Market Lessons, from Henrik Andersson
These were the top 5 lessons that the market taught me last week:
1. When the meme is to trade everything on the second derivate it doesn't matter that earnings beat by 114% as the container division in x did.
2. When the world is balancing between hope and despair (double dip) the risk premia ought to be very high and thus the risk reward owning equities unusually favorable.
3. It often pays to stick to your long term cases and not change around too often. xx lost a quarter of the market value last week. xxx finally moving higher etc…
4. Bad/ low quality companies have a tendency to be unlucky. xxxx comes to mind
5. Analysts have a predisposition not to care what is priced in. xxxxx is lower now than before the news that Finland will not implement the windfall taxes.
Steve Ellison writes:
The lesson I have learned this summer is that the counting trader should consider the data mining bias and expect that returns from a backtested method are likely to be lower in real trading than in the backtest. As Bacon recommended, having a trial period for a new method in which one paper trades or trades with tiny amounts can be valuable for assessing whether the method performs well out of sample.
Relatedly, I first heard of the Hindenburg Omen in 2005 (here is a link from that time. Therefore, while some suspect data mining or curve fitting, I am certain that at least the last four occurrences were out of sample. In 3 of those 4 cases, the S&P 500 was lower three months later (and I evaluated only three-month returns, not other intervals, which would have been another form of data mining):
Date S&P 500 3 Months S&P 500
of Signal Close Later Close Change
10/5/2005 1207.8 1/5/2006 1281.3 6%
4/18/2006 1324.5 7/18/2006 1245.7 -6%
10/19/2007 1515.3 1/18/2008 1325.3 -13%
6/18/2008 1341.6 9/18/2008 1203.2 -10%