InvestorsHub Logo
icon url

Tom K

09/07/03 12:36 PM

#1511 RE: exp #1507

exp, haven't done any quantitative backtesting but would love to. The main hurdle is my lack of programming/scripting experience. Depending on the price tag, I wouldn't mind paying someone to do this for me.

All of the indicators I use are based on both bullish and bearish periods, some as far back as 1990. Most of them are also based on the indicators and methods used my Ned Davis Research that have been tested for much longer time periods (see The Research Driven Investor by Tim Hayes).

I have tested my system in real time since last summer - results have been excellent, granted, the period is relatively short and was mostly bullish.

Responses to your general points:
(1) It's definitely a trend following system with this caveat: I use several contra-trend indicators as well as a mode scoring system to counter-balance the overweighting of trend indicators. For example, my NYSE model crossed above zero on October 10th 2002, one day after the lows for $NYA. How can that happen? A combination of extremely bullish (bearish) mode scores for contratrend indicators in combination with breadth trend indicators that have lower mode values (-1 or +1 vs. -3 or +3) because they're oversold (overbought) can signal a an imminent trend change BEFORE your typical trend-following indicators do.

(2) True.

(3) True.

Specific points responses:

(1) Always "late entries". The model is heavily weighted with price and breadth trend indicators. Again, contra-trend indicators and mode scoring usually signals entry/exit points earlier than using a pure trend-following system.

(2) Yes, backtesting would really help in determining optimal number of positions and size per position. These levels are fairly arbitrary based on the past year's performance.

(3) I agree. This could change if Profunds would increase the number of sectors they track. For example, Airlines are currently ranked within the top 4 in RSC - but Profunds only tracks 15 sectors.

(4) Agreed. 100% long translates into 150% long.