News Focus
News Focus
Replies to #33 on The Darvas Method
icon url

S Chun-Li

11/28/02 1:41 PM

#34 RE: MechanicalMethod #33

I adjusted all first hour trades to zero profit and the system was still down nearly 200 pts on the period.

Maybe it works better on the S&P/NYSE but I suspect that any TA is going to find things difficult if this rally is being helped by PPT. $VXN has bottomed and reversed upwards a few days ago. That is a very reliable bottom signal but the Nas comp still made a new high since then.

icon url

S Chun-Li

11/28/02 2:17 PM

#35 RE: MechanicalMethod #33

I have formed some "by eye only" impressions of the oddball system as follows:

It looks like the idea behind the system is sound in that increasing advancing issues lead to higher prices.

I think the weakness lies in how to identify/define increasing advancing issues.

For instance, the advancing issues on a Tuesday in the third hour of trading might be above normal and, if spotted, provide a signal to higher prices. BUT, if Monday's third hour of trading had a bigger advancing issues acceleration, then it makes Tuesday's advance look like a decline.

So my immediate concern is to find a better way of comparing than just using the same period of the previous day.

Maybe use an average of the difference between the two periods for the prior two or three days.

I intuitively believe that markets are influenced by a phenomenon called "weekends". On a Monday there are pent up investor decisions that have been building up for three days. On a Friday there is a reluctance to hold positions over a weekend. Also, futures are closed for an extended period over the weekend, so there isn't time for imbalances to be ironed out by the "invisible hand" of the market.

I don't think a trading signal derived from comparing a Wednesday to a Tuesday can be calculated the same way as a signal derived from comparing a Monday to a Friday. Maybe the comparison period should be a whole week...

I will look into these things but I seriously need a source for intraday data going back a lot further than the <three months' data I have at hand.

Ideally, I would like 5 years of data to ensure that bull market and bear market are both represented.