(FORTUNE Magazine) – It is a curiosity of the Long-Term Capital debacle that, although it has been endlessly and ably explained in the newspapers, it remains a mystery to many readers. Indeed, the affair has so many facets and is so complicated that it almost confounds complete understanding. For that reason, FORTUNE does not claim that this story will tell you all you need to know about Long-Term Capital or give more than a glimpse of all that is surely yet to unfold. But we do know we understand, to an extent that no other publication can, the role played in this affair by noted investor Warren Buffett, who is a longtime friend of this writer. Buffett's role in this saga was important, dramatic, and not without humor, coming to a climax in four particular September days. Moreover, his role is possibly not ended.

The outlines of what happened to Long-Term Capital in September are among the few clarities of this affair. For most of the month, the huge hedge fund hung on the edge of bankruptcy and was finally saved only by a $3.6 billion infusion from a consortium of 14 banks and brokerage firms--all creditors who feared what the bankruptcy of a fund that owed upwards of $100 billion would do to their own finances. Very much in the act also was the Federal Reserve Bank of New York, which wished to avert the domino effects that a fast liquidation of the fund's holdings might have exerted on an already reeling global securities market. So the New York Fed played godfather to the rescue deal, bringing the creditors together at its own offices and overseeing their negotiations.

In the center of this affair--always--was the hedge fund's founder and principal manager, former Salomon Brothers whiz John Meriwether. Meriwether had surrounded himself with brilliant colleagues, many of them Ph.D.s in mathematics and finance, and put them to work devising computer-driven trading strategies that were supposedly bulletproof. In the Greenwich, Conn., office of the hedge fund's management firm, Long-Term Capital Management, there may be more IQ points per square foot than in any other institution extant. There are certainly more Nobel Prize winners per square foot. LTCM has two: Myron Scholes and Robert Merton, who less than a year ago went jubilantly to Sweden to receive the world's highest accolade for achievement in economics.

The disaster that befell this brain trust recalls the first line of Allen Ginsberg's beatnik poem "Howl": "I saw the best minds of my generation destroyed by madness..." At LTCM the best minds were destroyed by the oldest and most famously addictive drug in finance, leverage. Had the fund not grievously overextended itself, it might still be trucking along, doing its thing, working those brain cells. But now its strategists have swapped their laurels for the booby prize of the financial markets, which is the ignominy of being largely wiped out and viewed as bumbling losers.

All of which raises the question of why Warren Buffett would want to pick up this particular piece of road kill--and yet he made a stab at doing exactly that. His move came on Wednesday, Sept. 23. It interrupted days of meetings that the New York Fed had been holding with the banks and brokerage firms that Long-Term Capital owed huge amounts of money to, which were being asked, in their own self-interest, to put up new money to stave off the fund's bankruptcy. The participants were an extraordinarily powerful group--every bank and firm that attended had sent its CEO or another top honcho--and the atmosphere was contentious. None of the members really wanted to ante up the money, and yet they dreaded a bankruptcy.

As the creditors' group was about to reconvene on this Wednesday morning, William McDonough, president of the New York Fed, learned from Goldman Sachs--itself one of Long-Term Capital's creditors--that it had an alternative financing proposition to present. McDonough recessed the meeting and listened to Goldman make the case for a bid, headed by Buffett's Berkshire Hathaway, to take over the fund. The terms were complicated--more on that later--but essentially the Buffett group proposed to put up $4 billion and make itself the manager of Long-Term Capital. Of the $4 billion, $3 billion was to come from Berkshire, $700 million from insurer American International Group, and $300 million from Goldman Sachs.

This bid then went to Meriwether, and that's when it began to die. There is disagreement as to why. Buffett believes that the deal did not happen because Meriwether and the other LTCM principals simply did not want to accept his terms, which would have left LTCM's principals with little money and no jobs. For their part, Meriwether and other LTCM partners told McDonough that the Buffett bid was structurally flawed and therefore not feasible.

Buying that argument, McDonough reconvened the meeting and told the creditors, in effect, that they were back to being the only game in town. One of the CEOs there told FORTUNE recently that McDonough explained that the other deal was not going to happen because it had "structural" problems. Grudgingly, the CEOs went back to negotiating their rescue plan, with each participant ultimately putting up very big money: $100 million to $350 million.

Suppose, FORTUNE asked that same CEO, that McDonough had said the deal would not happen because Long-Term Capital did not like its terms. "I think," answered the CEO, "that I would probably have told Meriwether to go screw himself." Of course, he adds, negotiations might then have turned into a game of chicken. Perhaps Meriwether would have countered that he preferred bankruptcy to Buffett. Then the creditors would be back to the question of whether they could stand bankruptcy. Squawk, squawk, which side flinches first?

To understand how both Long-Term Capital and the creditors--and the Fed, to boot--got themselves into this amazing bind and why Buffett volunteered to step into apparent chaos, it is necessary to look at the character of this hedge fund and its boss, John Meriwether. Until August 1991, Meriwether, now 51, was a vice chairman of Salomon, with responsibility for fixed-income trading and also "proprietary trading," which is a firm's investment of its own money. Meriwether was a huge money-maker for the firm, and was liked and admired to the point of reverence by the people who worked for him. "British Guiana actually comes to mind," said a former Salomon Brothers executive recently. "John had a kind of cultlike following."

Then came Salomon's Treasury bond scandal, precipitated by Paul Mozer, one of Meriwether's senior people. In the chain of disclosure, Mozer admitted to Meriwether in April 1991 that he had falsified bids for Treasury securities; Meriwether immediately went to his bosses, John Gutfreund and Thomas Strauss, with the news; everybody concluded that the Fed must be told; and then nobody did the telling. When the scandal finally erupted in August, Mozer was fired, and Gutfreund and Strauss resigned under pressure. To the rescue came Buffett, head of Salomon's biggest shareholder, Berkshire.

In the first two days of Buffett's regime as chairman of Salomon, Meriwether's fate was uncertain. Many of Salomon's managing partners had turned against him and wanted him out. But Buffett was not sure that it was fair to fire Meriwether, since he had acted responsibly in immediately disclosing Mozer's sins to the firm's bosses. In the end, Meriwether himself made the decision to resign, telling Buffett that he thought it was the best thing for the firm.

Gradually, a number of men who had worked for Meriwether left Salomon to join him in business. The specifics about what they were to do surfaced in 1993, when Meriwether announced plans to start a hedge fund called Long-Term Capital. Actually, Long-Term Capital is made up of a variety of investment vehicles, some partnerships, some corporations; some domiciled in the U.S., some in the Cayman Islands. Why the Caymans? So that offshore investors, as well as certain kinds of U.S. investors, can avoid U.S. income taxes--a point that grates, of course, when we recall that the Fed was scrambling around in September on the fund's behalf. However, all of these instruments feed their money into a master fund called Long-Term Capital Portfolio L.P., a Cayman partnership.

It is this fund that has taken the positions that its investment managers, Meriwether and crew, dictated. From its beginning, the fund was distinctive in that its managers proposed to carry out trading strategies that would take time--six months to two years or more--to deliver profits. For that reason, LTCM ruled that it would not allow investors in the fund to withdraw their money quarterly or annually, which are the common practices of hedge funds, but would instead lock up their money until the end of 1997.

As for the strategies, the fund's managers proposed to make relatively few trades that carried "directional" risk and instead to concentrate on capturing small profits from carefully hedged positions, particularly in the fixed-income markets. A couple of illustrations will illuminate some distinctions in risk, which are important in this tale. Imagine that an individual investor decided to buy the stock of Cendant, on the theory that it was headed up. That would be a directional trade, or a "position risk." Assuming the investor didn't buy on margin, the percentage gain or loss on each dollar at risk would be exactly equal to the percentage rise or fall in Cendant stock.

Long-Term Capital's main strategy, in contrast, was to make hedged trades that it thought could succeed regardless of the general trend of the markets. The fund put its financial technology and brains to work, for example, at identifying sectors of the bond market in which yields had gotten out of line with yields in related sectors. It would then buy one of the securities and short the other. A trade of this type can deliver a profit regardless of whether interest rates in general go up or down or stay flat. All that matters is that the two yields eventually converge.

In this kind of trading, the fund never expected to earn a large return on each dollar at risk. An investor in Long-Term Capital recalls talking by phone to some of its top managers in 1996 and asking them just how much they were annually realizing per dollar. The answer was 67 basis points, or 0.67 of a cent.

Returns like that were never going to make investors happy unless the invested capital could be greatly enlarged, and that's where the leverage came in. Confident that its position risk was very low, the fund took on a colossal amount of what's called "balance-sheet risk," piling a huge slab of debt on top of its relative sliver of capital. At the moment in 1996 when that investor was getting his answer of 67 basis points, the fund had a ratio of $30 in balance-sheet debt for every $1 in capital. Leveraged 30 to one, a 0.67% return on each dollar at risk produces a healthy 20% return on capital.

Leverage that steep sounds terrifying--and eventually it was. But LTCM always maintained that its financial technologies and its meticulously constructed hedges gave the fund a conservative risk profile. In October 1994, LTCM spelled out this proposition in a paper sent to its investors. In the paper was a table that delineated a range of returns that the fund might aim for in a year and paired them with the probabilities of loss if things went bad. For example, the table said that if the fund were shooting for a 25% return--which was indeed a typical goal for the operation--the probability that it actually would end up losing 20% or more was an insignificant 1 in 100. The table never even contemplated a steeper loss.

That would all be funny if it weren't tragic. In the first eight months of 1998, the fund lost nearly 50% of its capital, and in September it hemorrhaged still more, with the loss going to around 90%. "What that tells you," says Warren Buffett, "is that underneath the mathematical elegance--underneath all those betas and sigmas--there was quicksand."

At least one of Long-Term Capital's investors was troubled by the same thought. The investor who was told about the fund's 67-basis-point return wrote himself two notes for the file after that conversation. He recently showed them to FORTUNE. One note said, "Are they no different from any Wall Street firm--Bear or Salomon or Goldman? It's the proprietary desks without the people, overhead, or agency business."

The second note leaned on a Yiddish word, kishka, which figuratively means "gut instincts." Said the note: "There is no kishka override over the portfolio. One gets a Nobel Prize-winning computer system. Will that collapse one day? Would one prefer a person, with kishkas, at the controls?"

From the looks of the fund's vitality in its first four years, from the late winter of 1994 until the spring of 1998, not many investors were worrying about kishkas. The structure of Long-Term Capital calls for its investors to pay the management company unusually steep fees: first, an annual 2% on the capital that an investor has in the fund, and second, 25% of profits earned. Even so, the fund's investors earned 20% in the ten months of 1994 that it was in operation, 43% in 1995, 41% in 1996, and 17% in 1997. Furthermore, monthly performance figures showed that the fund was having a smooth ride, experiencing very little volatility in its results.

Like a siren song, these facts brought in a flood of new money. That was true even though the fund's managers kept a tight veil of secrecy over their specific trading strategies, leaving their investors with little clue as to just how the fund was making its profits. By the fall of 1997, the combination of contributions and reinvested earnings had raised the capital of the fund to around $7 billion.

At that point, things took an unexpected turn: LTCM itself concluded that the fund had too much capital for the investment opportunities open to it and forced many of its investors, notably those who had come in late, to withdraw their money. Many of those expelled went kicking and screaming, and at least one protested so angrily that LTCM allowed him to stay onboard. The anger arose in part because LTCM's principals and other insiders were allowed to maintain their full stakes in the fund. In other words, maybe attractive investments weren't so scarce after all. When the smoke had cleared at year-end 1997, the fund had $4.7 billion in capital, of which about $1.5 billion belonged to insiders.

At that time, the fund also had about $125 billion in debt, which means that its balance-sheet leverage ratio was about 25 to one. But that far understates the true exposure, since the fund has always aggressively entered into off-balance-sheet derivatives contracts that by their nature create additional leverage. Only an insider would know, at any point, what the derivatives exposure amounted to--but be assured that if the balance-sheet leverage at year-end 1997 was 25 to one, the overall leverage was greater.

In a sense, all this debt owed its existence to another kind of near-cult following that Meriwether had acquired, this one made up of banks and brokerage firms enamored with the prospect of lending to him. Creditors love customers who are big, active, and solid, and that was precisely the image that Long-Term Capital projected. Most of the fund's creditors do not appear to have looked critically at its leverage. In fact, many were accommodating in granting favorable terms on loans--the waiving of collateral, for example. Many also happily entered into derivatives contracts with the fund, increasing their total exposure to what, unknown to them, was about to become a train wreck.

This year of the wreck, 1998, began calmly enough, but in May, volatile markets cost the fund a 6% loss, and in June, another 10%. Those lashes of the whip, unheard of in the fund's history, were enough to make Meriwether shoot off a special letter to his investors. It said that LTCM was "understandably disappointed" with the May and June results, but nevertheless believed that "the future expected returns [for the fund's investment strategies] are good." In July the fund was essentially flat.

And then came August--terrible, terrible August. Russia devalued its currency, and the world's investors fled to quality. That was not what LTCM had been banking on in several different trading strategies. For example, it had at some point done paired trades that in essence bet on a narrowing of the spread between the yields of AA-rated corporate bonds and comparable U.S. Treasuries. The bet looked good because the spread was relatively wide compared with its historical levels. But when the flight to quality began, Treasury yields plunged, and the gap between government and corporate yields dramatically and incredibly widened still more. The outcome was disastrous. The fund lost 40% of its capital in August alone, leaving it with just $2.5 billion. Astoundingly, it was still carrying about $100 billion in debt.

Around this time, the fund's principals and at least some creditors got really nervous. Stuffing their pride into the closet for the moment, the LTCM crew went on a search for investors who might buy out some of their trading positions or, alternatively, put fresh capital into the fund, thereby giving it a stronger base. The team solicited, among others, certain investors who'd been pushed out of the firm at the end of 1997, as well as the hedge fund moguls George Soros and Julian Robertson. All said no.

On the night of Sunday, Aug. 23, LTCM opened the Omaha front. The man assigned to call Buffett was Eric Rosenfeld, an LTCM principal and onetime Salomon trader who had earned Buffett's regard by loyally helping Salomon climb out of its hole after the 1991 crisis. Rosenfeld, 45, is normally laid back. But on this occasion, Buffett says there was a sense of urgency in Rosenfeld's offer to sell him some of the fund's large equity arbitrage positions. Buffett said no.

By Wednesday, Rosenfeld was back on the phone, joined this time by Meriwether. Could Buffett meet the following morning in Omaha with still another LTCM principal, Lawrence Hilibrand? At the meeting, Hilibrand, 39, also a former Salomon trader well regarded by Buffett, presented a rough, though still guarded, picture of Long-Term Capital's portfolio and urged Buffett to become a large investor in the fund. Hilibrand also pinpointed a reason for hurry: When August ended a few days later, the fund was going to have to tell its investors and lenders how much it had lost for the month and would be pleased to tell them also that it had secured financing arrangements that would ease the fund's leverage. Sorry, said Buffett politely, but he simply had no interest.

Later, Buffett talked to Berkshire's vice chairman, Charles Munger, and remembers telling him that though the fund's positions might have been logical, he had no desire to make Berkshire a hedge fund investor. He also pointed out the absurdity of "ten or 15 guys with an average IQ of maybe 170 getting themselves into a position where they can lose all their money."

After that, Buffett did not again think much about Long-Term Capital until late Friday afternoon, Sept. 18, when he returned a call from a Goldman Sachs partner he knew, Peter Kraus. Kraus said that Long-Term Capital's net assets--its capital--had sunk to $1.5 billion, and that Goldman was calling around looking for people who might want to become big investors in the fund. No interest, Buffett said once again. But the two men then bounced around other ideas relating to Long-Term Capital. Finally, Buffett got to focusing on the very-big-picture thought that Berkshire and Goldman might jointly bid for the entire fund, chuck the Meriwether management, and take on the fund's management themselves. The concept, says Buffett today, was to patiently stick with most of the fund's positions and gradually liquidate at what Buffett thought could be a decent profit.

It was a major idea, of a size suitable for Berkshire's billions in available cash, but it could not have come at a worse time. Buffett was trying to leave his office to get to a granddaughter's birthday party. That night he was scheduled to fly to Seattle to join a Bill Gates group that was going to spend nearly two weeks touring Alaska and a galaxy of Western parks. It was a highly uncharacteristic move; Buffett ordinarily has no interest in scenery of any kind. But his close friendship with Gates had lured him and his wife, Susan, into signing on for the trip.

If a bid was to be made, then Buffett would have to work it out with Kraus over the phone from the wild. And that's how it happened, although not without moments of exquisite frustration. Suffice it to say that not even a megabillionaire can establish and keep a phone connection from the canyonlike depths of an Alaskan fjord.

Nevertheless, the $4 billion bid took shape over four hectic days, with Goldman declining to be more than a $300 million partner and with AIG, at Buffett's suggestion, asked in. That invitation occurred because Buffett has had good relations with AIG's chairman, Maurice "Hank" Greenberg, and thought this gesture might further cement them. Another facet of the deal called for Goldman to be the fund's administrator, once the property was in the bidding group's hands.

It is essential to understand what the group proposed to buy and at what price. It wanted to own the fund's entire portfolio, and to accomplish that, it wished to purchase the master fund, Long-Term Capital Portfolio L.P. This is the fund that had $1.5 billion of capital on Friday, Sept. 18. By the following Wednesday, though, when the Buffett bid was made, the fund's net assets were maybe in the range of $600 million. Nobody knew for sure. In any case, Buffett, never a man to overpay, wanted to buy the fund at a discount, and his bid was $250 million. Had it been accepted, the sum would have gone to the fund's old investors--Meriwether's management team included--and would have tacitly said that a fund worth $4.7 billion at the start of 1998 had sunk in value to about a 20th of that. Once the $250 million was paid over, the Buffett group would have immediately put up an additional $3.75 billion, which would have restored capital adequacy to the fund.

Think a bit more about the $250 million. Once it went on the table, it would become the figure that Meriwether, given time, would quite understandably try to "shop," looking for more dollars. Also, when you are talking a $4 billion proposition and the markets are turbulent, it is not good to let a bid sit around and go stale. So Buffett put a tight time limit on his offer. The bid was faxed to Meriwether at 11:40 on Wednesday morning, and he was advised it would expire within an hour, at 12:30.

Meanwhile, McDonough and the creditors' group had already been trying for days to assemble its rescue plan and were meeting again that day. So even before the bid was faxed to Meriwether that morning, Goldman briefed McDonough on the deal, and McDonough called Buffett, now at a Montana ranch, to confirm the offer. Buffett says that McDonough seemed elated at the sudden appearance of an alternative plan to the one he'd been nursing with the creditors. Buffett then called Meriwether and told him a bid was on its way. Buffett made it clear that his group wanted to buy Long-Term Capital Portfolio in its entirety. Meriwether was noncommittal.

But when the call ended, Meriwether went into action, fast. After a quick conversation with LTCM's in-house lawyer, he got on the phone to McDonough and told him that Buffett's bid wouldn't work because Meriwether did not have the authority to sell. Meriwether is a general partner of the so-called portfolio company, but it also has other partners, and Meriwether contended he would have to seek the vote of each one before agreeing to sell. And obviously, he said, that could not be done before 12:30. Backing Meriwether up in the conversation was another LTCM principal and a former colleague of McDonough's, David Mullins, who was vice chairman of the Federal Reserve Board until he joined LTCM in 1994.

Buffett's group had not gone into its bid without its own legal opinion. Studying the Long-Term Capital documents available, incomplete though they were, the group's lawyers concluded that the bid was workable from Meriwether's side. Besides, everyone on the Buffett side believed that if Meriwether really had any interest in the bid, he could tentatively accept it and then tangle with the complications later.

But Meriwether displayed absolutely no interest when he spoke to McDonough. The Fed president concluded that he could not dismiss Meriwether's legal argument or, for that matter, stretch his authority as a central banker to force Meriwether to sell. Resignedly, McDonough returned to his surly band of creditors and flogged along their $3.6 billion rescue plan.

How do Meriwether and the fund's other investors fare under that scheme? Part of the answer is that they keep whatever shrunken capital they still have in the fund, which totaled something around $400 million in late September. That is the correct figure to compare with Buffett's $250 million, though it is worth mentioning that Buffett was to hand over cash, whereas the $400 million remains tied up in the fund.

As for Meriwether and the LTCM crew, they still run the fund, though the creditors have roughly halved their fees. The creditors have also put six of their people in LTCM's offices with instructions to liquidate the fund as fast as it can be sensibly done. Finishing within a year would be splendid, says one senior Wall Streeter.

There is, of course, another scenario: a prompt move by the creditors to find a bidder--a Buffett, say--who would step up and take this mess off their hands immediately. Would Buffett do that? Who knows, since he never telegraphs his moves. But he is aware that certain creditors have been thinking about him.

In any event, Buffett has returned from vacation, and he managed to catch C-Span's Oct. 1 coverage of the House hearings on Long-Term Capital. The vignette he liked best: Congressman James Leach's memory that his father had told him to avoid anyone doing business out of the Cayman Islands. That reminded Buffett of a favorite line that seems just about the perfect wrapup for Long-Term Capital: "You never know who's swimming naked until the tide goes out."

REPORTER ASSOCIATES Patty de Llosa, Carol Vinzant