how important is start time for time frame bars

I've read lots from various sources that goes along the lines of "wait for the bar to close over the XXX indicator", or another example is candlestick patterns that depend on the O-H-L-C values for the shape of the candle.
However, I think it is possible to change the shape of the candles, or candle patterns, or whether the close/open is around an indicator, by changing when the interval starts.
In the attached example, it is an hourly chart; usually charts start the bar on the hour, and close it before the next hour. For my study, I've extracted hourly bars from 5-minute data, starting on the hour (ex, 1:00, 2:00, 3:00 ...). For the comparison, I start the bar at the half hour (ex, 1:30, 2:30, 3:30 ...).

The 2 look generally close but with some differences. I marked 3 examples that stood out to me with blue arrows and labeled as 1, 2, 3. (I think they match up as the corresponding bars.)

In case 1, the On-the-Hour shows a tall wick and a short body. But the On-the-Half-Hour, these are 2 bars with long bodies, and no wick. Additionally, if you were looking for a close over an indicator, I'd guess the On-the-Hour would probably not signal, whereas the On-the-Half-Hour would.

Are these types of differences significant? Or do they somehow balance out with a large enough data set? I suppose they would affect pattern recognition or technical signals, the patterns or signals could change based on when the summarizing bar starts and ends. I suppose this is why they say draw trend lines or patterns based on highs and lows instead of closes, they seem to have consistent values between the 2 interpolations.

 

here's the image -- didn't get it on first post.

 

let's see if it will show the picture in the post:



ok, this might make it easier to view on the thread

 

looks like the extreme highs and low remain unchanged, so what is the issue again?

 

Interesting study. Thank. Raises some issues.

 

I don't really see the issue here.

 

Thanks -- that offers support for my guess, that highs and lows are much more important than closes for chart patterns. The higher time frame summarizes the data, and does not tell the path within it. I think it was Redneck, in another thread, who said the ability to see how the lower time frame would form the higher time frame bars was a good ability.

Seeing how the closes can change widely (like in case 1 and case 3), I'm going to take technical indicators based on the closing price with a bit more caution.

 

===========
Agree, Double N;
especially since closing prices for 1, 2 days does not mean much@ all.Even though a day or 2[closing prices , in OCT 87 is worthy of note.LOL

Closing [pro] prices are however much more important than open [emotional in many cases]open/start prices.

Hi in uptrends[especially weekly, mo...];
low in downtrends[especially weekly , mo...]mean much .

Hope this helps,Double N;
helped me for many years

 

thanks Mr Turtle ... and other above posters who already knew this ... yes, it all makes sense now, now I understand my nagging suspicion indicators were missing something, Donchian channels were best, and all my strategies tend towards randomness over time ...

Highs and lows are the only values that will be consistent, opens and closes can vary based on the time sampling. Closes at the standard time may have some use because so many other people watch them. Donchian channels are somewhat useful because they're based on highs and lows which are mostly constant values. Indicators based on the closing price can vary, arbitrarily capturing the phase of price action that just happens to be occurring at the end of the bar. If the indicator value can vary, it probably will not be reliable or produce reliable results over time. I could research this, but probably won't due to time and effort, instead consider it a safe assumption.

I'll have to look thru indicators for more that are based on highs and lows. Then, change the ones based on closes to highs and lows. Lookback periods are probably useless too, I'll have to do some research to investigate that.

Conclusion (with noted assumption above): Time should not be the x-dimension, the sampling sizer, it is not the independent variable that matters to price. I should probably review my research and develop my trading approach over again from the beginning in view of this.

 

===============
Like the Don channels also;
All the weekly data, all monthly data, all daily data...............................