                             Team Efficiency
                             ---------------

How do you measure a teams performance in a season?

Lets start at the top and then peel back the layers to see if we can gain
any insights that are not obvious at first glance.

In some sports leagues, most notably in domestic soccer leagues in all
parts of the world, the big prize goes to the team that amassed the most
points in a nine-months-long season.

In baseball, we give the biggest prize to the team that (a) qualified for
the postseason tournament and (b) won that tournament.  And so we can say
that Atlanta was the top team in 2021.

Then again, if baseballs reward system was like soccers, we would say
that San Francisco was best.  They won 107 games, one more than their
rivals in Los Angeles, and 19 more than Atlanta.  And because the two
leagues play mostly separate schedules, we must also mention Tampa Bay,
which won 100 games in a league that, based on the results of inter-league
play, was a little stronger.

For the sake of our peeling-back-the-layers notion, lets say that these
things  titles and wins  make up the top layer.

The next layer consists of the things that produce wins.  In baseball, that
would be runs.  You win by scoring more runs than you allow.

By this measure, Los Angeles was the best team in 2021.  They outscored
their opponents by 269 runs.  San Francisco was excellent, too, with a run
differential of +210.  Over in the other league, Tampa Bay was again on
top, with +206, with Houston only one run behind.

This is a good place to stop and consider the relationship between runs and
wins.  Most of the time, these things are fairly tightly coupled.  Outscore
your opponents by 80 runs and youll most likely win 90 games, plus or
minus a few.  Get outscored by 50 runs and your win total will be close to
76.

There are exceptions, though, and we saw a historically big one this year.
More on that in a bit.

Peeling back the runs layer reveals the things that produce runs -- drawing
walks, hitting singles, and clouting extra-base hits.  Teams can also
increase their scoring chances with bunts and steals and good baserunning,
but hits and walks are the biggest drivers.

The most straightforward way to outscore your opponents is for your hitters
to produce more bases on hits and walks than your pitchers give up.  Well
measure this with something we call total bases plus, or TB+ for short.
This is the total bases on hits -- one base for a single, two for a double,
and so on  plus walks and HBP.

Subtracting TB+ allowed from TB+ produced gives us another good measure of
team performance.  It stands to reason that if a team leads the league in
TB+ differential, it has a very good chance to lead the league in run
differential and wins.

By this measure, the top teams in 2021 were Los Angeles (+618) and San
Francisco (+586).  No surprise there.  Over in the other league, we have a
new top-team candidate.  Houston outproduced its opponents by 450 bases,
comfortably ahead of Tampa Bay and Toronto, both at +405, and Chicago,
with +361.

When we talk of team efficiency, were examining the relationships between
these layers.  How well did a team translate its TB+ differential into its
run differential?  And how well did a team convert its run differential
into wins?

Sometimes the hits and walks that make up TB+ are clustered into big
innings and sometimes theyre scattered and dont lead to much.  Even over
a long season, this doesnt always average out.

The same is true with runs and wins.  Weve all seen examples where teams A
and B play a three-game series in which A wins by 3-2 and 6-5 scores while
B wins the other game 12-1.  A has more wins even though B took the run-
differential battle by a big margin.  This sort of disconnect can also
happen over a full season.

The biggest outlier weve seen in a long time was Seattle, whose hitters
produced 176 fewer TB+ than their opponents and, not surprisingly, were
outscored by 51 runs on the season.  These numbers are consistent with a
win total in the low-to-mid-70s, but this team somehow managed to post a 90-
72 record, fully 15 wins better than their run margin justifies.  Thats
among the biggest differences in baseball history.

We must travel far to the east and quite a bit south of Seattle to find the
most significant under-achiever.  We speak of Toronto, which outscored its
opponents by 183 runs, third-best in the league, and was tied for second in
TB+ differential.  This would normally be enough to produce 100 wins, but
they won only 91 and missed the postseason by a single game.  Perhaps their
strange Covid-related journey  splitting home games between Florida,
Buffalo, and Toronto  threw everything off kilter.

The two teams that edged Toronto on the final weekend of the season were
Boston and New York, and neither had the statistical underpinning to
justify their places in October.  Both finished with run differentials more
than 100 lower than Torontos, and their expected wins were 89 and 86,
respectively.

Youve probably already put the pieces together for San Francisco and Los
Angeles.  The latters 106 wins were fully supported by their underlying
stats  they were very good at converting TB+ to runs but offset those
gains by coming up short on converting runs to wins.  San Franciscos
production, on the other hand, was more consistent with 100 wins than 107.

Atlanta was a very good team that was at its best at the right time.  They
finished with 88 wins, but their +134 run differential was good enough for
a 95-66 record.  They offset that shortfall a little by squeezing a couple
of extra wins out of their TB+ differential.  All in all, its fair to see
them as a 92- or 93-win team.

Milwaukee also had a very good season.  Their TB+ differential of +277 and
run margin of +115 were fourth best in the league, and they finished with
the third-highest win total.  They were a little inefficient in run
prevention, though, finishing with a team ERA that was about 20 points 
higher than normal given their underlying stats.  So they could have won a 
few more games than they did.

The remaining exceptions can be found at the bottom of the standings.
Several teams won six or eight more games, or six or eight fewer, than their
runs and TB+ would normally generate.  But because these teams  some of
them, at least -- were carrying out aggressive rebuilding programs and not
really trying to compete this year, its not clear that theres much to be
learned by digging into their underlying stats.

What does all this mean for the Diamond Mind Baseball game and our annual
Season database?

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

When a real-life team's win-loss record is out of sync with the production 
of its players, as was the case for several of the examples we just wrote 
about, there's no way to accurately reproduce both player performances and
real-life win-loss records at the same time.

The only way we could force some teams to match their real-life win totals is
to distort the ratings of their players.  If we make them perform better or
worse in your simulated games than they did in real life, we can make their
teams come closer to their real-life win totals.  But we dont do that.

Those distortions would be highly undesirable in DMB leagues where these
players are drafted or traded onto new rosters, so we choose to maintain
the integrity of the player ratings.  We let the teams perform as they
would have in real life had they experienced normal relationships between 
(a) offensive events and runs and (b) run differentials and wins.

For instance, suppose you autoplay the 2021 season once, in the most
realistic way possible -- using real-life transactions and lineups.
Seattle will almost certainly fall far short of 90 wins.  If you do this
over and over again, theyll reach 90 wins only about once every 200
seasons.  One can only wonder whether the same would be true if the real-
life season was played that many times.
