The biggest single loser in the LTCM debacle was UBS, which was forced to write off Sfr950 million ($682 million) of its exposure. The UBS involvement with LTCM pre-dated the merger of Union Bank of Switzerland and Swiss Bank Corporation in December 1998. Various heads rolled, including that of chairman Mathis Cabiallavetta (formerly chief executive of Union Bank of Switzerland), Werner Bonadurer, chief operating officer, Felix Fischer, chief risk officer, and Andy Siciliano, head of fixed income (who had been with SBC).

UBS's deal with LTCM was a variation on other attempts to turn hedge funds into a securitized asset class with a protected downside. However in this case UBS was protecting the downside and LTCM was taking a good deal of the upside. The sweetener for UBS was a structure that looked more like an option than a loan, turning any income into a capital gain, and an opportunity to invest directly in LTCM.

For a premium of $300 million UBS sold LTCM a seven-year European call option on 1 million of LTCM's own shares, valued then at $800 million. To hedge the position - the only way it could be done - UBS bought $800 million worth of LTCM shares. UBS also invested $300 million (most of the $266 million premium income) directly in LTCM. Such an investment had to be held for a minimum of three years.

This transaction was completed in three tranches in June, August and October 1997.

The deal was calculated so that the $300 million premium was equivalent to a coupon of Libor plus 50 basis points over the seven years.

Assuming that LTCM performed well the deal provided UBS with steady, tax-efficient, return plus a share in the upside, through its $266 million stake.

But if ever its hedge looked like falling below the $800 million strike price it was looking at a loss. The only way to hedge it would have been to sell LTCM shares.

But there were various impediments to this. UBS could not just dump the shares. It was obliged to convert any shares it sold into a loan at par value, maturing in 2004.

Shares in hedge funds aren't liquid, and LTCM's were no exception. It was impossible to mark them regularly to market. LTCM reported to shareholders only monthly. If UBS did sell LTCM shares in a falling market, and then LTCM's performance picked up again, there was no guarantee it could rehedge its position. No one was making a market in LTCM shares.

Theoretically there was a volatility cap on the arrangement: if the fund's volatility exceeded a certain level a cash sum would be reckoned in UBS's favour, payable at the end of year seven. But it is not clear how that would have left UBS market-neutral.

In the climate of mid-1997 it is understandable how UBS risk managers might have overlooked the horrible implications of a worst-case LTCM scenario. LTCM had a fantastic reputation for big-number but low-risk arbitrage. (There is a parallel in the reputation that Nick Leeson enjoyed at Barings before March 1995).

But it is clear now that UBS risk managers never faced the possibility of a collapse of LTCM which would have left them with $766 million exposure ($800 million hedge, $266 million investment, less $300 million option premium). That is, they didn't wake up to it, apparently, until around April 1998, in a post-merger review, when it was too late to do much about it.

Credit Suisse Financial Products, which did a similar deal for $100 million, set that as the maximum it was prepared to lose.

An interesting aspect of the UBS deal is to consider it from LTCM's point of view. LTCM secured $800 million new investment capital at Libor plus 50 basis points. It had a call on all returns above that level. UBS's obligation, to convert any shares it wanted to sell into a loan, provided LTCM with a synthetic seven-year put on its own performance. Was this an added incentive to roll the dice? It was a cheap gambling stake.