[pcabc]

Maroon alert!

As discussed in my earlier post in this thread I'm trying to come up with a system to alert me when there are serious breadth warning signs. My previous short / longer term breadth without these warnings is OK for a trader but not stern enough for an investor. I've recently moved to an investor position with an aim to hold on to positions that I simply would have got out of as a trader long before then.  That has meant that I have ridden a few positions and funds down.

I've made some further progress on my alert system.  Its not a magic bullet, it is work in progress and I'm putting it up here to see if sharing the ideas are useful to anyone.

Now:
   
1. Is the aggregate breadth indication.
Green=Buy
Grey = hold
Yellow=Sell/hold short term (tight stops if hold).
Red=Sell/hold long term (tight stops if hold).
Maroon=DANGER - price still high - divergence.
Black=DANGER - price has declined.

2. Maroon alert two weeks before the drop.  This is mainly due to the position of the Weinstein momentum relative to its 20 and 50 day MAs , 3, and the behavior of the % stocks above their 200 day EMAs, and the averages of that line, 4.  Other parameters are less useful, e.g. New Highs minus new lows, 5 only seems to act as confirmation.  The bars below each line show amber and red for danger.  The lower bars have the signals stretched out to give them more overlap so that when they are combined there is more change of overlap and multiple short signals are not missed.

I've been working on the advance / decline line as both now and in 1987 there was a divergence from price. I've been trying to measure the time since price and the a/d line touched the top of their respective 120 day channels.  But the differing time scales of the divergances means picking up a useful signal is tricky.


1987:
   
Again I picked up black and maroon danger warnings about 3-4 weeks before the precipitous drop, 1.  Again Weinstein momentum and the % stocks above their 200 day EMA diverged, 2 & 3.  There was a divergance in the A/D line too, but the time period was too long for my code to pick it up.    

I am trying to work up code that picks up the breadth decline signs for both these drops.  Whilst I could produce code that might pick up each separately and then AND the results together that would be fine tuning for specific circumstances and miss the next one.  The other challenge is not to pick up too many false alarms, as false alarms devalue the system leading to it being ignored.