Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Why Does a Team Outperform its Run Differential?
1. Why Does a Team Outperform
its Run Differential?
Greg Ackerman
Syracuse University Sabermetrics Club
2. SU Sabermetrics Club
• SABR Student Group Affiliate
• Justin Mattingly
• Joey Weinberg
• Colby Conetta
• Ray Garzia
• Mallory Miller
• Zack Potter
• Marcus Shelmidine
• Brandon Love
• Matt Tanenbaum
• Bryan Kilmeade
• Justin Moritz
• Stephen Marciello
• Kyle O’Connor
• Willie Kniesner
• Michael Rotondo
• Matt Russo
• Sam Fortier
• Matt Filippi
• Isaac Nelson
• Zack Albright
• John Van Ermen
• Colton Smith
• Chris Karasinski
• Zach Tornabene
3. Basic Premise
• Explain the Difference between Actual Win Percentage and Expected Win Percentage based on Run
Differential (Expected Win Percentage based upon Pythagenpat formula from run differential)
• X = ((runs scored + runs against)/games)^.285
• If achieve run differential to possibly put team in playoffs – do not want to squander it
• If borderline run differential for playoffs – could be difference in attaining playoff spot
• Will focus on 3 key factors that may influence teams outperforming or underperforming their run differential
• Performance of Bench
• Relief Pitching
• Pitching Depth
• Part II – Add managerial decisions to the model
• Pinch Hitters Used
• Defensive Substitutions
• Relievers Used
• Etc.
4. Charts
• Average of (Actual Win % - Expected Win %)
• Standard Deviation of (Actual Win % - Expected Win %)
• Variables calculated from www.baseball-reference.com
• Team Examples of Difference in Actual Win % - Expected Win %
• San Francisco Giants
• St. Louis Cardinals
• New York Yankees
• Toronto Blue Jays
• Colorado Rockies
12. Measures of Bench (Hitters) Performance
• OPS+ - On-Base Average plus Slugging Percentage – Adjusted for Park and
League
• HR – Home Runs
• SB – Stolen Bases
• CS – Caught Stealing
• Calculated from Baseball Reference – using only bench players listed for
each team – weighted average based upon plate appearances of each
player
• Ultimately, only included OPS+
• Other variables did not add statistical value to the regression model beyond OPS+
13. Measure of Relief/Depth Pitcher Performance
• FIP – Fielding Independent Pitching
• ERA+ - Earned Run Average adjusted for ballpark
• SO/W – strike out to walk ratio
• Calculated as a weighted average based upon innings pitched
• Calculated for group of “relievers” noted on Baseball-Reference – includes
closer and top 4 used relievers
• Calculated for group of “depth” pitchers noted on Baseball Reference-
includes pitchers not included in “starters” or “relievers” categories
14. Regression Model I
• After different regression model incarnations – settled upon the
following to illustrate results:
• (Actual – Expected Win %)i = α0 +β1 (Bench OPS+) + β2 (Relief variable)
+ β3 (Pitching Depth variable) + εi
16. Results
• Variables have expected signs for bench (hitter) performance, relief
pitching, and pitching depth
• Only statistically significant result is for relief pitching performance
• Specifically – FIP
• FIP has a negative and significant impact on (Actual Win Percentage –
Exp. Win Percentage)
• As FIP increases – has negative impact on dependent variable
• More likely to underperform run differential
• As FIP decreases – has positive impact on dependent variable
• More likely to outperform run differential
17. Sample Bench OPS+ Relief FIP Depth FIP
top 10% Seasons -
Outperform Run Diff
81.56304 3.66215 4.691264
Bottom 10% Seasons -
Underperform Run Diff
81.63799 3.981814 4.769715
% Differential Between
Samples
-0.09% -8.03% -1.64%
20. Managerial Decisions
• For Next Step: Added Managerial Decisions to the Data Set
• To measure managerial decisions – used the Bill James Handbook
• Attempt to measure the impact of various managerial decisions on
the ability to outperform (underperform) a team’s run differential
24. Results
• Two statistically significant managerial variables:
• Defensive Substitutions – (+) – significant at the 1% level
• Pitchouts Ordered – (-) – significant at the 5% level
• Defensive Substitutions – more defensive substitutions used – greater
likelihood to outperform run differential
• Part is managerial decision
• Part is roster flexibility
• Pitchouts Ordered – more pitchouts ordered – greater likelihood to
underperform run differential
• Part is wasting a pitch
• Part is lack of faith in catcher/pitcher
• Likely a proxy for risk averse behavior on part of manager
25. Results
• Relief Pitcher Innings Pitched – (-) but not quite statistically significant
(15% level)
• When Managerial Statistics included – impact of FIP-Relievers is
lessened as well – no longer statistically significant
• Tried including one or the other – not quite statistically significant
• Appears to still have some marginal effect on ability to
outperform/underperform run differential
30. Conclusions
• Aimed to determine why teams outperform/underperform run differential
• Is it just luck? – or are there factors that contribute to its explanation?
• Without Manager Data – it appears that Relief Pitcher Performance
(measured by FIP) plays an important role
• Increase in FIP by Relievers – more likely to underperform
• Decrease in FIP by Relievers – more likely to outperform
• With Manager Data
• Defensive Substitutions – more defensive subs – more likely to outperform
• Pitchouts – likely proxy for risk aversion (poor catching performance?) – more
pitchouts – more likely to underperform run differential
• Starting point of our research – hope to learn more in future – open to
different variables/approaches to help determine answers
Editor's Notes
We are not looking at starting pitching and starting lineup, rather we are looking at performance of bench, relief pitching, and pitching depth.
Talk about the Yankees, Giants, and Cardinals having high averages and have been outperforming their run differential between 2000 and 2013, while teams like the Blue Jays, Rockies, and
One interesting thing here is that the Giants have the lowest standard deviation between 2000 and 2013
0 is what we would expect, so the Giants have regularly outperformed their run differential
Last of team group
The key to this is that we used a weighted average for each team amongst bench hitters only, according to baseball-reference. We are really only going to be focusing on OPS+ because it is the most promising of the bench statistics.
This is a weighted average of FIP, ERA+, and SO/BBW. We used
The dependent variable here is actual – expected win percentage. The independent variables are bench OPS+, and a variable amongst FIP, ERA, or strikeout to walk ratio amongst the top five relievers and depth pitching.
These are all very small numbers because we are dealing with trying to explain a very small number, which is the difference between actual and expected win percentage.
This makes sense because if you are not pitching well in late innings, you are likely to underperform your run differential, while if you have good Fielding Independent Pitching amongst relievers, they are more likely to outperform their run differential.
This slide shows the best seasons and worst seasons in terms of outperforming and overperforming run differential. The only place we have a sizeable difference between the top 10% and bottom 10% of seasons is in relief FIP, as there is a 8.03% dropoff.
This just gives an idea of the average bench OPS+ amongst the team.
This is the difference between relief and depth FIP.