【英语财经】人类操盘手如何胜过电脑？ How human traders will beat the machines

When Garry Kasparov sat down to play the IBM Deep Blue computer the Russian chess grandmaster believed he had discovered a strategy to turn the machine’s greatest strength into a weakness.

Deep Blue relied on being able to compute a vast database containing hundreds of thousands of chess games played by past grandmasters, meaning Kasparov was not simply playing one supercomputer but in fact taking on the amassed knowledge of many of the strongest players from history all at once.

But to be able to make use of large parts of its database, Deep Blue required its opponent to play like a typical grandmaster. If Kasparov was to intentionally make a bizarre opening play rarely seen in high level games the computer would have vast parts of its database rendered useless, as it would have fewer games to reference, and the human could regain the upper hand.

Kasparov’s “man verses machine” struggle with Deep Blue and his eventual defeat is a frequent refrain in discussions about the future of financial trading, and the likelihood that ever advancing technology will render the human fund manager obsolete.

Many human investors are enduring a miserable start to 2016 as stock markets and commodity prices have slumped. At the same time some trend following computer-driven hedge funds have made double-digit returns during the sell-off. Is man once again on the verge of being trumped by the machine like the Russian grandmaster? The answer is no. Here are three reasons why the best human investors are likely to be able to beat the “bots” for a long time to come.

First, human investors can follow Kasparov’s strategy and seek to use the strengths of algorithmic trading programmes against them. To paraphrase Oscar Wilde, trend-following hedge fund computers know the price of everything, but the value of nothing. This means human investors who focus on value over the long term, rather than price trends, should always be able to profit.

Hedge fund computers rely on their analysis of millions of data points from past market movements to use this to predict how markets will behave in the future. Broadly speaking, when any large market moves in one direction for a period of time the trend following computer will be able to profit from this, as many have done this month.

These computers however are not investors in the true sense of the word, but semi-automated trading bots seeking signal within market noise. Their models analyse price data, rather than creatively assessing, for example, how accurately a share that represents fractional ownership of a real business reflects the present and future value of that business.

The ability to process millions of numbers faster than a human fund manager can tie their shoelaces provides an advantage, but it makes such models inherently prone to following the consensus. Episodes from last year, such as the collapse of the seemingly hugely profitable Hong Kong solar group Hanergy, or the debate over the business model of Valeant, show the limits to superficial, data-based valuation methods.

Until computer traders can develop genuine artificial intelligence they will remain unable to gain an edge over the best human investors in spotting a catastrophic disruptive threat to an industry, or a revolutionary emerging technology.

Second, the computer-driven hedge fund is run and designed by humans, meaning its cold-eyed technology must contend with the emotional reactions of investors who are human. When many so-called Commodity Trading Advisor hedge funds suffered large losses in 2013, their investors began to pull out their money.

This placed great emotional strains on the humans who had designed them. No matter if a mathematical model shows that a large loss is statistically possible, the temptation to fiddle remains great. Even if over the long run, or maybe even given an infinite timeframe, your model will be statistically profitable, repeated large losses mean your hedge fund will be forced to close through human weakness.

Third, the advancement and wider adoption of computer-driven trading systems may sow the seeds of their eventual obsolescence. If, in the future, they ever become advanced enough to come close to “solving” trend following in financial markets, then the machines will gradually erode each other’s ability to make money. When everyone is using the same trading strategy, no one is left with an edge — the super computers will cancel out each others’ advantage.

Raw processing power may have its uses in financial markets, but until scientists develop a truly intelligent investing system, rather than a trend follower, the truly skilled human fund manager has no reason to fear.

当加里?卡斯帕罗夫(Garry Kasparov)坐下来和IBM的计算机“深蓝”(Deep Blue)对弈的时候，这位俄罗斯国际象棋大师相信，他有办法让机器的最大强项变成弱项。

“深蓝”依靠一个庞大的数据库进行运算，这个数据库包含过去的国际象棋大师的数十万棋局，这意味着卡斯帕罗夫不仅仅是和一台超级计算机对弈，而是在一盘棋局中与历史上许多最强大棋手的知识的集合体进行较量。

但如果“深蓝”要利用其庞大数据库，就需要对手像一名典型的国际象棋大师那样下棋。如果卡斯帕罗夫故意走出一个在高水平对弈中十分少见的古怪开局，计算机数据库中的很大一部分就变得毫无用处了，这意味着它能够参考的棋局更少，人类对手可以夺回优势。

在有关金融交易的未来和日益先进的技术淘汰人类基金经理的可能性的讨论中，卡斯帕罗夫与“深蓝”的“人机之战”以及他最后的败北常被提及。

随着股市和大宗商品价格暴跌，许多人类投资者在2016年经历了一个惨淡的开局。同时，一些利用电脑操盘的趋势跟随型对冲基金在抛售中斩获了两位数的回报率。是否就像那位俄罗斯国际象棋大师一样，人又一次处于被机器打败的边缘？答案是否定的。有三个理由相信，在今后相当长时期，最优秀的人类投资者有望胜过自动程序。

第一，人类投资者能够效仿卡斯帕罗夫的策略，尝试利用算法交易程序的优势来胜过它们。借用奥斯卡?王尔德(Oscar Wilde)的一句话，趋势跟随型对冲基金的电脑知道一切事物的价格，却不知道任何事物的价值。这意味着关注长期价值（而非价格趋势）的人类投资者应该总能盈利。

对冲基金的电脑对过往市场变动的数百万数据点进行分析，据此预测市场未来的走势。大致上说，只要任何大型市场在一段时间内朝着一个方向变动，趋势跟随计算机程序就能从中盈利，本月就有很多这样的例子。

然而，这些计算机程序并不是真正意义上的投资者，而是在市场的噪音中搜寻信号的半自动交易程序。它们的模型分析价格数据，而不是进行创造性的评估，比如一只代表一家实体企业部分所有权的股票在多大程度上准确反映了该企业现在和未来的价值。

程序处理数百万数字的时间比一位人类基金经理系鞋带的时间还短，这能提供一种优势，但这些模型本质上倾向于跟随共识。去年的几件事，比如看似利润丰厚的香港太阳能公司汉能薄膜发电(HTF)崩溃，或者围绕制药公司Valeant商业模式的辩论，表明了基于数据的肤浅的估值方法的局限性。

计算机交易员在发展出真正的人工智能之前，是无法在发现行业面临的灾难性威胁或某种革命性的新兴技术方面胜过最优秀的人类投资者的。

第二，使用计算机程序的对冲基金是由人类执掌和设计的，这意味着冷硬的科技必须应对身为人类的投资者的情感反应。2013年许多所谓的商品交易顾问(CTA)对冲基金蒙受巨大损失时，投资者开始撤资。

这给自动程序的人类设计者带来巨大的情绪压力。即便数学模型表明从统计上说有多大可能发生巨亏，人类忍不住“自行其是”的诱惑依然很大。即使从长期来看，乃至在时间无限的条件下，你的模型在统计上是盈利的，重复出现大笔损失意味着你的对冲基金将因为人的弱点而被迫关闭。

第三，利用电脑操盘的交易系统的进步和推广或许会播下淘汰它们自身的种子。如果在未来，它们变得足够先进，接近“破解”金融市场中的趋势更随，那么机器将逐渐侵蚀彼此赚钱的能力。当大家都使用同样的交易策略，那么没有一方能够“技压群雄”——超级计算机会抵消彼此的优势。

原始数据处理能力或许在金融市场有其用处，但在科学家开发出真正智能的投资系统（而非趋势跟随者）之前，真正有本事的人类基金经理没有理由担心。

译者/许雯佳