Measuring 2023 Team Efficiency
------------------------------

By Everett James
December 10, 2023


In this article, well see what we can learn about the 2023 season by 
examining team performance at three levels:

  Level 1  how many games did they win?
  Level 2  how much did they outscore their opponents?
  Level 3  how much did they outproduce their opponents statistically?

Obviously, the results at Level 1 decide who takes home the trophies.  But 
the results at Level 2 put you in the best position to lead the pack at 
Level 1, and the best way to come out on top in Level 2 is to dominate at 
Level 3.

If you examine the data across many real-life seasons, youll see that the 
relationships between levels are quite strong.  Runs predict wins and 
statistical performance predicts runs to a high degree.  But not perfectly.  
And the exceptions make for some surprising outcomes at Level 1.

We use the term team efficiency to describe the ability to turn batting 
events into runs and runs into wins. An efficient team is one that produces 
more wins than expected given its run margin, produces more runs than 
expected given its batting events, or allows fewer runs than expected given 
the hits and walks produced by their opponents.

In the 2007 edition of this article, we showed that teams that are unusually 
efficient (or inefficient) have exhibited a strong tendency to revert to the 
norm the next year. That is good news for some teams and unwelcome news for 
others. If you would like to find out who falls into which category, read on.

Converting runs into wins
-------------------------

The rules of baseball say that the winner of any one game is the team that 
scores the most runs, so of course theres a strong connection between runs 
and wins. 

Over a full season, outscoring your opponents by a lot of runs means youre 
very likely to win a lot of games.  Bill James used a formula that he called 
the Pythagorean method to calculate the mostly likely win total (where RF is 
runs for, RA is runs against, and ^2 means were squaring the number):

                                RF^2
  Projected wins =  Games * -----------
                            RF^2 + RA^2

Some teams come out ahead of or behind their predicted win total because of 
an unusual distribution of runs between games.  If you play a three-game 
series and outscore your opponents by a total of 9 runs to 6, the formula 
predicts that you would win twice.  But its also possible to sweep that 
series with three one-run victories or lose the series with a 6-1 win and 
losses by scores of 2-1 and 3-2.  

Over the course of a full season, these short-run variations tend to cancel 
each other out, but there are exceptions, and those exceptions can play a 
major role in the outcome of the season.

Since 1962, the first year when both leagues played a 162-game schedule, 
nearly half of all teams have finished within 2 wins of their predicted total 
and 84% of teams have finished within 5 wins.  Most often, the exceptions 
resulted from winning a lot of blowouts (that pad run differentials) and 
losing a lot of close games.  Or vice versa.

The biggest exception on the upside was the 2021 Seattle club that allowed 
51 more runs than they scored and still managed to win 90 games.  Normally, 
that run margin would produce 75 wins.

The biggest exception on the downside was the 1993 Mets, who were outscored 
by 72 runs.  Thats not a good result, but it would normally lead to an 
almost respectable 73-89 record.  Their actual record was an ugly 59-103.

In 2023, the biggest over-performers were Miami, which parlayed a run margin 
of -57 into 84 wins and a playoff spot.  That was 10 wins more than predicted.
Three teams  Baltimore, Detroit, and Pittsburgh  won 6 games more than 
their run margins would normally support.

The biggest under-performing team this season was San Diego, which would 
normally be expected to cruise into the postseason with 93 wins on the back 
of a +104 run margin.  Instead, they finished with only 82 wins and missed 
the tournament.  Much the same could be said about their NL counterparts in 
Chicago, who outscored their opponents by 96 runs but won only 83 games.  In 
the AL, Texas fell 8 games short of expectations, but made up for it in 
October.

Converting offensive events into runs
-------------------------------------

Just as there is a strong relationship between runs and wins, it is also true 
that the more walks, singles and extra-base hits you produce, the more runs 
you score. Sometimes, of course, a productive team comes up short on the 
scoreboard because they did not hit in the clutch, did not run the bases 
well, or hit line drives right at people in key situations. 

To shed light on this relationship, we need a way to take batting stats and 
turn them into a measure of overall offensive production. There are several 
good options here, including Runs Created (Bill James), Batting Runs (Pete 
Palmer), and Base Runs (David Smyth).

For this exercise, we will use the sum of total bases, walks, and hit by 
pitch, or TB+ for short. TB+ is not a perfect measure, but it does capture 
the most important things a team does to produce runs and it is easy to 
figure without a computer.

As with other statistics, a team's TB+ total can be heavily influenced by its 
home park. For that reason, we focus on the difference between the TB+ 
produced by a team's hitters and the TB+ allowed by its pitchers. This 
effectively removes the park from the equation and helps us identify the 
teams that outproduced their opponents by the largest margins.

2023 AL
-------

The following table shows the offensive and defensive TB+ figures for the 
2023 AL, along with the difference between these two figures and each team's 
league rank based on those differences. It also shows runs for and against, 
the run differential, and the rankings based on run differential.  If you 
see a team that ranked much higher on TB+ than runs, or vice versa, that's
an indication that the team was unusually efficient or inefficient.

The second table lists each team's win total, league rank, Pythagorean win
total, difference between actual and expected wins, and record in one-run 
games.  

                   ---------- TB+ ---------   -------- Runs ---------  
Team                Off    Def  Diff   Rank   Off   Def   Diff   Rank  
Baltimore          2870   2711   159     7    807   678    129     3t   
Tampa Bay          3061   2604   457     1    860   665    195     1   
Toronto            2936   2767   169     6    746   671     75     7   
New York           2719   2741   -22     8    673   698    -25     9   
Boston             2917   2979   -62     9    772   776     -4     8   
                                 
Minnesota          3033   2683   350     3    778   659    119     5   
Detroit            2647   2714   -67    10    661   740    -79    11   
Cleveland          2629   2783  -154    12    662   697    -35    10   
Chicago            2540   3076  -536    14    641   841   -200    14   
Kansas City        2640   3032  -392    13    676   859   -183    13   
                                    
Houston            3053   2833   220     5    827   698    129     3t  
Texas              3183   2750   433     2    881   716    165     2   
Seattle            2928   2628   300     4    758   659     99     6   
Los Angeles        2936   3056  -120    11    739   829    -90    12   
Oakland            2550   3287  -737    15    585   924   -339    15   

Actual wins versus Pythagorean wins, record in 1-run games

Team               Actual   Rank    PythW    Diff     1-run
Baltimore           101       1       95      +6      30-16
Tampa Bay            99       2      101      -2      22-25
Toronto              89       5       90      -1      25-20
New York             82       8       78      +4      17-23
Boston               78       9t      81      -3      18-25
                                          
Minnesota            87       7       94      -7      19-27
Detroit              78       9t      72      +6      21-20
Cleveland            76      11       77      -1      27-31
Chicago              61      13       60      +1      19-30
Kansas City          56      14       62      -6      21-20
                                          
Houston              90       3t      95      -5      20-21
Texas                90       3t      98      -8      14-22
Seattle              88       6       92      -4      25-26
Los Angeles          73      12       72      +1      22-21
Oakland              50      15       46      +4      20-27

Summing up for each team:

Baltimore  first in wins but only seventh in TB+ differential, mainly 
because their offense was unusually efficient (9th in offensive TB+ but 4th 
in runs) and they excelled in one-run games ... as a result, theyre 
unlikely to come close to 101 wins total in most DMB season replays

Tampa Bay  statistically the most impressive team in the league

Toronto  a good season at all three levels

New York  eked out a winning season on a slightly below average TB+ 
differential by outperforming their Pythagorean record

Boston  if theres such a thing as a respectable last-place finish, this is 
one

Minnesota  won the division comfortably, but it could have been a romp if 
they had turned their #3 ranking in TB+ differential into the expected 
high-90s win total

Detroit  their 78 wins were a little higher than their stats would normally 
suggest

Cleveland  their run margin and wins were in line with their stats

Chicago  a very bad season at all three levels

Houston  very good at all three levels ... a big drop from the 2022 team 
that produced a TB+ differential of +604 

Texas  a little better than Houston in every measure except wins

Seattle  statistically good enough to make the postseason

Los Angeles  a below-average season at all three levels

Oakland  the second-worst TB+ differential in the 162-game-season era

2023 NL
-------

                   ---------- TBW ---------   -------- Runs ---------
Team                Off   Def   Diff   Rank   Off   Def   Diff   Rank
Atlanta            3408  2786    622     1    947   716    231     1
Philadelphia       3034  2703    331     3    796   715     81     5t
Miami              2705  2841   -136    10    666   723    -57    11
New York           2816  2862    -46     8    717   729    -12     7
Washington         2686  3193   -507    14    700   845   -145    14
                                                               
Milwaukee          2723  2622    101     6    728   647     81     5t
Chicago            2963  2750    213     5    819   723     96     4
Cincinnati         2955  3155   -200    12    783   821    -38     9
Pittsburgh         2733  2947   -214    13    692   790    -98    12
St. Louis          2931  3103   -172    11    719   829   -110    13
                                                               
Los Angeles        3241  2705    536     2    906   699    207     2
Arizona            2817  2921   -104     9    746   761    -15     8
San Diego          2942  2697    245     4    752   648    104     3
San Francisco      2700  2721    -21     7    674   719    -45    10
Colorado           2730  3376   -646    15    721   957   -236    15


Actual wins versus Pythagorean wins, record in 1-run games

Team               Actual   Rank    PythW    Diff     1-run
Atlanta             104       1      103      +1      23-18
Philadelphia         90       4       90       0      29-24
Miami                84       5t      74     +10      33-14
New York             75      12       80      -5      25-28
Washington           71      13t      66      +5      28-21
                                  
Milwaukee            92       3       91      +1      29-18
Chicago              83       7       91      -8      21-24
Cincinnati           82       8t      77      +5      34-29
Pittsburgh           76      11       70      +6      22-17
St. Louis            71      13t      70      +1      17-26
                                  
Los Angeles         100       2      102      -2      16-15
Arizona              84       5t      79      +5      21-21
San Diego            82       8t      93     -11       9-23
San Francisco        79      10       76      +3      25-19
Colorado             59      15       59       0      23-24
  
Summing up for each team:

Atlanta  best in the league at all three levels, and the second-best TB+ 
differential in Atlanta history, behind only the 1998 team

Philadelphia  fully worthy of their record and their deep postseason run

Miami  unimpressive statistically but excellent in one-run games ... will 
not be able to replicate their 84 wins in most DMB season replays

New York  nearly a .500 team based on the underlying numbers

Washington  71 wins flattered a team that was second last in TB+ 
diffential and run margin

Milwaukee  a solid season at all three levels

Chicago  shoulda been a contender ... outperformed Milwaukee in both TB+ 
and run differential

Cincinnati  their 82 wins masked a 13th place finish in TB+ differential

Pittsburgh  posted a respectable win-loss record despite being third last 
in the league in TB+ differential

St. Louis  not a good season, but they might have sneaked into third place 
in the division if their wins and runs reflected their TB+ differential

Los Angeles  excellent at all three levels

Arizona  like Miami, they snagged a playoff spot despite weak TB+ and run 
differentials ... unlike Miami, they got hot in October and almost went all 
the way

San Diego  definitely one the of the leagues four best teams ... might have 
made some noise in October had their dismal record in one-run games not kept 
them on the sidelines

San Francisco  a little below average at all three levels

Colorado  historically bad ... last by a large margin at all three levels 
in 2023 ... since 1962, only three teams have had a worse TB+ differential, 
and all were expansion teams in their first six years of existence

Rating the players
------------------

We have created a database in which all players match their real-life 
performances when the season is simulated many times and the results are 
averaged.

But this article is about the times when the sum of the parts (the aggregate 
performance of the players) doesnt match the whole (the teams win-loss 
record).

When those two things are out of sync, there is no way for DMB to accurately 
reproduce both the player performances and the real-life win-loss records.  

We could force teams to match their real-life win totals by distorting the 
ratings of some of their players, but those distortions would be highly 
undesirable in DMB leagues where players are drafted or traded onto new 
rosters.

Instead, we choose to rate the players accurately and accept that some teams 
wont match their real-life records.  In a season replay, DMB teams will 
produce win-loss records they would have in real life had those real-life 
teams experienced normal relationships between (a) batting events and runs 
and (b) run differentials and wins.
