-Note: Mea culpa to anyone who may have read this (fewer than nine people, since I don't advertise, write about, or tell anyone about this site), I may have been doing the digital equivalent of holding my graph upside down. I just started using Tableau software as part of my analysis, and I had interpreted percent difference relative to last price to mean last price/current price, when in fact it is current price/last price. This would mean the results are the opposite of what I wrote below. I am looking into this as this is in stark contrast to earlier tabulations I have performed in excel, common sense, and so forth. I'm looking into this further.-
I have been tracking discounted cash flows since December 2012, and have used a fairly consistent methodology since January 2013. Over this time frame, I have not found much evidence that investing in companies that are trading at a discount to their net present value results in gains. However, I have found that traders and investors are consistently rewarded when they purchase stocks that are currently increasing in calculated net present value.
It is very common to see traders attempting to prove a system by touting the results of a backtest. I believe that this approach is highly prone to "sharp shooter" bias. Given enough arbitrary parameters, timeframes, and rules, a system that *would* have worked can be carved out of any idea. It worked, provided you avoided it when it didn't.
Alternatively, when I test an indicator, I look for a positive correlation over every the entire universe of stocks, that is consistently reproducible. This existence of such a correlation does not suggest a precise trading strategy- but I believe it does prove that the indicator in question actually does indicate something.
The Tableau visualization below show my full results thus far.
The weekly results test a holding period of one week. The change in NPV value over the previous week is plotted against the price change over the next week. This has resulted in a positive correlation in 17 of the 24 weeks for which I have results. The old adage "sell in May and go away" has held true, as all seven of the counter-trend weeks occurred during the summer.
In the full term results, no time frame is specified. The historical change in NPV value is measured from the first week on record, February 8th, and the resulting price change is measured through the present for each weekly dataset. Measured in this fashion, a positive correlation between historical changes in NPV and future price changes has developed in all 24 weeks.
Note that for both full term and weekly results, the first and last week in the analysis are not meaningful. For the first week there is no historical NPV, and for the last week the future price change is not yet known. I apologize, as I can not figure out how to turn off the visualization of those weeks without turning off the data as well.
And of course, I intend to continue measuring these results every week.