August, 2008

What Might Have Been

mul�ti�verse [ m�ltee v�rss, m�ltī v�rss ] noun Definition: hypothetical cosmos of multiple universes: a hypothetical cosmos that contains our universe as well as numerous other universes and space-times.

I have often dreamt of the possibility that in some parallel universe, the Cowtippers are currently six-time BDBL champions, and the '08 'Tippers are locks to make it seven titles in ten seasons.  Not only that, but I haven't traded John Danks and Jon Lester, Miguel Cabrera is enjoying an MVP season, Kenji Johjima is hitting his weight, Erik Bedard and Dustin McGowan aren't injured for the season, and the 2009 Cowtippers are looking like surefire locks to win it all again next year.

No doubt, in that alternate universe, I am the only human GM remaining in the league, as all of the other owners have grown so tired of seeing me win year after year, they've all departed, leaving me to manage every one of my games against the MP.  But still, I am happy.

The beauty of Diamond Mind baseball is that you can simulate this dream scenario and play out the season several times in order to see what might have happened had a few random bounces gone differently.  Since the inception of this league, it has been an unwritten rule that we should not simulate any full seasons using the league disk, as it would "ruin the surprise" of the coming season, and may prompt teams to make premature decisions based on a few random sims.  However, I'm going to break that unwritten rule for the simple reason that I have nothing else to write about this month.  Don't try this at home.

For the purpose of this exercise, I ran ten full simulated seasons using our current player disk.  The downside to this method is that I'm using the rosters as they exist today -- after a ton of trades have been made.  It would be more interesting to run these sims using our Opening Day rosters, but that would take too much work, and I'm very lazy.  Another big difference is that not all MPs may be set to their optimum usage.  And another difference is that the sims don't follow our usage rules, so some SUS's may get more playing time than they are allowed.  And the sim seasons also include random injuries that last more than four games.  But for the purpose of this exercise, this is all acceptable enough, I think.

One thing I have learned through this exercise is that the regular season is FAR more random than I had ever suspected.  Below are the results of my ten sims, along with some projected numbers from our current season.  I'll explain what all of this means after I show you the numbers:

The first three columns (W, L, Pct) display the average wins, losses and winning percentage for the ten simulations.  (Note that the wins and losses have been rounded, while the PCT is the result of the unrounded numbers -- which explains why the Infidels and Irish Rebels own the same W/L records, but different winning percentages.)

Next, I show the average runs scored and runs allowed from these ten sims.  The "pRF" and "pRA" columns show the projected runs scored and allowed for every team, using our current stats projected over a 160-game season.  The "Diff" columns show the difference between the current projections and the simmed seasons.

The "Div" column tells us how many times a team won its division in the ten sims, and the "HiW" and "LoW" column gives us the maximum and minimum number of wins for each team in any of the ten sims.

Finally, the "pW" column shows the projected wins for our current 2008 season, based on our records as of 7/22.  And the "Diff" column shows the difference between this total and the average wins total for the ten sims.

Got all that?

The most intriguing part of this whole table, to me, is the enormous spread between one season to the next.  For example, there is a 34-WIN DIFFERENCE between the Kansas Law Dogs' best (103 wins) and worst (69 wins) performances during these ten sims.  How on earth can the same team, playing the same schedule, in the same order, win 100 games one season and nearly lose 100 games the next?  It seems inconceivable.  It seems illogical.  It seems SO incredibly random!

In general, the difference between wins from one season to next (i.e. the standard deviation) for all teams is 6.2.  To me, this seems like a HUGE disparity, as most division races are decided by fewer than six wins per year.  It's certainly a much higher rate of randomness than I had expected.

The results of the ten sims are also very interesting, in that they tell us which teams are currently performing well outside "the norm."  Let's review three of these teams:

New Milford Blazers

No surprise here.  The fact that New Milford continues to hover above .500 at this late point in the season has been the source of countless neck injuries.  (All that shaking of the head, back and forth, in disgust.)  It makes no logical sense.  And that's because it's an enormous statistical outlier.

In ten simulated seasons, the Blazers won no more than 77 games.  Yet they're currently on pace to win 86.  They averaged 97 losses per season in the ten sims, yet they're on pace to lose just 74.  What are the statistical odds?  I'm not smart enough to calculate it.  But I do know they're greater than 1-in-10.  If I had run 100 sims, would New Milford ever win more than 85 games in any one of those sims?  Unlikely.

So what has caused New Milford to win at such an alarmingly illogical pace all season?  I've already covered this in an eariler FTDOTC article.  But it's also interesting to note the difference between the "RA" and "pRA" columns above.  The Blazers have allowed 100 fewer runs this season than they averaged in the ten simulated seasons.  How on earth could this be?

New Milford's starting pitchers have compiled an ERA of 5.26 this season -- the fourth worst rate in the BDBL.  New Milford's relievers have compiled an ERA of 2.98 this season -- the fifth best rate in the BDBL.  And New Milford's starters have pitched an average of just 5.3 innings per game, which is the lowest figure in the league.

This means one of two things: either the Blazers have scored a TON of runs this year, and it doesn't matter how poorly their starters have pitched, or the Blazers have won a LOT of games in late innings after falling behind early.

Using BASE, we can see that New Milford has come from behind five times after the fifth inning.  That's not enough to explain the discrepancy.  A look at run scoring by inning isn't particularly revealing, either.  And looking at situational hitting stats, no unusual patterns leap out.  New Milford hits roughly the same in the clutch as they do in other situations.

So what on earth has led to this 100-run disparity?  Your guess is as good as mine.  Perhaps Peburn deserves some of the credit for this performance.  Realizing the shortcomings of his starting rotation, Peburn has yanked his starting pitcher far earlier than the MP likely did on most occasions.  And because the bullpen has performed so well -- better than they did in any of the ten simulated seasons -- the Blazers have profited as a whole.  Perhaps the bullpen has far exceeded their expected performance because Peburn has been able to cherry-pick the right match-ups.  Or perhaps he's just lucked out.  I'll let you be the judge.

Not only has the bullpen stepped up their performance by a HUGE amount, but the New Milford offense has also out-produced their alternate-universe sim cousins by 73 runs -- the second-largest disparity in the league.  It's difficult to pinpoint what caused this disparity, but perhaps it's also a function of playing the right match-ups at the right times.

Let's give Peburn some credit.  But not too much.  The Blazers have been both good AND lucky, and as a result, they're outperforming even their loftiest possible projections.

San Antonio Broncs

At the other end of the spectrum, the Broncs are UNDER-performing to the tune of 16 wins.  The culprit here is the offense, which has undershot its average runs scored by 80 (projected) runs.

But believe it or not, the primary reason for this offensive slump is not Jimmy Rollins.  In ten simulated seasons, Rollins hit just .252/.302/.423, which isn't too far off from his current BDBL average of .235/.279/.367.  Well, okay, it's still pretty far off, but still well below what he hit in MLB (.296/.344/.531.)  Evidently, DMB just doesn't like Jimmy Rollins.

In ten sim seasons, the Broncs never lost more than 84 games.  Yet this season, they are on pace to lose 93.  The reasons for that go far beyond the performance (or lack thereof) of Rollins.  In one sim, San Antonio finished with a record of 96-64.  What was the difference between that team and the current Broncs team?

In that 96-64 season, the Broncs hit .242/.309/.409 as a team -- hardly awe-inspiring, and barely better than their current .239/.306/.383 numbers.  The main difference here was that Shelley Duncan was able to amass 548 at-bats in the simmed season, due to the absence of usage rules.  He hit 55 homers and created over 100 runs.  Substitute his performance for Conor Jackson's, and you could knock 40 runs off the tally, and perhaps subtract as many as four wins.  But that still makes this a 92-win team, which is far better than the Broncs team we see in this dimension of the universe.

The most noticeable difference between the two teams is on the pitching side of the ledger.  John Lackey (18-6, 2.70 ERA), Mark Guthrie (15-6, 3.28), Danny Haren (14-13, 3.35), Chad Durbin (11-5, 3.95) and Wandy Rodriguez (10-9, 4.12) formed a formidable starting rotation in the simmed season.  In our current season, Lackey (9-10, 4.40 ERA) has been a huge, $17 million disappointment.  Haren (3.13 ERA) has pitched well, but has lost 12 games in 18 decisions.  Guthrie (3.49 ERA) has also pitched well, but has lost nearly twice as many games as he has won (4-7.)  Durbin owns a 5.93 ERA in just 13+ innings, and Rodriguez owns a 5.03 ERA in 107+ innings.  Taking Durbin's innings is Orlando Hernandez, who has posted a nice 3.84 ERA, but has also lost over twice as many games as he has won (2-5.)

Another major difference is the performance of the bullpen.  Renyel Pinto posted a 3.79 ERA in 73+ innings in the simmed season, but is currently sporting a 6.42 ERA in 40+ innings for the Broncs in reality.  Edinson Volquez (3.83 ERA in 47 IP) also played a big role in the simmed season, while he's hardly pitched in reality.

Finally, another big difference is the alternate-universe Broncs' performance in the clutch.  That team went 28-27 in one-run games, 12-6 in extra innings, and came from behind 11 times after trailing in the seventh inning.  The current Broncs team is 15-22 in one-run games, 7-4 in extra innings, and has come from behind after trailing in the seventh inning just once all season.

So, basically, the difference between a 96-win team and a 67-win team boils down to one good hitter, better performances from two starting pitchers and a couple of relievers, and better luck in tight games.

Akron Ryche

Akron GM D.J. Shepard has recently announced that he is throwing in the towel on the 2008 season.  But in our ten simmed seasons, the Ryche (as presently constituted) won an average of 93 games (third best in the Eck League) and eight division titles (second only to the SoCal Slyme.)

Akron won 105 games in one of the sims, and only 79 games in another.  Yet the current version of the Ryche are projected to win just 78 games -- and that was before their recent four-game sweep was reported.  So what happened?  How did the wheels fall off the Akron bandwagon?

Like the Broncs, the Ryche aren't hitting nearly as well (-53 runs) as they did in the sims, nor are they pitching as well (+78 runs allowed.)  Let's take a look at Akron's 105-game- winning team to figure out where things went wrong.

Here, the differences are even more subtle than San Antonio.  Aramis Ramirez hit a little better in the sims (.332/.391/.598) than he has in "real life" (.275/.333/.561.)  Todd Helton (.316/.415/.508 sims, .285/.372/.493 BDBL) also hit a little better.  The simmed Ryche also got a full season (544 at-bats) out of Dan Ortmeier (.322/.357/.550), which certainly helped.

But the biggest difference in performance was Brad Hawpe, who hit .283/.407/.543 in the simmed season (30 HR, 112.6 RC), but is hitting just .215/.311/.358 with 11 HR and 37.7 RC for the Ryche.  These differences add up, and the result is that the simmed Ryche hit .275/.343/.460 as a team, while the present-day Ryche are hitting just .250/.318/.411.

On the pitching side, the simmed Ryche own an ERA of 3.70, while the actual Ryche's ERA is 4.25.  And the difference can be attributed mostly to three pitchers:

• In the sims, Carlos Villanueva's ERA is 4.15 over 158+ innings.  In reality, Villanueva's ERA is a stellar 8.27 over just 16+ innings.
• In the sims, Bronson Arroyo owns a 4.24 ERA in 225 inning.  In reality, Arroyo's ERA is 5.34 in 119+ innings.
• In the sims, Bradon Looper's ERA is 4.74 in 188 innings.  Not pretty, but good enough for a 13-7 record.  In reality, Looper's ERA is 5.04, and he owns a 2-8 record in 103+ innings.

The sims prove that the Akron team is certainly talented enough to win the Hrbek Division.  The odds tell us that if Akron isn't completely dismantled over the final two chapters, they could catch the Black Sox and win the division.  But odds aren't a guarantee.  That's why they're called "odds."

Overall, only the SoCal Slyme won their division in every one of the ten sims.  The Slyme won no fewer than 95 games in any simmed season, and yet they are on pace to win 12 more games than their simmed average of 102.

Another interesting team is the Allentown Ridgebacks, who have pretty much performed as well as expected this season.  (Their sim average is just six wins more than their current pace.)  Yet the Ridgebacks currently trail the first-place Kansas Law Dogs by three games.  And if the season ended today, Allentown wouldn't even be in the playoffs.  In ten simulated seasons, the Ridgebacks won their division eight times, and yet they may not make the playoffs this year.  It just goes to show that there's a reason why we go through the trouble of actually playing these games.  And I can't help but wonder how much of a difference it makes actually managing these games versus simply letting the computer manager make your decisions for you.  As we've seen with New Milford's performance this year (and in past years), it could make a huge difference.

Another interesting race is taking place in the Griffin Division, where the Bear Country Jamboree currently lead by four games over the Los Altos Undertakers.  In our ten simmed seasons, Los Altos and Bear Country each won the division five times.  Los Altos averaged 91 wins per season, while Bear Country averaged 90.  You couldn't get a much tighter race than that.  But this season, the Undertakers have scored 58 fewer runs than their ten-sim average, while Bear Country has scored 64 runs more than their sims.  And this has made all the difference.

In yet another tight race, Manchester and Ravenswood finished with identical 82-78 average records after ten sims.  Yet in reality, Ravenswood owns a TWELVE GAME LEAD over Manchester.  How on earth can that be?  Well, Manchester has scored 59 fewer runs than their ten-sim average, while Ravenswood has outscored their sim average by 10 runs.  And while Ravenswood has allowed 26 fewer runs than their sims, Manchester has allowed just 5 fewer runs.  The striking difference between these two teams is that Ravenswood is 17-8 in one-run games while Manchester is just 9-16.  Ironically, after trading Matt Cain (and Carlos Quentin) for Jonathan Papelbon in the pre-season, Manchester GM Jim Doyle still may not have acquired enough bullpen to win those tight games.

This exercise has taught me a great deal about the randomness of this game.  Up until now, I had always thought that the long, 160-game season evened out all of the randomness and gave us an accurate picture of a team's true talent.  But now, I realize that there can be a swing of as many as 34 games for one team from one season to the next, and that randomness plays an enormous role in whether a team dominates with 100+ wins or merely plays "also-ran" baseball with a .500 record.

This exercise also makes me appreciate the truly momentous achievements we have seen in our league through the years.  It has made me appreciate just how rare it is to see a team win 114 games, or a hitter like David Ortiz club nearly 80 home runs.  Instead of whining about how "unrealistic" the game is, we should be celebrating these achievements for what they are: a one-in-a-thousand rarity.

Viva las Blazers.