|
PureBytes Links
Trading Reference Links
|
OK, Robert I'll take on this one and post to te list because the first request was public.
More,it can be of interest to the readers of the list.
Was ( Re: "DON'T WORRY - BE {{{ CONFIDENT}}} CRAPPY")
----- Message d'origine -----
De : Robert W Cummings <robert.cummings@xxxxxxxxxxxxxxxx>
À : pierre.orphelin <pierre.orphelin@xxxxxxxxxx>
Cc : Robert W Cummings <robert.cummings@xxxxxxxxxxxxxxxx>
Envoyé : lundi 13 septembre 1999 00:37
Objet : Re: "DON'T WORRY - BE {{{ CONFIDENT}}} CRAPPY"
> Let me try and put it another way on the stochastic indicator. I wanted to
> know the thinking around using 100 multiple using an average of lowest low,
> highest high, and average close then he calls it fast D, why the D why 100.
> Then on K he uses a .5 of D.
First, the 100 factor is a scaling factor because people may better understand when using percent
values.You may remove the 100 factor, the stochastic will have the same shape.
It's only for convenience.
Stochastics came from the raw indicator known as RawK ( PercentR in TradeStation).
I'll explain the names after.
They are only smoothed values of th K indicator , for a better reading ( average remove the spikes,
at the cost of lag).
What is K and why K ?
The K indicator and its avatar are probably the best oscillator ever devised.
First, K is nothing that the representation of the position of the last close compared
to the highest HH and the lowest LL of a given lookback period.
== == HH========================100%
__________________________________ Close (K=85 %)
== ==============LL============= 0%
It can be seen as a normalization of the price values on a scale that is 0 for LL
and 100 for HH.
So C is of course in between and moves inside an relative interval of 100 %
But, as we calculate the HH and LL value on a sliding window, the 0 -100 scale
is always changing whn comparted to price move.
This has one advantage ans on inconvenience:
The bad new is that a 0 -100 reading has no meaning in itself over the long run.
You may observe huge move of the K oscillator when the price stay in a narrow range.
The good new is that there is a sie effect ( a window effect):
When volatility occurs, the HH-LL value is modified and the K value goes quickly
to the extremes, because the price C moves, but also the HH or LL value that will go into the same direction
In this sense, the stochastics acts as an oscillator with a booster effect when the trend starts ( HH or LL mobe acts like a momentum overlaid to the oscillator
For a downtrend starting:
== == HH=cancelled=======================.
******************************newHH = 100
*****************************************newHH =100%
__________________________________ Close quickly moves to LL
= ==============LL======================= 0%
Once established
Th close stick to the descendinc new low LL and prodces a relaible 0%, indicating a downtrend
******************************newHH
*****************************************newHH
= ============================C=LL 0%
= =============================C==LL 0%
= ==============================C==LL 0%
The sensitivity is very high, so it is necessary to smooth the K oscillator.
The first smoother is a 3 bar exponential like moving average applied to K.
The name is FastD ( or slowK,or %K)
Then you may smooth again D to get D % ( or Slow D or %D)
Extremes readings are good trend indicators for the stochastics.
The interpretation is far different when the volatiliy is low or high.
If you want to devise a stochastic system, you need different rules
according to volatility or trend strength (both are the same.)
Now, to find the rules, you can spend a lot of time on it, and they are not in a book
Suppose that you download the evaluation of the Safir -X software.
( I wrote " suppose")
The examples provided use only 2 stochastics of 10 and 20 bars.
You will see that by building complex rules with a very small set of indicators that make sense,
provided that the rules are well written and precise.
In this case, 300 to 500 rules are generated to interpret the 2 stochastics.
Ans the system works well after on several dozens of thousands of unseen bars, different markets.
>Where did the K come from and why .5 of D just
> to have it move a little faster than D?
According to George Lane, he named his trials from the alphabet.
The K oscillator was his trial number 11, so the name was K
Same for D ( 4 th trial for smoothing).
Do not forget that Lane's stochatics were calculated by hand and is work
if far before Welles Wilder (who had a small HP palmtop calculator in 1978)
> Then he uses the same variations
> with percent d and percent k when describing slow by grouping an average of
> the same.
The other advantage of comparing %D and %K is that you perform moving average analysis on
not the price, but a pseudo normalized value of the price that , for the reason ofthe side effect explained above,
can move faster than the price when the trend starts.
So, the lag of the moving average is compensated by the %K speed, giving clean signals with less noise and les lad than any other simple method ( providend that the price serie is not too much noisy).
This is of course not true for high low readings of the stochastics, but mainly this is the reason of the averaging technique.
It shoud be clear that the key is a different interpretation according to external condition.
This is the case for all classical indicators, that are not so stupid.
What is stupid is what you read in the books and the poor 2 - 8 rules that our brain is only able to conceive and apply.
This is the reason why we have spent so much time on Safir-X, because we will never be able to beat its complexity.
You, maybe, but not me...
As an expert PO what tells you when to use these math equations
> when plotting momentum, volume, trend, price, volatility. I mean you have
> no energy here only a tick count then from this all other indicators evolve
> by using math,trig etc. The most obvious is the MA with time based on
> history that is then recycled using more history and we use that to trade
> the future.
> Just wondering how you looked at this after all your experience doing it.
>
I'm not an expert,and I do not believe in prediction power of technical analysis.
Let be a follower, but very skilled, and it's enough to make a living
My definitive opinion is that technical analysis work with classical indicator,
but the very good solutions are not available to human being with their naked brain.
I cannot tell you when to plot a momentum.
I have a vague idea like i explained above for the stochastics, but I'm unable
to explain the exact rules because I do not know them.
The best progress that I have mad was to understand where were ( hi hi) my limitations.
So, I stopped to code trading systems by hand
( I have found some interesting, but too much time consuming),
but I do not want to do this anymore because
I do no tbelieve that the ( my) potential is here.
I have found my solution(s) by 100 % computer mean
You can too, there are an infinite number of possibilities.
I of course speak as a systematic trader.
For visual use, I'm not very good at it,a nd I'm not interested to see the charts.
Indicators have been removed from the TradeStation display for years now.
They are integrated as the initial part of the computerized decision.
For the rest , I wait for the active order window.
Sincerely,
Pierre Orphelin
www.sirtrade.com
|