他她話題:電子交易系統的風險

Knight Capital, the market maker at the centre of the latest malfunction in electronic share trading, bills itself as: "the science of trading, the standard of trust".

近日遭遇電子股票交易故障的美國做市商Knight Capital Group以"交易科學，信任標杆"作爲宣傳語。

Yet that public trust in US equity markets is being severely tested now. Following on from the technical glitches that marred the flotations of Facebook and BATS Global Markets this year, comes Wednesday's erroneous trading in 148 shares listed on the New York Stock Exchange.

然而，美國股票市場的公信力正遭遇嚴峻考驗。今年以來技術故障已導致Facebook和BATS Global Markets首發上市出現問題，上週三紐約證交所(NYSE)又出現了148只股票交易異常的情況。

Once more investors and traders in the US equity market are left bewildered by another episode of an algorithmic or "algo" trading system running wild.

這起算法交易系統故障再次令美國股市的投資者和交易員感到無所適從。

More than two years after the flash crash of May 2010 – when Wall Street tumbled hundreds of points within minutes – it demonstrated how much of the US equity market in effect operates on autopilot. It also serves as a reminder that trading firms, exchanges and regulators still face huge challenges in safeguarding how stocks trade in a superfast dealing environment.

在2010年5月"閃電暴跌"(flash crash)事件發生兩年多以後——當時道瓊斯指數在半小時內暴跌650點——這件事折射出了美國股市自動化交易的普遍程度，對於交易商、證交所以及監管當局而言也是一記警鐘，提醒他們：在高速交易環境下保障股票交易正常進行方面，他們仍面臨艱鉅挑戰。

While the wider market was unaffected by Knight's "technology issue", suggesting that regulatory and exchange reforms since the flash crash helped to prevent a panic, the episode has revived a long-running debate over the benefits and dangers of relying on complex computer systems.

儘管整體市場未受Knight"技術問題"影響（似乎表明監管當局和交易所自"閃電暴跌"以來推行的改革措施對防止恐慌起到了作用），但這件事重新引燃一場持續多年的辯論，其主題是依賴複雜電腦系統的效益與風險。

Much trading in equities now comprises sophisticated algos that are created to trade in specific ways, reacting to changes in prices, sectors, volatility, economic data and other parameters across the fragmented trading landscape.

如今，美國股市的很大一部分交易由複雜的算法執行，這些算法的設計宗旨是以某種特定的方式進行交易，對價格、行業形勢、市場波動、經濟數據以及跨越交易版圖上不同板塊的其他參數的變化作出反應。

By replicating the way human traders ascertain where other investors want to buy and sell stocks, the algos spit out a torrent of orders, many of which are quickly cancelled before a trade is executed, all within fractions of a second.

通過複製真人交易員判斷其他投資者買賣意向的方式，算法程序發出一連串交易指令（其中很大一部分在執行前被取消），整個過程耗時不足幾分之一秒。

Since the advent of such trading in the past decade, the average trade size has dropped from 1,000 to 200 shares. Some market participants and observers worry about the risks of an "algo" running out of control.

自從過去10年裏算法交易誕生以來，單筆交易的平均規模已從1000股下降至200股。一些市場參與者和觀察人士擔心算法交易存在失控風險。

"We are still vulnerable to all kinds of glitches, no matter what changes and precautions we make, there will always be an unanticipated scenario," says James Angel, associate professor at Georgetown University.

喬治敦大學(Georgetown University)副教授詹姆斯•安吉爾(James Angel)指出："我們在各種技術故障面前仍然十分脆弱。無論我們採取什麼改革或防範措施，總會有意想不到的情形出現。"

In the case of Knight, which suffered a loss of $440m, Tom Joyce, chief executive, told Bloomberg yesterday the implementation of new software contained a bug that sent erroneous orders into the market. "So we ended up with a large error position which we had to sort through the balance of the day," he said.

Knight公司首席執行官托馬斯•喬伊斯(Thomas Joyce)上週四對彭博(Bloomberg)表示，發生技術問題是由於公司新安裝的軟件存在缺陷，向市場發出了錯誤的交易指令。他說："結果我們持有規模龐大的錯誤頭寸，被迫在當日餘下的時間進行整理。"

Traders said they believed a large order for a number of stocks was executed in five minutes rather than a longer period of either one or up to five days. The scale of orders pushed prices sharply higher or lower and overwhelmed the liquidity of the market.

交易員們認爲，涉及多隻股票的大額交易指令在5分鐘內就執行完畢，而沒有在一整個交易日甚至五個交易日的更長時段內分批進行。這些指令的龐大規模導致股價劇烈波動，超出市場流動性的承受能力。

It also sparked wild trading in exchange traded funds; the Vanguard Utility ETF experienced volumes of 4m, well beyond its usual 128,000 trades a day.

該指令還導致交易所交易基金(ETF)的交易量暴增，先鋒(Vanguard)公用事業類ETF的交易量達到400萬筆，遠遠超過每日12.8萬筆的正常水平。

Since the flash crash, the market has put in place circuit breakers, designed to limit wild swings in the price of stocks beyond a rise or fall of 10 per cent during a five minute period.

閃電暴跌事件後，市場引入了"斷路開關"機制，以扼制股價劇烈波動，使股價在5分鐘內的漲跌幅度不超過10%。

In total at least 40 of the 148 stocks affected by the Knight trading algo experienced a change of more than 10 per cent, as circuit breakers do not start operating until 15 minutes into the trading day and reflect volatility, not order flow.

受Knight算法交易影響的148只股票中，至少有40只股票的股價波動幅度超過了10%，因斷路開關機制在開盤15分鐘後才投入運行，而且該機制反映的是股價波動性，而非指令流量。

Among the stocks that triggered circuit breakers on Wednesday were Molycorp, CoreLogic, Kronos Worldwide, China Cord Blood Corporation and Trinity Industries. The changes in their share prices went well beyond 10 per cent, with China Cord jumping as much as 151 per cent before falling back.

上週三觸發了斷路開關機制的股票包括莫利礦業(Molycorp)、數據供應商CoreLogic、化學品生產商Kronos Worldwide、中國臍帶血庫公司(China Cord Blood Corp)以及鐵路運輸商Trinity Industries。這些公司的股價波幅遠超10%，中國臍帶血庫公司股價在回落前曾暴漲151%。

"The industry has been well aware of defects with the circuit breakers," says Prof Angel.

安吉爾教授指出："斷路開關機制的缺陷業內早已知曉。"

Meanwhile, only trades in six of the 148 stocks were subsequently broken by the NYSE. That leaves Knight potentially facing substantial costs and raises broader questions. "This is not only a problem for Knight, but creates a whole new thing for regulators to dwell on and for investors to lose trust in the market," says one Wall Street analyst.

在148只受影響股票中，紐約證交所只中斷了其中6只股票的交易。這使Knight可能面臨重大損失，並帶出更廣泛的問題。一位華爾街分析師表示："這不是Knight的個體問題，而是監管當局應該仔細研究的新領域，也讓投資者有新的理由對市場失去信任。"

For some, the notion of breaking trades underlines the problem of using and relying heavily on high-speed computer technology. Once a trade is executed it soon triggers a chain of positions that becomes very expensive to break.

在某些人看來，停止交易的概念，凸顯出使用並過分依賴高速電腦交易技術的問題。一旦一條交易指令被執行，將迅速觸發一連串新頭寸，這導致停止交易的成本非常高。

"There is no kill switch; the minute you hit the send button on an algo it's running and you can't stop it," says Ken Polcari, managing director at ICAP Equities. "Maybe we do need to reflect on where the line is between humans and computers."

英國毅聯匯業(ICAP)股票交易部董事經理肯•波爾卡里(Ken Polcari)表示："不存在徹底停止交易的總閘開關，算法程序一旦啓動，你就無法阻止它。也許我們確實需要認真思考真人與電腦之間的界線在哪裏。"

Advocates of electronic trading say it has enhanced transparency and provides investors with more competitive and fairer prices.

支持者稱，電子交易提升了市場透明度，併爲投資者提供了更有競爭力以及更公平的報價。

Critics point to the glitches and heightened volatility. "There's an old bumper sticker that says, ‘to err is human, to really foul up requires a computer'," says Prof Angel.

批評者則指出，電子交易系統常有技術故障並導致市場波動性上升。安吉爾教授表示："一句印在汽車保險槓貼紙上的老話是這麼說的，‘是人都會犯錯，但要徹底搞砸還需電腦'。"