Yu Darvish and Updated Projections

We’ve updated our playing time projections and posted new sheets. The new downloads include a number of players who were missing from our initial forecasts including Yu Darvish.

Yu Darvish 2012 Projections


It’s worth noting that Steamer initially projected an ERA of 4.11 for Darvish until we informed Steamer that Darvish’s fastball has averaged 95.4 mph in spring training. That lowered his ERA to 4.01.

Why are the other systems so much more optimistic? A 4.01 ERA while starting for Texas is an excellent performance but are we underestimating his greatness?

19 thoughts on “Yu Darvish and Updated Projections”

  1. Thanks for posting these!

    One comment: It seems like you’ve lowered the projected IP for pitchers a little much.
    E.G.: In the previous version, you have Roy Halladay projected for approx 32 GS and 230 IP. This version has Doc at 31 GS, but losing 24 IP.

    Many pitchers see their appearances go down precipitously in this update; Craig Kimbrel goes from 71 to 52, for example. Is this by design, or a flaw?

  2. This is by design but I might not know until after the season whether they’re flawed. If you check the last post about projecting playing time, I found that relievers projected to throw a high number of innings fell well short (although Kimbrel and Venters were the two major exceptions last year) as a group.

    I think what we should really move towards, is providing 25%ile/50%ile/75%ile playing time projections. No doubt, a healthy Kimbrel throws more than 52 IP. That said, I think 52 IP is a reasonable mean expectation and probably closer to the median expectation that 71.

  3. Thanks for the quick reply. I’ll have to manually adjust for my own fantasy projections, as my league values relievers highly (HD, K/9, K/BB), and having relievers get enough IP gets their Z-scores in line with their end value. (I’ll quit complaining about doing even this minimal work now.)

    Thank you so much for all the work you put into this.

  4. These projections are great; although, I wish hitters had a projection for games. I always struggle with projections that are based on PA and don’t include Gs. For my points h2h league I prefer to be able to figure out the value of a player based on Pts/wk based on pts/g. If I can figure out pts/g, I can figure out a reasonable estimate for pts/wk for injury prone players when they are healthy.

    1. We could probably work backwards and get games since we could predict PA/G based on lineup position and team OBP. Would it be valuable to know how many of the predicted off days are due to long-term injury and how many are due to unpredictable days off?

      1. I don’t really focus on that. I just figure out pts/g and extrapolate a players pts/wk if his games are below the average at his position. That gives me a baseline value to compare with other players. Although, I came up with a new idea yesterday. Instead of doing per week, I’m going to do per 10 games by multiplying pts/g by 10 to get a more significant value than pts/g. I dunno why I didn’t think of it before this year. I used to do my own projections but I’m unable to find the time anymore. Thanks for listening to my suggestion.

      2. Games projections are critical for many competitive fantasy baseball leagues, apart from their value in projecting PA. In daily leagues, you should basically never have a player not playing. Even when their absence is the result of just being benched for the day, there is the chance in many leagues to replace them with another player off the bench or free agency.

        In my league, nobody finishes near the top of the standings without more or less getting 162 games per position. Knowing how a certain player got to 650 PA is a big deal in evaluating what someone’s counting stats actually mean.

        Alex Gordon bats leadoff, so he reached 690 PA in “only” 151 games. That means for the other 11 games, players in a standard 5×5 league can accumulate R, RBI, SB, and HR for “free” the rest of the time. Depending on the league, you can replace him with WIll Venable (below replacement level based on your SGP levels), and get another 4 RBI, 5 R, 1.5 SB, and 1 HR out of Gordon that way.

        It would be a great help to get a games projection.

  5. Are you guys aware that you can make a fair projection of quality starts if you know the number of games a pitcher is going to start, the number of innings per game he averages, and the number of earned runs per game he averages?

    I’m sure you guys could calculate an even better projection for QS than I did if you tried. I’m surprised you haven’t done so for people who play in leagues that use that.

    1. What formula did you come up with? I tried this with a smallish dataset a couple of years ago when I was playing in a league with QS and found that for QS% (QS per start):

      NL avg: 56%
      AL avg: 52%

      subtract 8% for every ERA point above average and add 8% for every IP/GS above average. Does that sound reasonable? I feel like the relationship between IP/GS and QS% shouldn’t be linear but I never got into that level of detail.

  6. The updated total innings pitched projections look very realistic now at 41,327, very much in line with the average IP each year. Why, however, does the GS column sum to 5,022 and not 4,860 (30 teams * 162 games)?

    1. Good catch on this. A number of teams are well over 162 starts (Cleveland, Philly, Boston, Baltimore and the Yankees are well over). We’ll have to look over their depth charts at some point — although I can’t promise I’ll get to this soon.

    2. There will always be injuries, nobody knows exactly who yet other than the usual suspects like Bedard, JJ, etc, who should have injuries built-in to their playing time. We also don’t know who will even make the opening day rotations. Looks like Bard could be going back to the bullpen, though, which should help straighten out the Boston situation.

      I think it’s best to get an idea of the most likely projection for each individual player instead of each individual team, or the league as a whole. Let’s say hypothetically one player from each team misses 20 starts from a freak injury. We don’t know who it will be, so we can take off a handful of starts from each pitcher to account for it. Or we can make a prediction for each player assuming they are not the one to suffer the freak injury, which allows us to maximize the amount of pitchers with accurate projections, because the guy who misses 20 games is going to be wrong anyway so being a little bit more wrong is insignificant. 4 good playing time projections and one really bad one v/s 4 mediocre playing time projections and one really bad one. It’s not that black and white in real life but is still the better approach, in my opinion. And that leads to over-estimating the amount of GS, IP, and ABs if you sum up the projections.

  7. Thanks. I don’t think we’ll be able to add more players this year *but* in the future we should be able to be completely comprehensive and this might include a midseason update.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>