Monday, May 24, 2010

Late-Day Market Surges Revisited

Some people astutely noted that this study on late-day market surges triggered on Friday. Unfortunately, since the original publication, the edge has not persisted. Below is an equity graph to demonstrate.

I've market the original study date on the chart. Up to that point there had only been 2 out of 12 instances that rose the next day. Since then it's been a coinflip but the gainers have strongly outsized the losers.

In the original post I listed all of the instances up to that point. Below I have listed all the instances since then. Market behavior is always evolving. It is important to continually monitor edges to make sure they are still providing an edge. This is something I constantly do in the Subscriber Letter. Over the last year and a half since I first published the study the market has: 1) Shown a much greater inclination to experience these late-day market surges. There have been more instances in the last 2 years than in the previous 25 years. 2) With the market more inclined to experience a surge, it's meaning also seems to have changed. It no longer seems to suggests a downward edge for the next day.


Cobra said...

Haha, you mean I'm that astutely guy? Well, at least I bring you some traffic, not only from my blog, but also from :-)

Anonymous said...

Terrific blog. I just wanted to ask you a question regarding this study. Late-day surges have been particularly profitable over the years (defined slightly differently than you have here)for a trader named Gary Smith (He wrote the book "How I Trade for a Living". Of course he was more interested in staying with a developing trend than in just the next day's outcome. Have you considered looking at the outcome several days, weeks and months after such late-day surges? Especially if they occured after a down day or period of down days in the market? You may want to explore Gary Smith's book to see how he defined late-day strength. It is more flexible than your criteria here but also a bit more difficult to define for your backtests. Anyhow I would love to hear your reply! Thanks...