After reading this blog for six months, I noticed the funny way of today's market close. Thanks Bill for teaching me.

When computers are as intelligent as humans, they may be able to trade. We are still many, many years from this event.
Do computers have any significant impacts on markets? What is it? Can we take advantage of it? I am thinking that low short-term volatility of markets in recent years can be attributed to computer algorithms that try to take advantage of immediate imbalances, so I look for mid-term trades.

I also noticed that many small-cap stocks open with a wide-open spread between bid and ask price to fool people who use market orders. The spread will disappear in 10-20 minutes after market opening. This was not a case few years back, especially for many stocks with decent volume. Perhaps, there was some consolidation and some market makers do not serve those stocks anymore or their computers were quietly reprogrammed.

I've recently noticed odd bursts in trading volumes for some stocks, mostly showing up at the end of the regular session (4:00 pm). Seems like it was the case today when looking at the volume window for today's Yahoo DOW chart:
http://www.aloha.net/~scottp/b.png

Here is a graph showing todays (5/31) trading volume for FCX:
http://www.aloha.net/~scottp/FCX%20Trading%20Volume.htm

Does anyone know if the volume bursts seen at the beginning and end of the regular trading session is related to what Bill is describing above?

Thanks....

As Bill mentions from time to time, to trade is like to dance... a nice way of describing the minimum skills needed if you want to play within a Complex system (http://en.wikipedia.org/wiki/Complex_systems). Those "quantitative PhD-armed staff" mentioned in the article tease and torture information from the historical data, model parts of the system with it and then try to automate... at least for a song or two. Unfortunately more than a few hedge/funds have found out the hard way: brilliant pure stats/math does not consistently a trading system make. Worse, all that math and stats have been known to lull or blind the management team into complacency. Interdisciplinary creative stat/mathematicians who also know how to dance are not usually the personality types to rush out of uni and work in "exciting" bank jobs, that I have ever witnessed anyway :).
Complex systems have relationship feedback loops (http://en.wikipedia.org/wiki/Feedback) that move to remove the advantage being exploited. This does not just affect automated systems however; for a recent example how many discretionary traders were short the financial's after successive rate hikes: since "everyone" knew they had to go south, only to watch them rocket north? More than a few I'll bet, all following predictable "reliable" historical models. Other classic examples would be seasonal effects suddenly not being, or one very popular trend following service for the Q's who's performance has hit a bit of a wall. Due in no small part (from my research at least) to it's membership having around $30 billion under management of the system. So now the service is scrambling to diversify the markets where their members receive signals, to try and get back under that (apparently) invisible barrier that fall's without warning to squash exploitable inefficiencies in the system. Further, only those select few with access to the exchange order flow have the true finger on the pulse of the system, complete with little niceties such as future order info sitting bright on the screen - a true crystal ball into the future. Everyone else is must play catch up on the dance floor and hope that their actions (discretionary or my preference: algorithmic) won't shine too bright on the big screen when they make their next dance step ;).

Everything said in this blog is rubbish. You clearly have absolutely no clue how these algorithms work, either for the buy side or the sell side. Buy-side quantitative algorithms are by design aware of risk and they usually have a better understanding of volatility and its distribution than humans do..algorithms provided by brokers merely try to efficiently trade positions. They very rarely use technical indicators. And 4 does come after 1,2,3.

There is something to quantitative trading:

http://www.fool.com/news/commentary/2006/commentary06060203.htm

There's no other business on Earth like the hedge business. In 2005, according to Institutional Investor, the average top-25 hedge fund manager made $363 million. James Simons, head of hyper-quant Renaissance Technologies, made $1.5 billion! Renaissance's black box speed-trades any public, liquid security, which it can put into its math models. The flagship fund produced remarkable returns (60% before fees) and even more remarkable fees -- 5% of assets and 44% of the gain.

From HedgeWeek:
http://www.hedgeweek.com/articles/detail.jsp?content_id=25065

James Simons of New York-based Renaissance Technologies Corp earned an estimated USD 1.5 billion in 2005, making him the top earner in the latest ranking by Alpha Magazine of the world's highest-paid hedge fund managers.

Snippets from other sources:

The 68-year-old financial guru, who once served as math department chairman at Stony Brook University, rose to the No. 1 spot in a banner year for top hedge fund managers, who saw their average $363 million in earnings jump by 45 percent from the year before, according to the survey. Simons' own income more than doubled from the estimated $670 million he earned in 2004.

For years, Simons' firm has favored a math- and statistics-based approach to his managed portfolio, often hiring those with doctorates in math and science to help oversee his company's investments.

On Long Island, the low-profile Simons is known as a benefactor to Stony Brook University and Brookhaven National Laboratory - where he recently gave multimillion-dollar donations - and several other charitable causes.

There is a very long article about Jim Simons at Institutional Investor
http://www.dailyii.com/article.asp?ArticleID=1028172

It's a very good read.

Teresa

There are lots of themes in this thread. Bill's basic points have to do with scale and homogenous thinking. We know from ecology that diversity in strategies make for a healthy system. When one strategy takes over, it succeeds only for so long until it crashes the whole system.

As for programmed trading, the success of the system rests in the expectations of it. If you expect that it will replace human intelligence, it will fail. People and machines are fundamentally different. You cannot program intelligence. My favorite book on this subject is "What Computers Can't Do" by H.L. Dreyfus. This book injects common sense into the AI discource.

People have been trying to replace intelligence with machines since the end of WWII. I view the tech bubble as a result of a general naivety by the investing public on this issue. The Lernout and Hauspie were at the core of the tech melt down. I think this particular scandal is as significant as Enron. At the core of it was classic fraud, but the real issue is: what was the public's expectation of voice recognition? I believe the level of investment plowed into Lernout and Hauspie revealed a very naive public that imagined HAL like computers driving our economy. To some extent, the entire bubble was being inflated by these naive Jetson Family style fantasies.

The over dependence on machine trading is an extension of this false god of AI. It is in a longer cycle that the tech bubble itself. But sooner or later, it is going to be revealed as an instrument of market obfuscation.

On the other hand, program trading is working at some level. This gets back to Bill's basic points about how it is distorting the market. These kinds of tools work in the context of human understanding, not independent of them.

Able Ape