Search THT:

Or you can search by:

March 11, 2012

A better ERA projection

The THT Forecasts have been as good or better than any at projecting the rates of home runs, walks and strikeouts allowed by pitchers. However, recent third-party tests have shown that we have fallen short in projecting pitcher's runs allowed, which is the most important number. After conducting a THT roundtable discussing possible solutions, and long hours coding, most modifications are now in effect.

I had been calculating runs allowed based on a non-linear formula that uses the wOBA allowed by each pitcher. This has shown to be accurate at all scoring levels, and I am still confident in its ability to project the number of runs created by the actions of the batters against each pitcher. However, that is not the only way that a pitcher can control run scoring.

The extent to which a pitcher can or cannot control base stealing is the largest factor. I have tabulated the steals, caught stealing by catchers, pickoffs, pickoff caught stealings and pickoff errors by each pitcher, assigning an extra 0.25 runs for each steal and 0.50 runs for each pick error more than expected, and a reduction of 0.5 runs and an extra out for each additional pickoff and caught stealing.

Likewise, a pitcher will reduce his projected runs allowed by 0.25 with every steal less than expected. Soon, wild pitches and balks above or below expected will be handled in the same manner.

Groundball pitchers and those with fewer walks and strikeouts will get more grounded into double plays per opportunity. GIDP is not included in the wOBA calculation, but I now assign a 0.5 run change and an out for each GIDP more or less than expcted. Pitchers with a higher groundball rate also suffer more batters reached on error, but those additional errors lead to a higher percentage of unearned runs. Soon, I will adjust the unearned run percentage for the pitcher's groundball rate.

The ERA baselines also have been adjusted. Previously, I was using a single value of the MLB averages from the beginning of Gameday in 2005 until the current date, but the accuracy of this has been called into question as the offense production in the major leagues has dropped over each of the past two seasons. The baseline MLB wOBA and runs scored per plate appearance are now based on a weighted mean of the same past three seasons that the individual projections are constructed from.

The last major item is whether the pitcher is performing as a starter or reliever. My comparison of the performances of pitchers who worked in both roles confirm research by Tango that showed that pitchers have a strikeout rate about 17 percent higher and a home run rate 17 percent lower when in relief, while the walk rate does not change.

Tango also found a reduction of the batting average on balls in play of about 17 percent in relief, compared to starting, but my research found a difference about half of that. When each of those components are adjusted, depending on their mix of skills, pitchers had starting and relief ERA projections that varied from about 0.40 to 0.80 runs.

These adjustments enable me to convert statistics compiled in relief to an equavalent starting pitcher performance before league conversion factors are calculated.

After the projections are completed—and customized to each pitcher's parent club's home ballpark—a final projection is created that is a weighted mean of the starting and relieving projections. For pitchers who work 100 percent in either role, there will be trivial changes in their projections. However, this gives us the flexibility to customize projections when pitchers are transitioning to a new role.

For example, Daniel Bard has had a 2.88 ERA over three years in the majors, 3.33 last year, working strictly as a reliever, but his 2012 projection as a starting pitcher is 3.96, reflecting the lower expected level of performance.
Posted by: Brian Cartwright

February 20, 2012

Rotovalue tests 2011 batting projections

Geoff Buchan has posted the results of his testing of the 2011 batting projections from several leading sites, our own THT Forecasts (aka Oliver), Baseball Prospectus' Pecota, Steamer, Rorovalue, and Marcel. I am pleased to say that, according to this set of tests, that Oliver had a very good year. Check out Geoff's article here


Posted by: Brian Cartwright

February 19, 2012

Park factor fix for Forecasts

The Forecasts update posted yesterday, Feb. 18, will have different numbers for every player from the week before, as I discovered a logic error in my code which was preventing park factors from being applied to each player's batting and pitching projections. Recently published Forecasts were park neutral but now are as intended, specific to the player's team.

Major league players' projections are customized to the weighted mean of all the ballparks his team played in the previous season. Projections for players in the minor leagues are based on their parent major league team. Schedules for 2012 may now be downloaded from mlb.com, and as soon as those are imported into the Oliver database, I will use the number of games scheduled to be played in each ballpark in the coming season instead of the previous.

I do not expect using 2012 instead of 2011 to produce major changes in the projections, but some players who play in extreme parks did see sizable differences when their park factors were correctly applied. Troy Tulowitzki solidified his rank as the overall best position player; playing half his games in Coors Field inflates his batting projection from a park neutral .285/.358/.506 to .302/.371/.544. On the other end of the ballpark spectrum, half a season in Petco Park (as well as a disproportionate number in parks such as Dodger Stadium and AT&T Park) drop Chase Headley's projection from .283/.353/.408 to .269/.342/.383.

I apologize for any inconvenience and welcome any comments from subscribers who suspect something may be amiss. Sometimes we'll have an explanation, but other times we have been able to catch errors. As a result of Matt Swartz's testing at FanGraphs of several projections, including Oliver, I am currently at work on some improvements on projecting ERA, as well as an existing project to be able to project a pitcher as either a starter or a reliever for those times when a player's role changes.
Posted by: Brian Cartwright

February 02, 2012

THT Forecasts - 2012 fantasy price guides

In addition to the Oliver database, reports, projections, and watch lists, THT Forecasts also has a fantasy price guide that you can customize with your league specifications (although note that not all stat categories you can imagine are present yet; we'll work on that!).

Also, the fantasy price guides are updated as the season progresses and new data are added to the Oliver database, thus changing the projections for the remainder of the season to assist you with waiver wire pick-ups, trades, etc.. You can then return to your price guide at the end of the season to evaluate players and see what they should have been worth before the season started (thus best to transfer and save your pre-season fantasy price guide for comparison).

Note that the fantasy price guide lists players according to their dollar values for an auction draft, but the rankings can still be applied to a snake draft. For a league that uses a points structure, you can switch to the customizable reports, download and open them with your favorite spreadsheet, and then include your stat-cat modifiers to calculate point totals for the players.

Finally, using Yahoo! Fantasy Baseball's default league auction and roto settings, here are the top projected offensive players:

Pos   Players      2012 Value
C     Mike Napoli     $34
1B    Albert Pujols   $55
2B    Dan Uggla       $30
3B    Jose Bautista   $40
SS    Troy Tulowitzki $32
OF    Matt Kemp       $57
OF    Mike Stanton    $47
OF    Ryan Braun      $44
Util  Miguel Cabrera  $52
Util  Joey Votto      $46

THT Forecasts is available for the 2012 season, with Brian Cartwright's Oliver database providing over 9000 projections for major and minor leaguers. Visit Forecasts today to learn more and sign up for only $14.95!
Posted by: Greg Tamer

January 25, 2012

THT Forecasts - 2012 Oliver projected-WAR starting lineup

Using the 2012 Oliver database, here is the top-projected-WAR starting lineup:

Pos Player       2012 Projected WAR
P   Yu Darvish          7.0 (Wow!)
C   Joe Mauer           3.5
1B  Albert Pujols       5.5
2B  Robinson Cano       4.5
3B  Evan Longoria       4.6
SS  Troy Tulowitzki     5.6
LF  Ryan Braun*         4.6
CF  Matt Kemp           4.9
RFl Jose Bautista       3.9
                       -----
                       44.1

* - If Braun indeed suspended for 50 games, will then substitute in Matt Holiday, 4.2 projected WAR.

Think you could construct a lineup with other players that will out-WAR the above lineup in 2012? Perhaps worth an investigation after the season!

THT Forecasts is available for the 2012 season, with Brian Cartwright's Oliver database providing over 9000 projections for major and minor leaguers. Visit Forecasts today to learn more and sign up for only $14.95!
Posted by: Greg Tamer


Click here for more THT Notes.