusted my projections for those players accordingly.

But that's just the first step. The next step is to investigate those perceived errors and see if there are any patterns - blind spots if you will. Are there players that are consistently getting ranked too high, or ones I'm consistently missing? Is it a particular category I'm undervaluing? Or a position? Or just a type of player?

I've started that process, and the one trend that I've identified is that I'm falling short on potential second-and-third year breakout players. I think this blind spot is system-based, in how I do my projections. Part of my process for projecting players is to start with a three-year weighted average for a player - it's not the final result, but it's the initial baseline. But there are problems with this process for less-established players.

- As you've mentioned on the show, this entails a starting bias. Sure, I can progress or regress a player from that starting point, but often it's not enough. Players don't develop in a uniform fashion, and that's especially true from those on the extremes. A 10% qualitative improvement across the board sounds great, but what if a player improves by an order of magnitude more than that?

- On a related point, players can improve in one aspect of their games but not in another, either by design (trading speed for power, for example), or just as a fluke.

- Younger players don't have three years of data, or they have incomplete data. Nolan Arenado played only 111 games last year, and played after a finger injury that could have depressed his totals upon his return. So you have inputs of varying relevance, and trying to guess how much each line is worth is at best an educated guess.

All of these issues apply not just for my projections, but also for the higher developed systems like Steamer and Oliver.

How do you work your way out of these traps when you're researching the player pool? Do you investigate whether you have any blind spots? What are they? And how do you think you can best get around them?