There should be a statistic average absolute close to close move divided by high -low and another statistic average absolute open to close move divided by high - low.

Iit would tell how well the strong have done about scaring out the weak during the day only to have them eating crow and wishing they had done nothing during the day, i.e. the importance of sanguinity and gravitas in market play.

Anatoly Veltman writes: 

PIVOT has been widely used for decades = (H+L+C)/3

The most popular use of it is: if the next session is trading above, then PIVOT is a support area. Conversely, if next session is trading below, then PIVOT is resistance area.

PIVOT's strength took considerable leaking with onset of 24-hour sessions, as opposed to prehistoric DAY-ONLY sessions. The reason: of course, every price traded on volume IS more meaningful than every price traded on a few lots.

Over the decades, at least two distinctive intraday set-ups where also developed, for cases of overcoming PIVOT early in the session, and for cases of overcoming PIVOT late in the session.

Also, I found Weekly, Monthly and even Yearly pivots to be useful.

In any case, despite the ease of coding the conditions for algorithmic PIVOT trades, I found that best uses of PIVOT were by layering a second indicator into the mix, and sometimes even a third. I never had the resources to code that myself — but I'm pretty sure it has been accomplished by now; including by a number of shops that I had tutored.

Gary Rogan writes: 

Not knowing any of this stuff myself, I'm curious how something this simple can work when you have quantum physicists programming ever more sophisticated algorithms and I'm sure some of which are of the learning and self-changing variety. Even the simplest control theory is orders of magnitude more complicated than this and so are rudimentary digital filters. Without giving this more than ten seconds of thinking, if I were to code up something like this I would at least do a continuously adjusting filter that would backtest the coefficients for each of the three components to something other than one while still adding up to three, variable time windows for back testing resulting in multiple variable windows rather than some fixed monthly, weekly, yearly periods, and variable coefficients for however many windows I would wind up with.





Speak your mind

2 Comments so far

  1. Bill Rafter on November 18, 2012 12:22 pm

    Collegially may I suggest that we just look at the Chair’s suggestion and not trying to change the topic to a subjectively favored indicator.

    Taking the liberty of further defining the suggestion:

    A: (abs change close X) / (range X back 1)

    B: (abs ((op X) – (close X back 1))) / (range X back 1)

    A question arises as to whether one should use the current day’s range or that of the prior day (which I chose). That can be considered subjective, although I believe the logic favors that of the prior day.

    N.B. the formula B should not be used on indices as they tend not to record opening values different from the prior day’s closing values. That is, B should be applied to traded assets and not concocted ones.

    The data on first inspection looks like static. However if you smooth it with any variety of tools it shows definite non-random activity. I tried moving averages, moving exponentials, moving spearman correlations, moving linear slopes, and moving parabolic slopes, all using various N-values. Again, let me emphatically state that it is not random, but its ability to define either the current milieu or the future has yet to be determined.

    I also tried the premise as voiced by an important personage that “chaos is when the intra-day activity does not match the inter-day activity.” That is, what is the effect of any moving correlation of the variables A and B as defined above. That also appears as non-random over many N-values, but without yet revealing any descriptive or predictive characteristics.

    Recalling earlier lessons learned, I also tried everything on a 2-day basis; that is, a 2-day change versus a 2-day range. That appeared on the surface to be more interesting, and I suggest further work in that vein for those interested.

    Bill Rafter

  2. Ed on November 19, 2012 11:42 am

    “statistic average absolute close to close move divided by high -low”

    Mr. Rafter Suggests using the prior day’s range. I suggest that using what technical analysts call the “true range” might work, as it would cover the same time period as close-to-close change it is compared too. Basically it the greater of today’s range or the difference between yesterday’s close to today’s high or low - If i remember correctly.


Resources & Links