Binary profiling of candlesticks

I am creating an EA that will use a genetic algorithm to create/test candlestick patterns. I need to classify different binary (true/false) traits that can be attributed to candlesticks - I will use these traits as the "DNA" input for multi-candle detection.

Lets think of some variables that are binary in nature and can provide (in combination with other binary variables) useful information about the nature of the candlestick.

For example:
var1: upper wick greater than body (true/false)
var2: lower wick greater than body (true/false)
var3: High[0] > High[1] (true/false)
var4: Close is further from High than from Low (true/false)
var5: trading range is greater than ATR (true/false)


If true is represented by 1 and false by 0, we end up with a "DNA" string of 1's and 0's that represent traits of (multiple) candlesticks. For example, the DNA string "01101;10010" represents two candles with 5 binary variables each.

So please help me think of good binary variables. I will post the EA once it is completed.

Save yourself alot of work and simply define the Open and Close as a Percentage of the whole Candle body. With the combination of the two you can create every candle possible.

If open is a 0% and Close at 100% it's a strong bullish candle and vice versa.
If Open = Close you have a doji. If Open = Close < 10% you have a pinbar.

The body (OPEN -to-CLOSE) going up is 1.25 * ATR
The body going up is 1.5 * ATR
The body going up is > 2 * ATR

The body going down - same thing.

[if you have to you can substitute another value for ATR, the reason I say ATR is that it will vary per pair you use it on. This is just an rough idea to get a clue about momentum.]

-------------------------------
After you have those then add the following:

The body of one of the above bars closes above or below 100 EMA
The body of one of the above bars closes above or below 200 EMA
The body of one of the above bars closes above or below 300 EMA

-----------------
the next step...

The body of candle that closes above 100 EMA and the 100 EMA is above 200 EMA and 300 EMA.

The body of candle that closes below 100 EMA and the 100 EMA is below the 200 EMA and 300 EMA

---------------

Hopefully this gives you some really good ideas.

These responses are excellent. I will implement both of your ideas; I'll define the candle itself by percentage of body and use binary variables for filters like what pip_seeker described.

I implemented TJPLD's idea. Additionally I included another variable called "percent error" which accompanies each "close percent" variable. I am working on pip_seeker's ideas.

What other variables would be interesting to profile price action?


ideas for other variables:
1. number of bars to the most recent fractal
2. slope of linear regression (positive or negative)
3. number of bars to the most recent Donichan channel breakage or other S&R level breakage (this could get complex).
4. RSI value?

You seem confused about Price Action. Just so that others don't get the wrong idea, I am going to point out that what you are talking about is not price action - just indicators. I doubt what you trying to do can be coded because it is too complex.

CodeMeister, I realize the things I mentioned in the last post are indicators but I am mainly interested specifically in price action. That is why I initially wanted a way to profile candles. I want to eventually add things like channels and triangles.