Feb 13, 2012

MockDraftCentral: Analyzing Rankings & ADP




For better or worse, MockDraftCentral is the fantasy baseball industry standard for average draft position data. If you read my Twitter, you may have seen me recently rant a bit about how the data is very skewed there and doesn't really tell us much of anything. The reason I say that is because their ADP data almost exactly mirrors their default site rankings which are presented to mock drafters. Since their site presents rankings to the drafters, their own bias influences the results as I touched upon last year.

In looking at the Top 150 players in their ADP at the moment, there is a 0.99 correlation to their site's default rankings. What that means is that if a player’s default ranking on MockDraftCentral (MDC) were to increase by one standard deviation (about 44 spots here), it's expected that their ADP on their site would increase by 0.99 standard deviations (43.7 spots on the ADP here). Which is why it's not a surprise to see Jered Weaver have a ranking of 32 at MDC (which is ranked one standard deviation away from the average ranking of 76 in our sample) and also have an ADP of 33.9. The link is extremely strong between the site rankings and the ADP data.

It's hard to really learn anything when you look at Chase Utley’s ADP of 77.12 versus their ranking of 76 for him. It gets especially confusing when you look at another mock draft site, CouchManagers, and see his ADP is 49.8 there (where he is ranked at 47 coincidentally). Looking at MockDraftCentral's data, you don't learn what the public thinks of a player like Utley -- you learn what MockDraftCentral thinks of him.

To go back to the statistics, the r-squared value is 0.98 when comparing the rank and ADP which means that a player's ranking at MDC is 98% of what goes into their resulting ADP. That does leave open a small 2% window that isn't explained by their rankings and could be explained by the actual preferences of the mock drafters. So, yes, there is a sliver of valuable information to be found.

For instance, there are some players whose ADP varies quite a bit from their ranking which makes them quite interesting. If we’re really trying to gauge how the public perceives a player then this is the information we can gain from looking at MockDraftCentral. When you see that MDC has Brett Lawrie ranked at 63 but his ADP is 54.04 (17% increase from rank) then this shows that the public is reaching for Lawrie in their drafts. On the flipside, Jayson Werth is ranked at 90 by MDC but has an ADP of 99.42 (9% decrease from rank) which shows that the public is choosing to pass on him at that spot because they think it's too high.

So, let's take a gander at the guys that public has seemingly strong opinions on. When looking at the players that the public is drafting ahead of their expected spot, there are quite a few pitchers on the list. It could mean that MDC ranked pitchers a bit too high or it could mean that the general public would rather pass on pitching in the early rounds this year. Regardless, here are some of the top names that the public is choosing to pass on:

Name Pos ADP MDC Rank Diff
Justin Upton OF 9.09 7 -23%
Justin Verlander SP 9.28 6 -35%
Roy Halladay SP 15.38 13 -15%
Cliff Lee SP 20.71 17 -18%
Tim Lincecum SP 24.79 22 -11%
Felix Hernandez SP 27.26 23 -16%
Cole Hamels SP 32.01 27 -16%
David Price SP 39.25 34 -13%
Dan Haren SP 43.37 39 -10%
Elvis Andrus SS 45.08 41 -9%
Yovani Gallardo SP 51.59 45 -13%
Jon Lester SP 52.2 47 -10%
Craig Kimbrel RP 59.14 53 -10%
Shin-Soo Choo OF 65.02 59 -9%
Mat Latos SP 72.29 64 -11%
Drew Stubbs OF 81.46 74 -9%
Drew Storen RP 82.39 72 -13%
Daniel Hudson SP 86.83 78 -10%
John Axford RP 91.25 83 -9%
Josh Beckett SP 92.98 84 -10%
Jayson Werth OF 99.42 90 -9%
Cameron Maybin OF 103.49 93 -10%
Johnny Cueto SP 116.32 104 -11%
If you're at the end of the first round and are hoping for Justin Upton, there's still hope for you as the public isn't buying into MDC's #6 ranking for him. Also, some other names like Drew Stubbs may not need to be reached for because it seems that the public would rather pass on him at his current ranking.

Now, let's look at some of those that the public is choosing to reach for ahead of their rankings:

Name Pos ADP MDC Rank Diff
Robinson Cano 2B 9.6 11 15%
Prince Fielder 1B 13.87 16 15%
Hanley Ramirez SS 18.83 21 12%
Mike Stanton OF 22.61 26 15%
Ian Kinsler 2B 23.55 28 19%
Mark Teixeira 1B 27.1 30 11%
Adrian Beltre 3B 31.04 35 13%
Mike Napoli C 45.62 51 12%
Brett Lawrie 3B 54.04 63 17%
Stephen Strasburg SP 62.33 68 9%
Pablo Sandoval 3B 68.9 81 18%
Asdrubal Cabrera SS 71.72 80 12%
Michael Morse 1B 77.56 96 24%
Jimmy Rollins SS 86.12 92 7%
Lance Berkman OF 90.87 100 10%
Josh Johnson SP 97.53 108 11%
Matt Moore SP 100.85 111 10%
Adam Wainwright SP 103.84 113 9%
Yu Darvish SP 117.59 137 17%
Dee Gordon SS 144.93 158 9%
Ryan Howard 1B 145.59 173 19%
Despite their ranking of 96 for Mike Morse, the public is grabbing him nearly two rounds ahead of that. If you want Morse, he will likely go earlier than expected because the public seems to be hot for him. There are a number of names here that you will likely want to reach for if you are hoping to get them on your roster for this year.

It's quite valuable to have an idea on what the general public may be thinking as you head into drafts. By being able to predict the moves of your leaguemates, you can better plan your own moves in the chess match that is draft day. ADP data is certainly one part of that puzzle. And, while MockDraftCentral doesn't really warrant being considered the industry standard for ADP data, there is still a bit of valuable information to be found if you dig deep enough.

Filed Under:

2 comments:

  1. I used MockDraftCentral for the first time this year and I can't say I'm surprised at the strong correlation between ranking and ADP. The interface only shows you a few players at a time and makes it hard to browse for players for which you might want to reach. I would suspect that you would find more variance in Yahoo mock drafts simply because of the superior browsing capability.

    That being said, mo matter what interface you use rankings will always influence ADP to some extent which makes one wonder if there's any way to collect ADP data without influencing the results? It doesn't seem realistic to expect users to participate in a mock draft where they don't have access to some kind of ordered list and even something as neutral as alphabetical order would favour some players over others (in this case, simply based on the spelling of their names).

    ReplyDelete
  2. I certainly agree. It's a complex problem. If a rival mock draft website was created that focused on trying to get reliable data, they'd likely risk being not user friendly and lose customers as a result. Ideally, I think the best design would be where the drafters had the ability to quickly and easily sort by a wide variety of stats without there being any rankings. Give people as much info as possible without any rankings. Or maybe just very generic rankings like "Early Round Candidate" or "Middle Round Candidate" and such... so drafters could still get an idea on who to focus on. It would certainly skew things but not as bad.

    Also, I think that MDC's timer is a bit too fast which gives you less time to actually research a pick... but I understand they don't want to lose the interest of other drafters and want to get drafts done very quickly.

    It's complicated but I know that there's got to be ways to improve the data. A 0.98 r-squared is just too high... If I had the technical know-how, I'd love to create my own mock draft site as another option to have out there.

    But, yeah, once more Yahoo mock drafts have been done (ESPN.com and CBS Sportsline too), I'll look at their results and see which one is the best in this regard.

    ReplyDelete