And here, even before going into a heavy duty backtesting analysis, we can check some simple statistics first. As a first pass, there’s no need for detailed model optimization, no need for sophisticated entry/exit methodology and no need for risk and money management rules, just a basic analysis of the underlying statistics. What matters at this stage is: the number of accurate occurrences (how statistically significant the results are) and the price behavior within 5 or 10 subsequent candlesticks after the pattern’s occurrence.
The most challenging aspect is to effectively convey the statistics and present them in such a way readily showing pattern efficiency, just after its occurrence. CandleScanner adopted very simple principle. If there is a bullish reversal or a bullish continuation pattern on the chart, the next 5 or 10 candles (depending on the user's preferences) should demonstrate its prognostic value.
Stop Loss order is used to have more realistic results. If Stop Loss level is reached, the algorithm stops and the so far the most extreme price is used to calculate efficiency level.
The user can set Stop Loss level, and ranges for FALSE, LOW, MEDIUM and HIGH efficiency. These settings may have a great impact on the efficiency readings.
Here is the summary of how the algorithm assessing a pattern's efficiency is implemented.
The algorithm evaluating the efficiency of a bullish pattern occurrence works as follows:
- If Stop Loss price level is reached on the first candle following the pattern, the occurrence is marked as FALSE.
- Is Stop Loss price level is reached within test period, the highest price reached prior to Stop Loss is used to assign the pattern occurrence to appropriate efficiency level (i.e. FALSE, LOW, MEDIUM or HIGH).
- If Stop Loss price level is not reached within test period, the highest price reached during the full period is used to assign the pattern occurrence to appropriate efficiency level (i.e. FALSE, LOW, MEDIUM or HIGH).
The algorithm for every pattern occurrence is run for the period of 5 and 10 candles, producing two separate results.
If there are not enough candles following pattern occurrence, the result is marked as NO DATA.
The algorithm works the same for bearish patterns, except that instead of using the highest price reached during the test period, the lowest one is used.
More info about CandleScanner: