Mid-Month Update: Well after a very strong start to the month, the model's May picks gave back most of the profits in the last downturn. Momentum positions have been hit worse than others resulting in the quick loss of profits. Transportation and industrials have also been hit hard partly due to the sudden rise in commodities prices. Lets see how the rest of the month follows: Model: -2.47% S&P: -1.48% Return since start of journal: Model: +10.53% S&P: -0.04% EDIT: I forgot to factor in dividends received from OSK and CRS. Numbers have been updated.
No the model is long only at this moment. I am working on a short equities model, as well as looking over some potential index trading model which can be used to vary exposure.
All positions liquidated. For the month of May: Model: -6.98% S&P 500: -2.32% Since start of journal: Model: +5.39% S&P 500: -0.90% I will post comments and the June list later today/early tomorrow.
Mahras I just discovered your thread which is quite interesting. I had a few questions about your backtesting procedures and live testing. 1) Survivorshop bias- Have you double-checked to see whether stocks like Enron or Worldcom have been totally removed from your historical data or precisely when they were dropped? Both of these stocks had "cheap" fundamentals about a year or two before they went bankrupt and may have been selected by your model during momentary strong technical periods. 2) What do you use as the entry price for Monday morning? Some momentim stocks open up on Monday morning, and you may get better results by assuming an entry price of the average of High and Low. In your live trades, it looks like you usually use the opening price, but I saw a few exceptions. For example, on May 1 the SWFT open price (on Yahoo Finance) is 30.30, but you bought it for 30.16. 3) How promptly is your live Fundamental data updated? Most services like Reuters have delays where a company might issue a new earnings report, but it doesn't get reflected in the database right away. I wonder how this would affect the back testing? In some cases, the back data might be updated later to reflect the correct earnings, but the live data would be operating under a one or two week delay when you pick new stocks.
Hey George_S. 1. There is no survivorship bias in the data. Many of the picks during the dot com bubble days, for example, doesn't exist anymore. This is one of the first thing I looked out for as I have been a victim of false backtests due to survivorship bias in the past. 2. My mistake on the SWFT error. I used by fill prices (30.16). It should be 30.30. I shall have this fixed as well as double check the other prices very soon. To report the results on the website I use the opening prices on monday to make calculations simple. 3. As far as I know the database I am using for backtesting was built with live data for the purposes of backtesting alone. I do not think it is affected by the delays but I shall double check on that. Thanks for bringing it up. I will post the list later tonight (I was away for the weekend).
Corrected the May results and posted up the June list. In case anyone notices any errors please post them up! www.aeroscapital.com
Yep momentum was the driving factor for that one. BTW for macro direction of the market I have been looking at potential fundamental models. For example here is one based on comparisons of stock yields and value: http://www.cxoadvisory.com/status/ I am also looking at something called the Arezi Ratio as well which is a yeild fueled model as well (compares stock yeilds vs cash). Here is the link: http://www.aetheling.com/MI/AreziRatio.html Both seem to be logical and shows historical results. What I am planning at this moment is to use a variety of fundamental models. When these macro forecasts determine that its "unsafe" to be in the market then short the index to hedge off exposure. If anyone else knows of any resources like this feel free to share.