|
PureBytes Links
Trading Reference Links
|
Didn't need to run your experiement as I'd figured this outsome time ago.
The problem is, with time bars you are arbitrairily dividing up the day
into fix period intervals that have no meaning in the underlying market.
There's nothing that says a reversal is going to take 3 minutes or 5
minutes or whatever bar period you choose or that it will start right at
11:05 since that's when your bar starts. It starts when it starts and
the only way you'll catch it if you are setup right for that reversal.
Using tick bars is an improvment as it divides the tick stream up into
fixed length units independent of time, but dependent on the amount of
activity. But the behavior of these bars will depend in part on your
feed as well so if your feed has problems, you loose.
Kase's Universal Bars (I think that's what she calls them) are an
interesting approach in that they ignore both time and tick volume and
focus on defining bars by the amount of price movement (similar to Point
and Figure charts). Of course you can't do them in TS today, maybe TS 5?
Another approach is to not use bars at all but use tick charts to
identify chart formations. This could be an interesting approach but
there is a lot of work in filtering out the noise at this level.
If you want to use candlestick patterns, stick to day or week charts.
Jerry
>I was having trouble understanding why candlesticks work at all. That is,
>there are various sources for lists of candlestick patterns that are often
>times reversal patterns, but memorizing the list is a different thing than
>understanding how it truly is that these patterns are not mere accidents.
>Perhaps the value of candlesticks for indicating potential swing high/lows
>is intuitively clear to you right from the texts. It wasn't clear to me.
>
>Here's an exercise has helped me to appreciate the value of candlesticks for
>indicating potential reversals at a given timeframe.
>
>1 Browse 5 or 10 minute S&P day session bars until you identify a VERY
>SIGNIFICANT swing high/low bar that ALSO has a clear candlestick reversal
>pattern per the standard candlestick texts. (e.g., shooting star, bullish
>engulfing, etc.)
>2 Center the chart on that bar and create a .gif file of that chart for
>later study.
>3 Change the start time for the S&P day session by adding 1 minute (to
>8:31 for the futures). (You can do this without going into the data server
>by formatting the price data and clicking the Universe button on the
>Settings tab.)
>4 Center the new diagram on the same reversal point you identified in
>step 1. (Notice that the look of the reversal bar set has changed!) Create
>a .gif file of this chart.
>5 Repeat steps 3-4 until you have increased the S&P day session start
>time 5 or 10 times depending on what bar period you selected in step 1.
>
>Print and compare all the .gif files you've produced.
>
>Conclusions:
>
>A I found in this exercise that SIGNIFICANT reversal swing high/lows
>ALWAYS had at least one or two or more clear candlestick reversal patterns
>associated with them within the complete set of .gif files produced for that
>reversal per the above steps.
>
>B It appears to me that if any set of bars that contain a reversal don't
>exhibit a candlestick pattern this is due to the timing of the reversal
>high/low within the bar period and the magnitude of the reversal.
>
>Questions:
>
>- I'd like to confirm that those folks doing the exercise above come to
>the same conclusions that I have. Specifically, after doing this exercise,
>do you agree with both conclusions A and B above? If you don't, why not?
>If you don't, what specific reversal date/time led you to your different
>conclusion?
>
>- I'd like to formulate a paragraph or two that summarizes these findings
>in a systematic manner. I've got the beginnings of thoughts for these
>paragraphs and will share them once complete. But I'd like to get other
>folks' thoughts about what these paragraphs should contain so as to increase
>the breadth of the insight these paragraphs will contain.
>
>Thanks in advance....
>
>Steven Buss
>sbuss@xxxxxxxxxxx
>Walnut Creek, CA, USA
>
>
|