August 14. 2005

Often times we really don't know why one team won while another lost.  We look at the numbers and if they don't fit within the parameters of our understanding we often attribute the successes and/or failures to luck.  Sometimes, it's not until some time later when analysis tools are greatly improved do we truly understand what happened.  I'll give you an example from history.  It has long been held that the English won the battle of Agincourt in 1415 because English longbowmen cut down French knights in droves as they advanced on the English position.  It was good King Henry V, English courage and the English long bow that carried the day on that St. Crispian's Day.  It's a version of the battle that has been celebrated for the last 600 years.  The problem is that the facts show nothing could be further from the truth. 

The use of bows in battle had been around for centuries, dating back to pre-Roman warfare.  So why would bows in the hands of ale-guzzling pale white fellows suddenly become substantially more lethal some 1500 years after they were first introduced?  The fact is, they weren't.  The french army was largely comprised of nobles, most of whom were wearing the best available armor of the day.  The material that armor was made from was a relatively new development called steel.  If you take a typical english arrow with its point of iron and fire it with a longbow at a 2mm-thick plate of steel (the common thickness of armor) the arrow head crushes against itself.  It does nothing to the steel, barely even scratching it.  The arrows were effective against the less-armored horses of the French cavalry, but that doesn't account for what happened to the other 10,000+ or so frenchmen on foot, not does it account for the actual guys who rode on the horses. 

So if the arrows weren't the reason, how did barely six thousand Englishmen manage to rout around 14,000 French on their home field?  There were three primary reasons and it wasn't until investigative science advanced significantly that they became obvious. 

The first reason was the field itself.  Agincourt's soil becomes a very clumpy mud when wet.  The individual grains of dirt are especially porous and after weeks of rain that preceded the battle, the ground was essentially a very sticky quicksand for anyone wearing heavy armor.  However, the mud wouldn't stick to more porous materials like cloth or leather.  So a steel-armored French knight would add 10 pounds of mud on each leg walking through it, and was additionally slowed from having to overcome the force of suction from pulling his foot out with each step.  But on the English side, a cloth-armored archer/footman could basically run unencumbered through the slop and only get a little dirty.  By the time the French knights closed on the English line, they were probably too exhausted and too encumbered to move.  The English archers, who became footmen in a melee, merely ran around them and stabbed the immobile and fatigued knights in the neck or eye hole with their daggers and pointy sticks. 

The second reason was that the topography of the battlefield was a funnel right to the English line.  Trying to cram 14,000 knights in armor into a field 300 yards wide is not an easy task, especially if a couple of knights tripped or fell.  The effect would be like the 1979 Who concert in Cincinnati where 11 people were crushed to death trying to get into the festival seating at Riverfront.  Once one fell, others would stumble over and the next thing you know there'd be a dog pile of french knights with the bottom guys either getting crushed to death or suffocated in the mud.  Moreover, people laying face down in the mud aren't particularly difficult to dispatch, irrespective of whose banner they fight under. 

The third reason there were such a high number of French casualties is that Henry V killed the prisoners he took for fear that they would re-arm themselves from the battlefield if there was a second French attack.  So hundreds, perhaps thousands of French nobles and soldiers were killed after they had stopped fighting and had surrendered their weapons.  It's quite different picture than what you read in Shakespeare, isn't it?  But these are the facts. 

So how does this pertain to baseball?  Well, just like Agincourt, contemporary analysis isn't always accurate or even close to the truth.  Sometimes the tools aren't applied correctly.  Sometimes they are simply inadequate to do the job.  Take the Pythagorean theorem, for example.  As applied to baseball, it maintains that a team's winning percentage should be very close to the square of it's runs scored divided by the square of its runs scored plus runs allowed.  But can it be used to determine a team's expected record using an incomplete season's sample, especially with an unbalanced schedule?  This has been forwarded extensively in an effort to analyze the Nationals' success (or recent lack thereof) this season.  But is that a reliable test?  What happens if a team reworks itself midway through the season with either trades, call-ups from the farm or a strategic change in manager style?  That would mark a significant change in the potential of the team, wouldn't it?  Is it then still accurate to use the overall season totals as the data?

For example, if you look at the Marlins' run differential in 2003 when they won the World Series, they scored 751 runs and allowed 692 over the full regular season.  But if you look at what they did after they fired not-so-competent Jeff Torborg 38 games into the season and went with largely competent Jack McKeon, it tells a different story.  They scored 588 runs while allowing 506 the rest of the way.  Had McKeon been at the helm from the beginning, at that pace for the full season they would have ended up scoring 768 while allowing 661.  The difference is nearly 50 runs.  In Pythagorean terms, that's the equivalent of five wins and a ten game swing in the standings.  And that's assuming that the general theorem actually yields an accurate result.  Also consider that the Marlins bolstered their bullpen down the stretch with the addition of Ugueth Urbina (who became their closer in the playoffs) thus making the team less susceptible to late inning blow-ups which are often camouflaged in the overall run differential.  People (myself included) just looked at the Marlins' overall season numbers and underestimated them as they entered the playoffs.  Our mistake was that we weren't actually appraising/analyzing the team that actually made the playoffs.  We were looking at a conglomeration of data of which a significant portion was no longer germane to the discussion.  In this particular case, an improvement to the bullpen and a change in the way the team was used had a significant impact.  The effect of these changes, however, can be hidden in the runs equation despite the obvious change in the won-loss record.  This may be one of the reasons why my Strat-o-matic team made the playoffs in the SOMBOE (Strat-o-matic Bevy of Experts) League. 

Just as I predicted at the beginning of the season, an effective bullpen can have a pretty significant effect on the standings.  Just comparing the overall runs scored versus runs allowed numbers, Tristan Cockroft's Bank of New York team (805 runs scored, 778 runs allowed) should have been expected to finish higher than my Montreal Funiculars squad (732 and 742) in the division.  According to the Pythagorean theorem, Tristan's team should have won the division by 4 games.  I believe the bullpen is where the difference in our respective won-loss records ultimately was determined. 

The computer who actually manages the games, HAL, is notoriously erratic in the way it employs each team's resources despite the number of restrictions you can place on it.  One of the things HAL likes to do is use the bullpen, especially in situations where no rational manager would ever consider... like when a starter is throwing a shutout through 8 innings and has only thrown 90 pitches... in comes the team's worst reliever to start the ninth.  It's not necessarily that HAL picks bad relievers every time out, but that it uses them so much that unless they get regular rest, they will become fatigued and lose effectiveness.  And that is one of the things I suspect helped me and hurt Tristan: both he and I began the season with just 5 relievers, but when I noticed worse than expected performances occuring regularly after the first month of games, I expanded my bullpen to 7 relievers.  After that, Eric Gagne and Eddie Guardado performed largely as expected, and Neal Cotts and Danny Graves surpassed my expectations.  On Tristan's squad, Latroy Hawkins, Cal Eldred and Troy Percival were much worse than anticipated over the course of the full year and I suspect it was largely due to HAL's overuse. 

And while his relievers were certainly respectable, even in the best of circumstances they were by no means a lock to close out the game.  Hawkins, Percival, Damaso Marte and Lance Carter all showed vulnerability in closing out the opposition with some frequency in 2004.  That showed up on their cards and apparently in their performance.  Eric Gagne and Eddie Guardado were much less vulnerable to meltdowns.  Just as in other sim games I had played previously, a great closer nearly always sealed the win once he was brought in and a strong bullpen kept close games winnable until the end.  It's impossible to avoid blowing leads entirely for a full season, but with a strong bullpen one can minimize the damage.  Fewer than average blown wins resulted in a better than expected winning percentage. 

The other aspect of my team's effectiveness that was obscured by the overall numbers is the changes in personnel I made over the course of the season.  After the experiments with Erik Bedard, Roy Halladay and Scott Kazmir proved fruitless after 60 or so games, I exchanged them for Mike Mussina, Cris Carpenter and Travis Smith.  I probably made up 20 runs in differential in the second half just from these changes to the pitching staff.   Both Carpenter and Mussina finished with ERAs in the 3s while Halladay and Bedard had struggled to stay below 5.  Smith ended up with an ERA of 2.75 in limited use, far superior to Kazmir's unspeakable disaster as the mop-up man.  Yet all those runs that Halladay, Bedard and Kazmir had allowed were still part of my overall season record thereby concealing significant improvements to my team's run prevention capability if you only looked at runs scored versus runs allowed..

I also improved the offense.  Ben Broussard had been highly regarded in Strat circles for his defense and doubles power, but he was a complete bust for me.  Same for Brian Roberts.  Broussard was supposed to hammer lefties, but since there were so few being employed as starters in our league that ability was largely useless.  Further lessening  his value was that even when he faced them he was completely ineffective.  Roberts was nearly as futile from both sides of the plate.  After a brief trial with Paul Konerko at first, I settled on Justin Morneau at first and Willie Harris at second.  As it turned out there wasn't a significant difference on either side of the ball between Harris and Roberts - just a few points in slugging - and Harris proved to be a much more aggressive basestealer.  The defensive drop-off at first from Broussard to Morneau, while noticeable, wasn't as significant as the improvement in slugging that Morneau provided.  Morneau lived up to his billing as a home run threat against right-handers and while he was awful against lefties, their scarcity effectively minimized his downside.

So by the time playoffs rolled around, the team I was fielding was not only much better than the one I started out with, but one that was comparable to the other contenders.  In fact, if you take home field advantage out of the equation and just look at road record, my Funiculars had the second best mark (45-36) in the league, a mark better than any team that made the playoffs.   There was a strong case to be made that not only was my squad deserving of a playoff spot, but was a real threat to win it all on merit despite an inferior overall record and run differential.  The first round series brought the point home: it was so evenly matched (against JP Kastner's Milwaukee Pub Crawlers club) that the series went to a 5th and final game that remained scoreless until the ninth inning.  It was ultimately decided by a solo homer off Roger Clemens, perhaps coincidentally, by Justin Morneau.

In the finals, I faced off against Lawr Michael's BALCO Ballerinas squad, a team that had stomped mine with impunity through the first half of the season.  However, since my mid-season roster changes the results had been a little more even albeit still in his favor.  Position by position, he still had a better position player at every position with the possible exception of shortstop (he had Jeter, I had ARod) and right field (Ichiro versus JD Drew).  Our starting staffs were pretty evenly matched, but it's difficult to tell because his team played in such an extreme pitcher's park.  However, there wasn't much doubt who had the more valuable closer (Otsuka versus Gagne).  Their numbers were surprisingly similar, except Gagne had thrown twice as many innings.  Nevertheless, I felt I had closed the gap enough that if fortune smiled on my team, they could make it an interesting series.  And that is exactly what happened: Morneau stayed hot, joined by Jose Guillen, JD Drew and ARod, my starting pitching kept every game winnable until the later innings and I got 14 scoreless playoff innings out of Gagne, Guardado, Smith and Graves to pull off the upset, 4-1, in the championship series.  My team's overall playoff record was 7-3, but maybe the key to winning it all was their road record of 4-2.

So what do I take from this?  Well, I don't fully understand how I developed a team that could win at everyone else's ballpark but my own.  That'll be a trick that's hard to repeat.  Previous simulation games I had played exposed me to the importance of a strong bullpen and that seems to have held true here.  And if it's possible to find cheap starting pitching in Strat-o-matic, it seems essential to have a terrific defense behind it.  But while it's satisfying to win, it's obvious that I still have a lot to learn about this game and the real one.  I'm also involved in a Strat-o-matic All-Time Greats league and using a similar formula - great team defense, strong bullpen - finished the regular season with the second best record in the league, both overall and on the road.  Unfortunately, the team with the best record in both regards is in my division.  We'll have to see how that ultimately turns out. 

Maybe legendary manager Whitey Herzog had it completely right when asked what it takes to be a successful manager.  His reply: "a sense of humor and a good bullpen".  He also said, "If you don't have outstanding relief pitching, you might as well piss on the fire and call out the dogs."  Sounds like Whitey might have played Strat-o-matic.