glippert@trwrba.UUCP (George A. Lippert) (05/24/85)
I am sure that the subject of computer-based commodity trading is not new to this newsgroup but I have just recently subscribed and our system's file of the newsgroup is small. I would like to hear about people's actual, or "paper", trading experiences, simulations, whatever, and descriptions of the computer programs that the trading was based on. My interest was piqued by an article by J.M.Keynes (Money DOS, 80 Micro, Sep. 1982) in which a short BASIC program was given that was reported to produce 38.9% net return yearly and was based on three moving averages of different lengths. Inspired by this, and by the discovery that a friend had several years of daily commodity data on his computer, I spent some time doing an exhaustive search for the optimal three MA lengths. I was pleased to find that with an initial seed of $7k in 1969, trading only one Chicago soybean meal contract at a time on the average of once a week, in 1980 the program returned $206k (using a $40 broker fee per trade). Now I'm no financial whiz (I have trouble with the short form) but this seems good to me. The optimal "magic numbers" used above for the three moving average lengths were (1,19,21) days, and were used according to Keynes strategy (Keynes' numbers for this commodity were (4,9,18) ). I have seen some advertisements for commodity programs where the fact that they don't use moving averages was highly touted. Why is this?