I made a trip to Arizona at the start of the month to take over the BP chair in the LABR draft run by USA Today. Jonah Keri drafted for us last year in the AL and won the league. He was back this year, with new partner Derek Zumsteg, to play in both the AL and NL as a representative of both ESPN and Derek’s new book “The Cheater’s Guide to Baseball.” Between facing the defending champion, and taking over as the man in the chair for the defending champion organization, I had big shoes to fill.
The major reason I went is that I’ve been working on a new program specifically designed for fantasy auction “drafts.” The numbers that go in come from PECOTA–I made a few personal tweaks to the numbers that came from our Depth Charts, usually to add unlisted players I thought had a chance of getting major league time–but the way they are treated is very different from the way they are calculated in the PFM. It is supposed to work in very much the same way as the Postseason Odds report I run during the regular season. At every step of the draft, it computes expected team totals, based on the players you have already acquired, and taking an average of the players who are left–although the more money you have left, the bigger share of the average remaining talent you get credit for. Just because you are projected to have 170 stolen bases, more than anyone else, doesn’t mean you will actually get 170 and finish first in the league. Every stat comes with a distribution around it, and I run random numbers to extract representative values of that distribution. I run 1000 seasons through, in order to find out what kind of change in the standings I get from adding, say, Jose Reyes to my roster.
The key to the whole project is that it would continuously update the value of stolen bases, home runs, and so on, based on what players you already have. If you add a Juan Pierre as the first player in the draft, you’ll immediately jump from the absolute middle of the pack in steals (where everybody starts) to maybe third–he’s worth four points in a 13-team league, and you went from seventh to third. Try to add a Jose Reyes now, and his stolen bases are worth, at most, two more points, bumping you up from third to first. Since you already have a lot of stolen bases, further steals are de-valued. The analogy is to runs in actual games–the run that gives you a one-run lead has more win value than the run that takes the lead from one to two, which has more win value than the run that takes it from two to three, and so on. My program works the same way.
Unfortunately, I really didn’t have it done to the level I wanted it at, and certain aspects had to be done through shortcuts simply because I didn’t have enough time to write the code for the full and complete way I wanted it. Still, I think it was doing rather well, and it was a huge improvement over a buggy version I’d used for a draft on Groundhog Day, one that didn’t subtract hitter’s values from the board as they were taken. Big oops there, and by the time I caught on to the problem, the hitters had been well picked over.
On to LABR. Even though I was only drafting for the NL side, I went to the AL draft the day before to shadow draft and give the program a test run. It was a good thing I did–not only did I find a small bug in the program which I could fix overnight, the pace of the bidding was a lot faster than the leagues I’m used to. I thought Jonah did exceptionally well, and my stats say he’s looking at another win.
The next day, I was tired. Anxious, nervous, worried if I’d caught all the bugs, if the shortcuts really approximate the total properly, if I really figured out how to get 1400 innings from pitchers on the Nationals roster–I didn’t sleep much. That was all forgotten when the draft started. Adrenalin kicked in, and I don’t remember feeling tired over the next five hours. I got a couple of funny looks when I started my program. Pretty much everybody else there had a computer, mostly Excel spreadsheets from what I saw. I’m not a GUI guy, and the program runs in the DOS command window through a pure text interface. It’s only 2007, Clay, welcome to the Nineties. On the plus side, it is simple to use and I don’t have to change any code between Windows (my laptop) and Linux (my desktop where I wrote it all).
Carlos Zambrano was the first name out, and he zoomed right past my $21 price point to 24. Pass. Ryan Dempster, Barry Bonds, Roy Oswalt, Bill Hall all come out, and get bought for too much for my tastes. Finally some of the big names come out, and go for prices you’d expect–Jose Reyes for $40, Albert Pujols for $43, Alfonso Soriano for $40, Chris Carpenter for $30. The key for me is that they are fair prices; I always want to get a buck or two under fair price. On the other hand, one of the other things the program does is try to keep you in line with money being spent; it will start suggesting higher prices if you’re sitting on too much money, and will reduce them if you are spending faster than your opponents, keeping you in line for the endgame. (I can’t stand being stuck with $1 for more than my last one or two picks, so I wrote the program to steer me away from that.) Since one of the goals was to evaluate the program, I really wanted to follow it pretty strictly, even if my personal judgment said to do something else.
So it wasn’t until more big money had gone off the board–Berkman at $33, Utley at $37, Howard at $40–that I finally got in on the 25th pick, getting Brandon Webb for $29. That was $2 more than the program started the draft with, but the program thought I was sitting on too much money, and upped the price. It would be a couple of more rounds before I picked up my first hitter, Kevin Kouzmanoff for $12, but once I got rolling I started zipping through players quickly. Jake Peavy? Yes. Adam Dunn? You bet.
That reminds me about something else I came in planning on: I specifically targeted players who had a really high gap between their Breakout and Collapse scores. I wound up with three such players in Adam Dunn, Prince Fielder, and Troy Tulowitzki, and scooped a fourth in the six-player draft phase with Yusmeiro Petit. Unfortunately, I didn’t get Chris B. Young, or Stephen Drew, or Jeff Francoeur, who were also pretty high on that list.
One of the things that needs upgrading is how the program tracks positions. I got stuck in a middle-infield muddle; the Rotowire guys filled their middle infield spots early, then kept throwing them out every time it came round to them. A good strategy on their part, to be sure, as all of the useful players disappeared while I sat on my hands. I like Tulowitzki fine, but paid full price for him, which is not what I want to do with a rookie–no matter how touted they are, or how good their stats are coming from the minors, not being “established” means you are at higher risk for getting sent to the minors if you happen to start the year in a two-for-25 slump, and that has to be taken into consideration. For the other two slots I wound up with Kevin Frandsen and Ryan Theriot–not good. But like I said, I was zipping right along, and ultimately was the first person to fill his team. The final roster looked like this, with the price I paid and the price I had pegged at the start of the draft after each player:
I took a lot of ribbing for Hatteberg, since he is anything but a “BP” type of player. My type or not, I had him for $11 value, so seven bucks is a nice price.
Middle: Troy Tulowitzki (14/14), Ryan Theriot (2/1), Kevin Frandsen (1/2)
Outfield: Adam Dunn (24/23), Dave Roberts (20/19), Corey Hart (16/18), Kelly Johnson (13/13), Matt Kemp (6/8)
DH: Kevin Kouzmanoff (12/14), by rule a DH only at the draft
Pitchers: Brandon Webb (29/27), Jake Peavy (28/32), Anthony Reyes (11/16), Jonathan Broxton (8/12), Cla Meredith (4/10), Kevin Correia (2/3), Tim Lincecum (2/6), David Wells (2/7), Bob Howry (3/5), Jorge Julio (1/7)
Draft (six player reserve): Michael Bourn, Yusmeiro Petit, Humberto Quintero, Russ Ortiz, Josh Hancock, David Newhan
Ortiz isn’t someone I liked, but thought he would be worth a flyer based on what we’d seen to that point–a noticeable loss of weight and improved velocity, which has stood up well through the spring. Three weeks later I’m still very happy with how most of these guys have performed in the spring: Paulino, Fielder, and Dunn have really been smashing the ball, Reyes has been outstanding, and the rumors from Florida are very encouraging for Julio’s value. Meredith and Wells have been shaky, as was Kemp, but you have to expect that someone’s going to stagger.
By my own stats, the ones I was drafting with that were pretty much entirely from PECOTA, I’ve collected $304 worth of talent for my $260, and I ranked myself first in the league by about a ten-point margin. The funny thing about that is how sensitive it is to your assumptions. I ran the numbers with Rotowire’s stats, and I came in tenth, as they were less optimistic about players like Dunn, Kouzmanoff, Tulowitzki, and Hart then I was. That is something else you need to remember in your drafts–whoever has the most optimistic projection for a guy is going to get him, assuming you stick to your lists. You’ll think you’re getting a bargain, because you got him $3 under your value; everyone else thinks you screwed up, paying $2 more than you should have.
Unless there were some big-time keepers, anything less than a first-place finish based on your own statistics really has to be considered a screwup on your part. The finals will be determined in part by whether your projections were really better than everyone elses, and I like PECOTA’s record over the last couple of years. The other part, of course, is what kind of trading and free agent picking you can pull off during the season, but we’ll see how that sort of mayhem unfolds as the season gets underway.