I’ve blogged about participating in a fantasy baseball mock draft last weekend, but I have a confession: I’d never before done a straight baseball draft. I’ve been playing fantasy sports for over two decades, and I’ve done fantasy football drafts often. Once we did a fantasy basketball draft when the delayed season start made it impossible to get the whole group together for an auction. But I’d never done a draft baseball, as my two leagues, the Ezra Stiles Rotisserie Association and the Park Slope Rotisserie League both do a full auction every year.
Now I had the arrogance of assuming that a draft must be simpler than an auction. After all, you only need to know who is better than whom; you don’t need to try to quantify how much more you should spend on Miguel Cabrera than on David Wright. It should be easy, right? Well, not so fast…
First, I underestimated how fast it would go. For our auctions, we literally take all day to allocate 270 or 280 players. The mock draft, with a 60 second time limit per pick enforced by a bot making your selection, took just two hours. Making good choices in that short time is harder than the more leisurely auctions I’ve done. Sure, it’s tense as the bid is rising on a player you care about, but you’ve got time to review his stats, look at your needs and other teams’ needs, and think about how high you’re willing to go. In a draft with a fast clock, there’s not as much time for reflection, and I wish I’d done more specific preparation for it.
While I was aware of concepts like average draft position, I just assumed that using a descending price rank from my auction analysis would make for a good draft cheat sheet. I still think, in theory, that it makes sense, but now I understand better what Mark Healy and Jay Ferraro meant when they were asking if I took Clayton Kershaw so highly just to get them to talk to me on air. Hey, I appreciate the exposure, but I wasn’t smart enough to make that a conscious strategy. Kershaw literally was the top choice using the consensus projections at my site – ahead of Cabrera, Mike Trout, and Ryan Braun, and he was top 5 under any of the projection systems. I wouldn’t have taken him with the top pick (I’d go with Cabrera), but I could see a case for it.
But apparently Kershaw simply isn’t going anywhere near that high in mock drafts. At Mock Draft Central their ADP Report (requires registration) today has Kershaw at #17, and he’s been taken no earlier than 11th and as late as 31st. Now Kershaw is the top pitcher (Justin Verlander is next at 25th). Another site, KFFL, had Kershaw #13, just behind Verlander at #2. This leaves me with two main thoughts:
- Is the draft market really that inefficient, or am I simply overvaluing pitching?
- Hey, if nobody else takes pitchers that early, then even if I think they’re worth a top-5 pick, I should wait until the 2nd round myself.
I genuinely am perplexed. I didn’t look at any ADP data before doing the mock draft, and I suppose it would make me a little more reluctant to take pitching early. Yes, I get that pitchers are more volatile; a single season isn’t enough time for ERA to converge very closely on a player’s current skill level. But I also have confidence in my pricing model, and in projections systems as a good way to estimate a player’s current skill level.
I do wonder whether there can be a herd mentality in effect in different places. Maybe the owners in my leagues are influenced by my pricing model, both because they can now see it, and because I use it and have had success. So perhaps my leagues’ auction prices tend to be similar to my model’s prices because the latter drive the former.
Conversely, people in draft leagues are likely influenced by what they see in other draft leagues. And so if nobody is taking pitchers in the first round, that can be a self-reinforcing cycle. So perhaps the ADP values are influenced by how others draft.
Now there’s two ways to look at being different: it can be an advantage (i.e. you’re seeing things others don’t recognize, and you’ll profit from that), or a disadvantage (you’re missing something others are seeing, and get hurt by it). I do think it’s interesting that my team was 3rd best in both my own projected standings using consensus stats and in the Mock Draft Central’s projection page (no link; I’m pretty sure it was a private league), and both sources agreed that BigChee had the best team on paper. So while I’m open to the possibility that my model (or perhaps the projections I fed into it, although they gave largely similar results) may overvalue pitching, if forced to pick, I’d say taking pitching early is better than waiting.
Well, I’ll hedge – waiting longer, but still being the first to take pitching, is even better. If I could have taken Matt Kemp or Robinson Cano at #4 and still gotten Kershaw (or even Verlander) at #20, then yeah, I’d have been better off than taking Kershaw 4th and getting Bryce Harper in round 2. So if I know drafts wait for pitching, I’ll want to wait longer myself, but be the first to snap up elite pitchers.
I would say that being very different may often suggest you’re wrong, but it is good strategy to win a fantasy league. To win against 11 other knowledgeable owners, you need both skill and good luck. If you hew too close to consensus, you can do well, but you’re more likely going to be 3rd or 4th than win. The converse is that by being different, you may increase your chances of winning, but also of finishing last or near the bottom. Depending on your incentives, that can guide your strategy.
A few things come to mind.
First, in a draft format a draft pick has a fixed value relative to the overall selection process and never more so than in the first round. There is a common wisdom in baseball drafting that your first round pick needs to be a highly productive player who is also low risk with stable performance. Touts and experienced players won’t avoid pitching outright in the first round, but they will place a high bar on expected reliability that it difficult for starting pitchers to surpass. In fact, one reason Verlander is drafting under Kershaw is probably due to risk assessment of Verlander having more pitches thrown, thus more wear and tear, than Kershaw. First round earned value is very volatile from year to year though and more pitchers will be on that list at the end of the season than Tout lists and ADP predict in the spring.
Second, I am pretty sure our leagues tend to value pitching more than average. I haven’t looked at this in a few years, but we are usually in a 59%-62% hitting allocation league wide and I think the global average is probably in a 64%-67% range. You may have access to industry information that could confirm or reject this.
Third, related to the first point, I think your model very effectively estimates skill based on past performance. Where it is less effective are the areas where models always struggle, growth trajectories for young players, playing time estimates, and adjusting for risk. Just my opinion from outside the black box looking in. I think your pricing is accurate within the range any model or estimate must accept, but less so on a risk adjusted basis (true for many other purely numerical models as well). People want prices, but there is probably not a big market for prices with confidence intervals. Although for all I know your personal output may show price in a range.
This is why the hosts felt your selection of Kershaw went against the grain, but I think they are wrong. I think Kershaw is a reasonable first round pick as the best combination of skill and reliability for a pitcher given what we know going into the season. The picks I would question would be the talented hitters with health concerns. Guys like Kemp, Votto, Bautista, and CarGo. Pujols is showing some aging volatility too. I would bid for them aggressively in an auction format, depending on when they came up and on my needs, but I’m not sure I would be comfortable using a first round pick on them.
I’d agree pitchers are more volatile (as long as the MLB season is, it’s not long enough for stats like ERA to stabilize well around a player’s talent level), and I agree with the assessment that Kershaw is lower risk than Verlander because the latter is older and has more mileage on his arm.
I actually have two models in the discussion, the projections of player stats, and the RotoValue pricing model to convert some set of stats into dollar values. But given how I weight categories, how much money spent on pitching depends on how much better pitchers are than the replacement level. The extra volatility of pitchers should result in top pitchers having somewhat lower projections, and bad pitchers somewhat better ones, meaning the replacement level rises. And as the replacement level rises, my pricing model would allocate less money to pitching.
So in theory, at least, my pricing model would reflect greater volatility of pitchers relative to batters so long as the projections captured it also by reducing expectations more for good pitchers than for marginal ones.
I think this post at Tom Tango’s blog is relevant here. Tango notes that a projection that you’ll finish last does not imply you’ll probably do so (chance > 50%), merely that you may be the single most likely team to do so. Similarly when talking about projections he notes that projected values will be less extreme than actual ones, because projections should be some sort of averaging of potential future values. We may think there’s no one player individually likely to hit 40 HR this year, but in aggregate we expect somebody will. It’s just at this time of year we don’t know who that may be.
When Tim S. says “our leagues” above, I should note he is a 5-time champion of my long-time NL league, the Ezra Stiles Rotisserie Association, last winning in 2010. His 5 outright titles are the most in the 24 year history of the league.