So much of by-the-book baseball management has been handed down from the game’s earliest days. Proper lineup construction was puzzled out in the 19th century.The pastime’s thinkers figured out that your agile Honus Wagners play shortstop and your lumbering Dan Brouthers stand around first, sort of like Jason Giambi does now.

Right or wrong, they also decided that the fast guys lead off, the hit-and-run guy bats second, the slugger bats third, and they handed that information off to succeeding generations. Eventually this was refined to the point that Casey Stengel could simplify the wisdom of the ages to, “You put a right-hand hitter against a left-hand pitcher and a left-hand hitter against a righthand pitcher and on cloudy days you use a fastball pitcher.”

Unfortunately, the ancients were obsessed with primitive manhood rites like riding one pitcher until his arm disintegrated, so they didn’t give relievers much attention. They have nothing to say to us on that subject, so even the Joe Morgans of the world, who cling to deadball-era wisdom the way Antonin Scalia clings to the intent of the framers, can’t point to the way it was done by John McGraw or Sparky Anderson. The magic formula for putting together a good bullpen is still the subject of anxious fumbling, leaving the modern general manager hoisting his lantern in the darkness like a horsehide Diogenes looking for one good reliever.

There is much about bullpen construction that seems almost random; relievers are a highly inconsistent lot. Comparing a list of the top 50 relievers from 2003 to that of 2004 would reveal that less than half of the pitchers carried over their success from one year to the next. Comparing the 2003 list to 2005 would show perhaps 10 pitchers ranked three years in a row; 80% of the list would be new.

That’s why so much of bullpen construction is guesswork. In baseball, a player is acquired for what he did last year.If a first baseman, 29 years old, has a five-year average of .280 AVG/.360 OBA/.510 SLG, a team can sign him with the reasonable expectation that he’s going to produce similar numbers for them. If the team is unlucky he might hit .270/.355/.490, and if it’s lucky he’ll hit .295/.380/.540, but barring illness or injury, it’s unlikely that he won’t hit at all.

Starting pitchers are less reliable, but because they put in 200 innings a year, their results are less open to variation. A reliever throws 50 to 90 innings each season, a small enough sample that it’s possible a reliever could go through an entire season being lucky or unlucky. Last year, Todd Jones was terrific for the Marlins, so the Tigers signed him to be their closer. At this writing, his ERA is just under five runs higher than it was last season. Jose Valverde was a 25-year-old closer on the rise for the Diamondbacks. This year he’s a 26-year-old millstone with an ERA higher than the national debt. These developments are not atypical.

Entering this season, the two New York clubs took divergent approaches to staffing their bullpens. As is their wont, the Yankees spent heavily, bringing in Kyle Farnsworth ($17 million), Mike Myers ($2.4 million), Ron Villone ($2 million), the rehabbing Octavio Dotel ($2 million plus more if he actually pitches), and picked up the option on flash-in-thepan Tanyon Sturtze ($1.5 million).

The Mets put their eggs in Billy Wagner’s $43 million basket and trusted providence to provide the rest of the staff. Duaner Sanchez is making less than $400,000. Pedro Feliciano is no millionaire. Chad Bradford went from celebrated “Moneyball” find to the discard pile in just two years; he’s making $1.5 million.Darren Oliver didn’t pitch in the majors in 2005. He played for four teams in his previous four years and was miserable with all of them. He’s making $600,000. Aaron Heilman, still lacking the service time for arbitration, is making close to the league minimum.

Wagner aside, most of these pitchers are not of the designer brand variety. Sanchez, who leads the pen in innings pitched as Wagner’s set-up man, is a great example. No silver-spoon Huston Street type, Sanchez entered baseball as a 16-year-old undrafted free agent with the Diamondbacks in 1996. An unexceptional minor leaguer, the Snakes traded him to the Pirates for a washedup Mike Fetters in 2002. After being bombed in two cup-of-coffee call-ups (Sanchez’s Pirates ERA is 16.20 in 8.1 innings), he was released. From there he found his way to the Dodgers, where he finally established himself. The Mets acquired him last winter for Jae Seo, a journeyman pitcher coming off of an unlikely good year.

The Mets’ no-names have been more successful than the Yankees’ mercenaries, allowing 3.84 runs per nine innings to the Bombers’ 4.35. In this, the Yankees have paid for the belated realization that even with Mariano Rivera and their fistful of expensive retainers, the pen was bleeding around the edges. Prior to their departure,Aaron Small (sent to minors), Tanyon Sturtze (rotator cuff surgery), and Scott Erickson (released yesterday) combined to allow 36 runs in 37.2 innings (bullpen assemblage may be random, but some bets have a better chance of paying off than others). The remaining staff has allowed 3.40 runs per game.

The results could just as easily have been reversed or totally different. The Braves will be the first to testify that collecting unknown relievers has just as good a chance of blowing up on a team as propelling it to a pennant.Their cadre of rookies and rag-arm vets has allowed 5.50 runs per game and turns over leads faster than the phone company will turn over your call list to the NSA.

On Sunday the Yankees lost a frustrating game to the Washington Nationals because their burned-out bullpen lacked anyone that Joe Torre felt good about using in the ninth inning. It wasn’t his fault, or even GM Brian Cashman’s; having such a pitcher available is more art than science.

Mr. Goldman writes the Pinstriped Bible forwww.yesnetwork.comand is the author of “Forging Genius,” a biography of Casey Stengel.