Sunday, February 9, 2014

Thoughts on Projections

The 2014 ZIPS projections for the Giants have been released on Fangraphs.  There are several projection systems out there, with ZIPS seeming to have the most respect in terms of accuracy.  Projection systems take historical data and use them to calculate future performance.  Some are quite primitive, such as simply taking the average of the last 3 season's performances.  Others like ZIPS use more advanced metrics including average career trajectories and park factors when players change teams.

Every year, these projections generate lots of discussion, sometimes heated, with lots of hand wringing from fans who think their favorite breakout candidate has been shortchanged.  Some of us can remember the days when we dreaded projection season, because it seemed like the Giants never fared too well, especially on the hitting side, usually for good reason.  We all still read them because there seems to be an insatiable appetite on the part of humans to know the future.

While the projections always seem to generate intense discussion when they come out, we rarely look back and see whether they were accurate or not.  Once the season is over, you know the outcome and want to move on to the Hot Stove and next season.  Who cares what ZIPS thought 6 months or a year ago?

First, some explanations.  Projections are not simply one line of numbers.  They are probabilities with a range of possible outcomes of variable likelihood with the published numbers generally being the mean or average of those probabilities.  Think of the published projection as being the center of a bell shaped curve.  By their nature, projections do not predict breakouts or busts well which is why ZIPS is only projecting three 6-WAR seasons in all of MLB, when the historical average per season is around 10. They also do not predict injuries well, but if a player misses time due to injury, it will impact the PA and counting stat projections.

Rather than simply list the 2014 ZIPS for the Giants and then comment on them in a relative vacuum, I decided to do a little extra work and put out last year's projections, last year's actual performance and this year's projections to possibly get a better feel for their accuracy or lack thereof.

Buster Posey:  2013 ZIPS- .299/.372/.482, 20 HR, 75 R, 86 RBI, 583 PA.
                        2013-          .294/.371/.450, 15 HR, 61 R, 72 RBI, 595 PA.
                        2014 ZIPS- .293/.367/.467, 18 HR, 71 R, 83 RBI, 583 PA.

Brandon Belt: 2013 ZIPS-  .264/.358/.429, 13 HR, 63 R, 67 RBI, 544 PA.
                       2013-           .289/.360/.481, 17 HR, 76 R, 67 RBI, 571 PA.
                       2014 ZIPS-  .264/.349/.441, 16 HR, 73 R, 70 RBI, 562 PA.

Marco Scutaro:  2013 ZIPS- .269/.316/.360, 5 HR, 53 R, 45 RBI, 542 PA.
                          2013-           .297/.357/.369, 2 HR, 57 R, 31 RBI, 547 PA.
                          2014 ZIPS-  .283/.332/.370, 4 HR, 65 R, 44 RBI, 547 PA.

Pablo Sandoval:  2013 ZIPS- .285/.338/.467, 18 HR, 68 R, 76 RBI, 547 PA.
                            2013-          .278/.341/.417, 14 HR, 52 R, 79 RBI, 504 PA.
                            2014 ZIPS- .279/.335/.448, 17 HR, 64 R, 81 RBI, 555 PA.

Brandon Crawford:  2013 ZIPS- .230/.292/.340, 6 HR, 49 R, 42 RBI, 475 PA.
                                 2013-          .248/.311/.363, 9 HR, 52 R, 43 RBI, 550 PA.
                                 2014 ZIPS- .238/.301/.351, 7 HR, 53 R, 46 RBI, 500 PA.

Hunter Pence:  2013 ZIPS- .262/.317/.410, 18 HR, 82 R, 90 RBI, 665 PA.
                         2013-           .283/.339/.483, 27 HR, 91 R, 99 RBI, 687 PA.
                         2014 ZIPS- .264/.319/.425, 20 HR, 87 R, 93 RBI, 665 PA.

Angel Pagan:   2013 ZIPS- .270/.320/.403, 8 HR, 81 R, 55 RBI, 610 PA.
                         2013-          .282/.334/.414, 5 HR, 44 R, 30 RBI, 305 PA.
                         2014 ZIPS- .266/.316/.388, 6 HR, 69 R, 47 RBI, 508 PA.

Gregor Blanco:  2013 ZIPS- .227/.322/.326, 4 HR, 50 R, 25 RBI, 400 PA.
                           2013-          .265/.341/.350, 3 HR, 50 R, 41 RBI, 511 PA.
                          2014 ZIPS- .238/.319/.333, 4 HR, 52 R, 33 RBI, 441 PA.

Joaquin Arias:  2013 ZIPS- .258/.285/.352, 4 HR, 39 R, 35 RBI, 371 PA.
                         2013-          .271/.284/.342, 1 HR, 17 R, 19 RBI, 236 PA.
                         2014 ZIPS- .259/.282/.350, 3 HR, 30 R, 28 RBI, 311 PA.

Hector Sanchez:  2013 ZIPS- .257/.285/.373, 6 HR, 34 R, 57 RBI, 362 PA.
                            2013-          .248/.300/.349, 3 HR, 8 R, 19 RBI, 140 PA.
                            2014 ZIPS- .250/.291/.362, 6 HR, 29 R, 47 RBI, 377 PA.

The team returns essentially intact from last year with the addition of Michael Morse.  What is his ZIPS, you ask?  .253/.304/.415, 14 HR, 48 R, 47 RBI, 427 PA.  I imagine Morse has a wide range of probabilities in his projection that you don't see here.  If he just hits to the projection, it appears he will be a significant upgrade on the offensive side.

As you can see, not only did Buster Posey fail to repeat his huge 2012 season, he fell short of his ZIPS projection by a significant margin.  A return to just his projection will be a significant improvement on 2013.

ZIPS does not see a continued upward career trajectory for Brandon Belt. In fact, it projects a small regression.  I think most of us who have watched his career unfold think he is still on a fairly steep upward trend.  It will be interesting to see how that plays out.

ZIPS is slightly optimistic for Scooter.  Conservative for Sandoval and extremely conservative for Hunter Pence.  It is optimistic for the bench.

Now to the pitching side of the ledger.  I'm just going go with ERA here to keep things simple and on the bottom line.  W's are a crapshoot and K's and BB's, BAA are all just components of predicting ERA.

Madison Bumgarner:  2013 ZIPS- 3.23, 2013- 2.77, 2014 ZIPS- 2.76.

Matt Cain:                   2013 ZIPS- 2.99, 2013- 4.00, 2014 ZIPS- 3.19.

Tim Lincecum:            2013 ZIPS- 3.67, 2013- 4.37, 2014 ZIPS- 3.80.

Tim Hudson:               2013 ZIPS- 3.61, 2013- 3.97, 2014 ZIPS- 3.28.

Ryan Vogelsong:         2013 ZIPS- 3.68, 2013- 5.73, 2014 ZIPS- 4.02.

Yusmeiro Petit:            2013 ZIPS- 4.68, 2013- 3.56, 2014 ZIPS- 4.10.

Sergio Romo:              2013 ZIPS- 2.25, 2013- 2.54, 2014 ZIPS- 2.39.

Santiago Casilla:          2013 ZIPS- 3.36, 2013- 2.16, 2014 ZIPS- 3.44.

Javier Lopez:               2013 ZIPS- 3.48, 2013- 1.83, 2014 ZIPS- 3.16.

Jeremy Affeldt:            2013 ZIPS- 3.44, 2013- 3.74, 2014 ZIPS- 3.48.

Heath Hembree:          2013 ZIPS- 3.80, 2013- 0.00, 2014 ZIPS- 3.51.

Jean Machi:                 2013 ZIPS- 4.87, 2013- 2.38, 2014 ZIPS- 3.47.

ZIPS projects sustained success for Bumgarner and a return to form for Matt Cain.  It seems to think Tim Hudson will benefit greatly from pitching in AT&T Park.  It projects modest improvements for Timmy and Vogey.  I think it sells the bullpen short, but that's just me.

Putting it all together, I still think the season hinges on the starting rotation.  If everybody in the rotation pitches to their projection, the Giants should be alright.


  1. This is a lot of extra work, thanks for putting that all together.

    As I commented in the article, I think that a healthy healed up wrist Michael Morse could beat the projections. It seems that his injury riddled year is getting weighed heavily here. The article portrayed him as a power only guy, but he's actually hit for good BA except for 2013 once he broke through.

    Scutaro handily outperformed, and I think its worth noting his high OBP in that 2 hole. The standard snark at Bochy is 2B hit 2nd, a semi-funny quip, but as the Giants always get knocked for "not getting obp" and overpaying for old vets, I thought it was worth noting that Scutaro is one of the best 2-hole hitters in the league last year (and projected by MLB's shredder to repeat fwiw).

    1. Like I said, Michael Morse probably has one of the widest confidence intervals out there, but even if he only performs to the projection, he will make a major contribution to the offense.

  2. If they do a combined team projection, one expects to see less volatility.

    It would be interesting to see a comparison between last year's opening day roster vs. the projected opening roster (unless there is some major surprises, the one or two guys making the last few spots shouldn't impact much versus the actual opening day roster 2014).

    And it would be interesting to see a projected injury-factored (some sort of injury probability - any stats on that front?) numbers for each players and the team as a whole.

    1. True about team projections, but I'm actually amazed at how accurate ZIPS is for individual players. Sure, there are some outliers here and there. You expect that, but man, it is extremely close on a lot of them! I thought it was almost eerie how many of the numbers came out exactly the same between the projection and the actually performance.

    2. While I am not aware of any way to predict injury, the system does project PA's and IP's. If a player, such as Pablo Sandoval misses time several seasons in a row, you will start to see it in lower PA projections and thus lower counting stats.

  3. ZIPS also factors in some methodology for regressing to the mean. The result being a "suppressed" output for players possessing real talent but without a couple of thousand major league at bats. The assumption in the model being that without a track record any above the mean performance is due largely to luck. For instance, ZIPS woefully under predicted Mike Trout's output in 2012 and 2013, but seems to be getting a little closer to reality now.

    Mike Trout
    ZIPS Projection -----> Actual Output

    2012: .267/.338/.414 -------> .326/.399/.564
    2013: .282/.361/.507 -------> .323/.432/.557
    2014: .300/.403/.535 -------> ??


    1. Hmm…..I guess ZIPS didn't see Mike Trout play against the San Jose Giants in the Cal League playoffs like I did. LOL!

    2. Ron Shandler writes in HQ's Baseball Forecaster that he predicted that Mike Trout would have a dropoff last year after his "unsustainable BABIP and home runs to fly ball ratio." As you know, nothing of the sort happened. Shandler shows humility in his opening piece titled "Nightmare," but sticks to his guns stating the methodology is solid. It will be fun to watch Trout this year. I may have to get over to watch the A's play the Angels this year.

    3. I had the rare pleasure of watching Mike Trout play in single A ball in a playoff game against the San Jose Giants. A fellow Giants fan and I attended the game to get a look at Giants prospects. We came away mightily impressed with Trout, to say the least. It was obvious that he was playing at least 2 levels higher than the second best player on the field and he was just 19 years old. He was in the majors less than a year later. We still talk about it in awe now. How many times do you get to see a once-in-a-generation talent play in the low minors and actually foresee their future success? I will always treasure that experience, even if it was a non-Giants player we witnessed that night.

  4. Terrific work to be able to see comparisons of 2013 ZIPS projections vs 2013 actual stats. I appreciate you putting in the time and effort. Being a big fan of Runs Created, I would like to see how projected Rc vs. actual RC came out.

    1. Thanks, appreciate you reading and commenting. Perhaps I will include RC when I review the outcome of the 2014 ZIPS next offseason?

  5. I can easily see Posey and Pence hitting 40-45 homers next year.

    1. That would be their combined HR total, of course. I would think we would be disappointed if they did not hit at least 40 total between the two of them.

    2. I meant 40-45 for each of them.

  6. PAs for Pagan and Pence are interesting over the past couple of years. Looks like ZiPs predicted both to be iron-men in 2013 (over 600PAs in Zips is pretty high), but of course Pagan's injury has changed things for this year's prediction. I'll have to look later for SB predictions, which contribute to both their games.