The Steamer and I reviews continue this week, which pits a player’s preseason Pod Projection against his Steamer projection. Today, I’ll review 2014 breakout Corey Dickerson, who unfortunately saw his plate appearance total cut in half due to various injuries. I was far more bullish on his prospects than Steamer was.
For our second Pod Projection review of the season, I will recap what was predicted and what actually happened for rookie Yasmany Tomas. Refresh your memory by checking back on my original Pod Projection post.
As a Cuban rookie with no minor league data to work with and limited statistics to analyze from his work in Cuba, Tomas was a challenge to forecast. Essentially, everyone was really just taking an educated guess and it was more like a 3rd grade education than grad school level.
Back in February, I compared preseason projections between reputed projection system Steamer to those submitted by FanGraphs readers, dubbed “FANS.” The concept was simple: identify National League outfielders whose Steamer and FANS projections varied wildly and predict a “winning” projection. (In the same vein, Community poster Bobby Mueller compiled some nice summary statistics.)
Alas, I needlessly task myself with determining who fared better: the Depth Charts — which are Steamer and other reputable projection system ZiPS, with playing time allocated by FanGraphs staff — versus the fans.
Because the present author, whose analytic capacity is debatable but authorship of this piece is absolute, retains sole proprietorship of quasi-analysis that has a moderate to high probability of spiraling out of control, he has chosen three statistics with which to compare qualified major league hitters: weighted on-base average (wOBA), an offensive rate statistic (not that it offends anyone, per se, but, well, you know); Fielding (Fld), a defensive statistic, probably; and wins above replacement (WAR), an overall performance metric.
Yours truly has elected to discuss only the most egregious differences in projections and declare winners between them accordingly. Granted three nominees within three categories, an outcome in which a winner is not declared is highly improbable. Alas, a true, rightful and, above all, 100-percent authoritative champion may very well be crowned in due time. So, who will it be: the wisdom of the masses, or the wisdom of two computers?
Well, my pain may be your gain. For a future Hardball Times article I am working on, I needed to create some historic pitcher groundball rate (GB%) projections. I decided to go ahead and run the values for 2016 and make them publicly available.
Here is the basic background behind creating the projections.
• I used a three-year weighting of .7 for 2015, .2 for 2014 and .1 for 2013.
• The data was regressed some to a league average rate.
• I didn’t use any aging factors.
• The final r-squared from the projection to actual results is around .67.
With that small amount of background information, here are the projections.
It’s never too early to prepare for next season, just as you can never have too many articles about the mechanics of projection systems. Well, ok, the second part of that statement is a lie, but it has been awhile since we’ve talked about how we use projection systems.
It’s second base week on RotoGraphs and that gives me the opportunity to perform the first review of my 2015 Pod Projection breakdowns. Anthony Rendon enjoyed a breakout year in both real and fantasy baseball in 2014 and given his previous top prospect pedigree, was then highly valued heading into this season. At the time of my Pod Projection, Rendon was being drafted 13th overall in NFBC leagues. Unfortunately, a knee injury that was supposed to only keep him out for a week or two to begin the season dragged on and on. He didn’t make his 2015 debut until early June and then missed another month with a quad strain. Ultimately, he finished ranked 35th in value…among just the second basemen!
For the first time this year, I decided to compare my Pod Projections to Steamer and discuss the players we disagreed on most. Of the hitters we both projected, it was clear I was much more optimistic than Steamer on the whole. However, for Jose Abreu I was actually significantly more pessimistic.
In the review of this series, I will be including my original Pod Projection, and Steamer counting stat projections extrapolated over the same number of plate appearances that I had projected. However, I included Steamer’s actual PA projection in that column. Also included are the player’s 2015 stats, plus the counting stats extrapolated over the number of PA I projected.
Yes, the 2016 projections are available thanks to Steamer. Well, they are kind of available. The projections’ playing times are all set to 600 PA for hitters, 200 IP for starters and 65 IP for relievers. There is no middle ground right now. The playing time estimates will be included from our depth charts once the postseason is over and the hot stove season is well under way.
I have gone ahead and included the Standard Gain Points value I calculated from the 2014 NFBC leagues from the 2014 season to give the players an overall rank. I will eventually get around to calculating the 2015 SGP values, but these numbers will give owners an initial estimate of the player’s value at the set playing time values.
Here are the hitters and pitchers tables and the values can be download with the link after both tables.
Well, sorry for no analysis with this final week of starting pitcher values. Eno and I are traveling and I waited as long as I could to try to get this week’s starters. I implemented the changes people recommended last week which is the team(s) the pitcher is facing and the park factor(s). Most of the over-explanation is in last week’s article. Just let us know if there are additional changes you would like to see.
Well, I have had making a two-start evaluator on my to-do list for a while. With the help of Jonah Pemstein, the project is done. A problem is its usefulness with only two weeks left in the season. I am going to roll it out anyway and take comments over the next couple of weeks. It will then be ready for full implementation to start next season.
The setup is pretty simple right now.
•Find the games a pitcher is expected to start this upcoming week.
• Determine the starters projected Steamer stats per start.
• Use my Standing Gain Points formula I calculated to start the season to give each pitcher a weekly value. Then rank them by this value.
This is all the data available for now, but running it for the first time I found a few ways to improve it going forward.
• I ran into a coding error for pitchers who have no more projected starts (end of the list like Hudson or Moore) according to Steamer. I am trying to get the depth charts updated to make sure this doesn’t happen in the future.
• On the same note, the stats for pitchers who are swingman (starter and reliever) may be a bit inflated because of the improved reliever rates.
• Wins are just funky but needed to show the value of the chance to pick up two Wins. Ideas?
• The strength of opponent and park factors are not shown or worked in yet. I am working on this portion now but wanted to make the list available before next week.