Nanex ~ 11-Dec-2012 ~ Money and SpeedMoney & Speed: Inside the Black Box is a true thriller that takes us to the heart of our automated financial world. On the basis of interviews with people directly involved and data visualizations to the millisecond, a reconstruction of the fastest and deepest drop in U.S. stock markets ever. |
| Commentary by Eric Hunsader. Much has happened since the filming of this documentary (end of October 2010). For one thing, I'm no longer intimidated from speaking what I know. For example, in the film, you'll see that I am reluctant to call the SEC report nonsense, despite proding from Marije Meerman (the director). Instead, I somewhat sheepishly point out what they missed. Since then, I've spoken with many people across the industry, on and off the record, and have come to understand a lot about what happened on that day (May 6, 2010). Enough to say unequivocally, the SEC report put out October 1, 2010 is complete and utter non-sense. For example, they completely misunderstand the term "liquidity" even though that word is used 249 times in 89 pages of that report. The entire opening paragraphs that talk about an algo that doesn't take time or price into account, and therefore begins selling more an more contracts as volume explodes: again, that is pure fabrication. How can I be so sure of this? Because I talked to the people who actually executed the 6,438 trade executions that made up the 75,000 contracts Waddell & Reed (W&R) sold that day. I (an apparently, no one else) was given those detailed trade records and matched every single one up to the 147,577 trades that took place in the eMini during that time. It is crystal clear from the data that the algo only sells at the offer. But in the film, listen to what Greg Burman says after Marije tells him what we (other folks) discovered: "5 different people can look at data and give you 7 different answers" This isn't economic forecasting. This is counting how many trades executed at the bid (never) or offer (all of them), and at what times (the majority after the market hit a low). This is equivalent to a math teacher grading her student's multiplication tests and a beaurocrat thinking that 5 different math teachers would come up with 7 different grades. Even worse, and this was very difficult for me to believe: the regulator never interviewed the traders that executed the W&R contracts until October 14, 2010 - that is 2 weeks after they wrote the report! Isn't that the primary purpose of investigating? You know, asking the people who are found guilty what happened. If only the regulator had asked the traders about the algorithm, they would have immediately learned it was passive: it sold by posting offers above the last executed price. It didn't cross the bid/ask spread. And the algo had price and time components: it would turn off for 15 to 45 seconds if the price moved to far and/or to fast: this is why that algorithm sold only 1,000 contracts during the time the DOW dropped 600 points. All these facts run directly contrary to the SEC report. 5 months to assemble the data. May 6th + 5 months is October 6th is when they'd first be able to analyze the data. Yet they wrote the report on October 1, which is 5 days before. "Interesting things tend to chunk up on the one second and one minute interval." Unbelievable. The ability to make up stories like that. Here's what really happened. They didn't have the capability to look at data in a finer resolution, but rather than admit that, decided to make up a story and hope the viewer doesn't notice. The report used 1 minute snap shot data (equivalent to inspecting one page out of a stack of paper 20 feet tall), and 1 minute samples (averaging all 20 feet of paper into one page): no surprise why they couldn't find anything. This is like looking for the Higgs Boson with a student microscope. |