NihabaAshi,
Thanks for your comments.
Ok I think my point was lost in the details of what I was doing to test my hypothesis. My software is fine. I created some custom studies that allowed me to shift the start times for these candles to test the validity of the patterns. And I agree, I've never heard of anyone doing this either (wanted to see if there was an edge in it). The result of my test was that many of the supposed candlestick patterns didn't appear when the time was shifted by a minute or two. Maybe all this proves is that there is no edge in shifting start times of candlestick bars?
*****
The premise of my testing was this: if a candlestick pattern is "valid" then it should occur regardless of a minor shift in time or # of ticks. If shifting time or # of ticks by a minute or two or a few ticks is enough to make a patterns not appear, then by my reasoning, this pattern isn't valid because it depends too much on the data and charting program.
******
The 5 minute candles was probably a poor example, as in my actual testing I was more concerned with longer time candle patterns like 15, 30, 60 minutes where in my view a 1-3 minute shift shouldn't significantly alter the outcome of the pattern. I tested many combinations such as 30 minute bars starting at 9:27, 9:30 and 9:33 respectively. I found that it did alter the outcome in many cases, so my conclusion was that candles are too dependant on start/end time and start/end tick in order to be of universal applicable value. In other words, I don't feel they are robust enough to rely on candlestick patterns for predictive value.
Yes pretty much so, although it would probably be more like comparing 12,26,9 to 10,24,7 and 14,28,11 to see if I get confirming signals among similar ranges of values. The premise is the same, if I get different signals using similar values in my MACD then perhaps the signals I am getting aren't robust enough to be relied upon? Some signals will come sooner, some later but most of the signals should be confirmed by the other MACDs with similar inputs.
And you are right with your other conclusions, I may just have been testing the wrong candle patterns in the first place, and as a result those patterns weren't profitable, regardless of how I shifted the start/end times. My study was by no means comprehensive or completely conclusive and I'm still open to candlesticks holding merit for trading. This is one reason I chose to make my first post in this thread. I wanted to share my experience and then see if anyone else was in fact getting positive results from the candle patterns they are analyzing.
Let me ask this, are you using some candlestick patterns that are in and of themselves predictive or contain an edge, without applying divergence or another indicator to the pattern for confirmation?
Thanks,
Thanks for your comments.
Ok I think my point was lost in the details of what I was doing to test my hypothesis. My software is fine. I created some custom studies that allowed me to shift the start times for these candles to test the validity of the patterns. And I agree, I've never heard of anyone doing this either (wanted to see if there was an edge in it). The result of my test was that many of the supposed candlestick patterns didn't appear when the time was shifted by a minute or two. Maybe all this proves is that there is no edge in shifting start times of candlestick bars?
*****
The premise of my testing was this: if a candlestick pattern is "valid" then it should occur regardless of a minor shift in time or # of ticks. If shifting time or # of ticks by a minute or two or a few ticks is enough to make a patterns not appear, then by my reasoning, this pattern isn't valid because it depends too much on the data and charting program.
******
The 5 minute candles was probably a poor example, as in my actual testing I was more concerned with longer time candle patterns like 15, 30, 60 minutes where in my view a 1-3 minute shift shouldn't significantly alter the outcome of the pattern. I tested many combinations such as 30 minute bars starting at 9:27, 9:30 and 9:33 respectively. I found that it did alter the outcome in many cases, so my conclusion was that candles are too dependant on start/end time and start/end tick in order to be of universal applicable value. In other words, I don't feel they are robust enough to rely on candlestick patterns for predictive value.
Yes pretty much so, although it would probably be more like comparing 12,26,9 to 10,24,7 and 14,28,11 to see if I get confirming signals among similar ranges of values. The premise is the same, if I get different signals using similar values in my MACD then perhaps the signals I am getting aren't robust enough to be relied upon? Some signals will come sooner, some later but most of the signals should be confirmed by the other MACDs with similar inputs.
And you are right with your other conclusions, I may just have been testing the wrong candle patterns in the first place, and as a result those patterns weren't profitable, regardless of how I shifted the start/end times. My study was by no means comprehensive or completely conclusive and I'm still open to candlesticks holding merit for trading. This is one reason I chose to make my first post in this thread. I wanted to share my experience and then see if anyone else was in fact getting positive results from the candle patterns they are analyzing.
Let me ask this, are you using some candlestick patterns that are in and of themselves predictive or contain an edge, without applying divergence or another indicator to the pattern for confirmation?
Thanks,
BTW, I'm sure there have been traders that have shifted bar data initialization forward a minute or two to see if they could possibly see a particular pattern print before the rest of us to have an edge. This might just work! :eek: