Commodities Hedge Fund Blowups Liquidity Natural Gas Position Sizing Risk Management Scenario Analysis

The Amaranth Blowup: What a $6.6 Billion Loss Teaches Systematic Traders About Position Sizing

Summary

Amaranth Advisors lost $6.6 billion in natural gas spread trades in September 2006 after holding positions equal to 40-81% of NYMEX open interest. A simple scenario analysis shows that a reversion to historical spread levels would have implied up to a 36% portfolio loss, visible two weeks before the collapse.

· 10 min read
Original paper The Amaranth Case Study
Hilary Till
Global Commodities Applied Research Digest 2018 DOI: https://www.jpmcc-gcard.com/wp-content/uploads/2018/05/GCARD_Summer_2018_CEC_Till_Amaranth.pdf ↗
Key findings
  • Amaranth held up to 40% of all NYMEX open interest in winter natural gas months and 81% of open interest in the December 2007 contract as of July 24, 2006.
  • A scenario analysis using six years of historical spread data showed the August 31, 2006 portfolio could have lost up to 36% under normal conditions, two weeks before the actual collapse.
  • Two spreads (November vs. October 2006 and March vs. April 2007) were 93% correlated with Amaranth's entire natural gas book.
  • The fund's January 2007 position exceeded 80,000 contracts, representing a volume of natural gas equal to the entire U.S. residential consumption for that month.
  • Citadel acquired Amaranth's portfolio for a net payment of $1.425 billion and unwound the positions over approximately six weeks with orderly execution.
  • Amaranth ultimately settled regulatory and class action claims for a combined $85.25 million plus a $750,000 individual settlement by its head trader.

In September 2006, Amaranth Advisors lost $6.6 billion in roughly two weeks. The fund had $9.2 billion in assets under management at the end of August. By mid-September it was effectively finished. The loss came from natural gas spread positions, specifically long winter contracts against short non-winter contracts, spanning delivery dates from 2006 through 2010. Hilary Till's case study, published in the J.P. Morgan Center for Commodities' Global Commodities Applied Research Digest (Summer 2018), reconstructs the positions from the U.S. Senate's 2007 investigation report and runs the numbers that should have prevented the trade from reaching those sizes in the first place.

What Amaranth Was Actually Trading

Amaranth was not speculating on the outright price of natural gas. The fund was trading calendar spreads: long winter delivery months (when heating demand drives prices up) against short spring and summer months (when demand is lower). These positions would profit if the gap between winter and non-winter prices widened, which could happen from hurricanes disrupting production, colder-than-expected weather, or storage constraints. The economic logic was reasonable. Natural gas cannot be shipped globally as easily as oil, storage capacity is physically finite, and weather events create genuine supply asymmetries.

The problem was not the thesis. The problem was scale. The fund's head energy trader held open positions worth tens of billions of dollars in commodities. By the end of August 2006, the fund's energy trades accounted for about half of its capital and had generated roughly 75% of its profits. When those spreads moved against the fund in September, there was no exit.

How Large the Positions Actually Were

The U.S. Senate Permanent Subcommittee on Investigations released a 135-page report with 345 pages of appendices in June 2007. The data came from subpoenaed records covering several million individual trades from NYMEX, ICE, Amaranth, and other market participants. The documented scale is difficult to overstate:

  • Amaranth controlled up to 40% of all NYMEX open interest in winter months (October 2006 through March 2007).
  • In late July 2006, the fund's January 2007 position exceeded 80,000 NYMEX and ICE contracts, representing a volume of natural gas equal to the entire amount used by U.S. residential consumers in that month.
  • On July 31, 2006, Amaranth's trading in the March and April 2007 contracts represented almost 70% of the total NYMEX trading volume in those contracts on that date.
  • The fund held 60% of all outstanding NYMEX natural gas futures contracts for delivery in 2010.
  • On July 24, 2006, Amaranth's position as a percentage of NYMEX futures open interest in the December 2007 contract was 81%.
  • On August 28, 2006, Amaranth accounted for over 40% of total ICE volume and over 25% of the entire combined exchange-traded volume on NYMEX and ICE.

For a financial entity with no physical natural gas assets, no pipelines, no storage facilities, and no ability to make or take physical delivery, these position sizes meant the fund's only exit was to find a financial counterparty willing to take the other side. When the spreads moved against them, that counterparty did not exist at any reasonable price.

The Scenario Analysis That Should Have Stopped It

Till's most useful contribution is a straightforward scenario analysis. Using the Senate report's documented positions as of August 31, 2006 (two weeks before the implosion), she identifies that two spreads were 93% correlated with Amaranth's natural gas book: the November 2006 vs. October 2006 spread and the March 2007 vs. April 2007 spread. She then asks a simple question: what would happen to the portfolio if those spreads reverted to levels that had prevailed at the end of August during any of the previous six years?

The answer: up to -36% could have been lost under normal conditions, with no hurricane, no crisis, and no unusual market event required. Just a reversion to historical norms. A 36% loss on a $9.2 billion portfolio is $3.3 billion, which would have been catastrophic enough on its own. The actual loss was nearly twice that because it triggered a liquidation spiral.

This is the core lesson of the case for anyone building systematic strategies. The scenario analysis was not complex. It did not require advanced risk models or Monte Carlo simulation. It required looking at where the same spreads had traded in prior years and asking what the portfolio would be worth if they went back there. Any systematic trader can run this kind of analysis on their own positions.

The Liquidation Spiral

By mid-September 2006, the fund had lost more than $2 billion month-to-date. On Thursday, September 14, the fund lost $560 million in a single day. The following weekend, Amaranth scrambled to transfer positions to third parties. Merrill Lynch agreed to take 25% of the natural gas positions for a payment of about $250 million.

The fund then lost another $800 million through Tuesday, September 19. On Wednesday, September 20, Amaranth transferred its remaining energy positions to Citadel Investment Group and its clearing broker J.P. Morgan Chase at a $2.15 billion discount to the previous day's mark-to-market value. The two firms shared the risk equally. On Thursday, September 21, the natural gas curve stabilized.

The critical-liquidation-cycle is a vicious circle: distressed liquidations cause even greater losses, which trigger greater margin calls from credit providers, which force even more distressed trading. The natural counterparties to Amaranth's trades were physical market participants, pipeline companies, storage managers, and retail marketers, who had locked in the value of forward production or storage. They had physical assets backing their derivatives positions and no economic need to unwind their trades on Amaranth's timeline.

As Robert Greer of PIMCO observed: "The market showed that someone can actually be so big that the market will punish them, rather than reward them for their size."

How Citadel Unwound the Positions

The aftermath provides a textbook example of how orderly position reduction works versus distressed liquidation. Citadel took over the entire Amaranth portfolio by the end of September, paying J.P. Morgan $725 million for J.P. Morgan's half of the positions. Citadel received all remaining concessionary payments from Amaranth, netting $1.425 billion in total payments for agreeing to take on the distressed portfolio.

Till tracks the natural gas spread evolution from September through December 2006. The spreads briefly recovered in late September (indicating a temporary absence of liquidation pressure), then smoothly declined throughout October and largely bottomed by the end of that month. Citadel's bond prospectus later confirmed the firm had reduced the risk of its Amaranth positions by two-thirds during the first two weeks of October and essentially finished during the last two weeks. Commercial market hedgers were apparently the natural other side to Citadel's orderly unwind, as they could realize their substantial hedging windfall at this time.

By November and December 2006, the natural gas curve was stable, indicating normal two-sided flow had resumed. Given how orderly the unwind was, Citadel likely sustained only relatively small losses during the process, meaning the firm probably realized substantial net profits from the $1.425 billion payment for absorbing the distressed positions.

The difference: Amaranth tried to exit a position equal to the entire U.S. residential gas consumption in a matter of days. Citadel took six weeks. Time and orderly execution turned a catastrophic position into a profitable trade.

What Systematic Traders Can Take From This

The Amaranth case is not about commodities or natural gas or hedge funds. It is about what happens when position sizes exceed the capacity of the market to absorb an exit. This applies to any systematic strategy at any scale:

  • Position sizing relative to market liquidity is a risk parameter, not an outcome. Amaranth's positions grew because they were profitable, and no one applied a hard constraint based on what percentage of daily volume or open interest the fund represented. Any systematic strategy should have a rule that prevents positions from exceeding a defined fraction of average daily volume.
  • Scenario analysis beats VaR for concentrated positions. The -36% loss was visible from a simple historical reversion check. Value-at-Risk models based on normal distributions and short lookback windows would have underestimated the risk because these models do not handle concentration or illiquidity. In RealTest, running a strategy across multiple market regimes and measuring the worst-case drawdown across regime transitions is the equivalent of Till's scenario analysis for equity strategies.
  • Exit strategy must be part of entry criteria. A financial entity that cannot make or take physical delivery has exactly one exit: find someone to take the other side. If your position is 81% of open interest in a contract, your exit depends entirely on whether a counterparty exists. Systematic equity traders face the same constraint in small-cap stocks where daily volume drops below a threshold.
  • The profitable side of a crowded trade is not free money. Amaranth's counterparties were physical market participants who had locked in favorable hedging rates. When Amaranth collapsed, those participants had no urgency to help. The lesson: if you are winning on a trade and cannot identify who is losing, the crowded side may be more fragile than it looks.

SetupAlpha strategies are built in RealTest with explicit position sizing constraints tied to average daily volume and maximum portfolio allocation per position. The RealTest Mean Reversion Trading Strategy for 2025 and the RealTest Short Term Mean Reversion Strategy include volume filters directly in entry logic, ensuring that positions cannot be entered in stocks where the strategy's capital would represent a disproportionate share of daily trading activity.

The Regulatory Aftermath

Both the CFTC and the Federal Energy Regulatory Commission (FERC) pursued actions against Amaranth and its head trader, announced in July 2007. The investigations focused narrowly on specific trading days in 2006, leading to allegations of attempted and actual manipulation. In August 2009, Amaranth agreed to pay $7.5 million to settle FERC and CFTC cases. In April 2013, a federal judge approved a $77.1 million class action settlement by Amaranth to resolve claims brought by other traders.

The jurisdictional question itself became a case study. A March 2013 federal court ruling found that FERC lacked jurisdiction to charge Amaranth's head trader with manipulation of natural gas futures contracts, ruling that the Commodity Exchange Act gives the CFTC exclusive jurisdiction. The trader ultimately settled with the CFTC in September 2014 for $750,000 and a partial ban on futures trading, neither admitting nor denying wrongdoing.

Limitations

Till's scenario analysis is based solely on positions documented in the U.S. Senate report's graphical appendix. Amaranth may have held additional positions not captured in those charts, which would change the sensitivity analysis. The report also excludes positions past the May 2009 maturity date and miscellaneous commodity investments.

The case describes a single event in a single commodity market with specific structural features (no global fungibility, finite physical storage, seasonal demand patterns). The liquidity and concentration dynamics apply broadly, but the specific spread behavior and counterparty structure are particular to natural gas.

The article does not quantify what risk management framework would have been sufficient to prevent the loss, only that a simple scenario analysis would have flagged the risk. The gap between identifying a risk and acting on it is organizational and behavioral, not analytical.

Citation: Till, H. (2018). The Amaranth Case Study. Global Commodities Applied Research Digest, Contributing Editor's Collection, Summer 2018, pp. 82-89. J.P. Morgan Center for Commodities, University of Colorado Denver Business School. https://www.jpmcc-gcard.com/wp-content/uploads/2018/05/GCARD_Summer_2018_CEC_Till_Amaranth.pdf

Key terms

Calendar Spread
A trade involving simultaneous long and short positions in futures contracts of the same commodity but different delivery months. Amaranth was long winter months and short non-winter months, betting on the seasonal price gap widening.
Open Interest
The total number of outstanding futures or options contracts that have not been settled. Amaranth held up to 81% of open interest in certain natural gas contracts, meaning the fund was effectively the market in those instruments.
Liquidation Spiral
A feedback loop where forced selling moves prices against a distressed position, triggering additional margin calls and further forced selling, each round at worse prices than the last.
Scenario Analysis
A risk assessment method that calculates portfolio impact under specific, defined conditions rather than statistical probability distributions. Till's analysis asked what would happen if spreads returned to any prior year's end-of-August level.
Concessionary Payment
A payment made by a distressed seller to a buyer as compensation for taking on unwanted risk. Amaranth paid $2.15 billion to Citadel and J.P. Morgan to absorb its energy positions.
Value at Risk (VaR)
A statistical measure that estimates the maximum expected loss over a defined time period at a given confidence level. VaR models typically assume normal distributions and underestimate tail risk for concentrated, illiquid positions.
Mark-to-Market
The process of revaluing a position based on current market prices. Amaranth transferred its positions to Citadel and J.P. Morgan at a $2.15 billion discount to the September 19, 2006 mark-to-market value.

Frequently asked questions

What was Amaranth Advisors?

Amaranth was a multi-strategy hedge fund founded in 2000 and headquartered in Greenwich, Connecticut. Originally focused on convertible bonds, it expanded into merger arbitrage, long-short equity, leveraged loans, and energy trading. By mid-2006, energy trades accounted for about half of its $9.2 billion in assets and 75% of its profits.

What caused Amaranth's $6.6 billion loss?

The loss came from natural gas calendar spread positions: long winter delivery months against short non-winter months, spanning 2006 through 2010. When winter-summer spreads narrowed in September 2006, the fund's positions were too large to exit, triggering a liquidation spiral where forced selling caused further losses.

How large were Amaranth's positions relative to the market?

Amaranth controlled up to 40% of NYMEX open interest in winter months, 81% of the December 2007 contract, and 60% of all 2010 NYMEX natural gas futures. Its January 2007 position exceeded 80,000 contracts, equal to the entire U.S. residential natural gas consumption for that month.

What is a liquidation spiral?

A liquidation spiral is a vicious cycle where forced selling causes prices to move further against the distressed position, triggering additional margin calls, which force more selling at even worse prices. Amaranth experienced this in mid-September 2006, losing $560 million in a single day and $800 million over two more days before transferring its positions.

Could the loss have been predicted?

Till's scenario analysis shows that if the fund's key spreads had reverted to any of their end-of-August levels from the previous six years, losses of up to 36% were possible under normal conditions. This analysis required no advanced modeling, only a comparison of current spread levels to historical norms.

How did Citadel profit from the Amaranth collapse?

Citadel received $1.425 billion in net concessionary payments for taking on Amaranth's distressed portfolio, then unwound the positions over approximately six weeks through orderly execution. Because the unwind was gradual rather than forced, Citadel likely sustained only small trading losses, resulting in substantial net profits from the transaction.

What does a natural gas calendar spread trade look like?

A calendar spread involves simultaneously buying and selling futures contracts on the same commodity but for different delivery months. Amaranth was long winter months (when heating demand raises prices) and short spring/summer months (when demand drops). The trade profits when the gap between winter and non-winter prices widens.

How does position sizing relative to market liquidity apply to equity trading?

The core principle is the same across all markets. If your position represents a large fraction of average daily volume, your exit is constrained because selling the position will move the price against you. In RealTest, this translates to volume filters in entry logic that prevent entering positions in stocks where your capital would be disproportionate to liquidity.

What is the difference between VaR and scenario analysis for concentrated positions?

Value-at-Risk uses statistical distributions and short lookback windows to estimate potential losses. For concentrated positions, VaR tends to underestimate risk because it does not account for illiquidity or the market impact of forced selling. Scenario analysis asks a specific question: what happens if this specific variable moves to a specific historical level? Till's 36% loss estimate came from scenario analysis, not VaR.

What regulatory actions resulted from the Amaranth case?

Both the CFTC and FERC pursued manipulation charges. Amaranth settled for $7.5 million with regulators and $77.1 million in class actions. A 2013 court ruling established that FERC lacks jurisdiction over futures contracts, giving the CFTC exclusive authority. Amaranth's head trader settled with the CFTC for $750,000 and a partial trading ban in 2014.

How can systematic traders apply these lessons in RealTest?

Set position sizing limits as a percentage of average daily volume so no single position becomes illiquid. Run backtests across multiple market regimes including stress periods and measure worst-case drawdowns during regime transitions. Evaluate whether max drawdown at any point in the backtest would have triggered a liquidation spiral given your actual margin constraints.

Related strategies

RealTest Mean Reversion Trading Strategy for 2025
Mean reversion strategy with volume-based position sizing filters that prevent entering illiquid positions, directly implementing the liquidity constraint lesson from the Amaranth case.
View →
RealTest Short Term Mean Reversion Strategy
Short-term strategy with daily volume filters in entry logic, ensuring position sizes remain proportional to available market liquidity.
View →
← Back to Research