Practical recommendations on using CandleScanner

Scanning only interesting patterns

CandleScanner recognizes 86 candlestick patterns (excluding basic candles). By default, CandleScanner uses all of them during the scanning process. You may, however, disable some patterns from the scanning process. To do so, please go to FILE » Settings » Candle Patterns.

Please note that basic candles are always included in the scanning process, i.e. they cannot be disabled as can candlestick patterns. This is due to the fact that basic candles are the building blocks for other patterns and, therefore, always need to be active during scanning.

Figure 1. In this window, you can select candlestick patterns that should be included/excluded during the scanning process.

Figure 1. In this window, you can select candlestick patterns that should be included/excluded during the scanning process.

Scanning speed and some performance recommendations

Overall, CandleScanner is very efficient when it comes to the speed of its scanning algorithms. Performed tests show that on a computer with an Intel Core i7 @ 2.70 GHz processor, the application finds on average 3,700 patterns per second. Nevertheless, it is good to follow few simple recommendations when using CandleScanner in order to make the work more efficient.

In general there are three main factors which have the biggest impact on the speed of CandleScanner:

  • Number of quotes per symbol in a given time interval (i.e. number of candlesticks per symbol on the chart)
  • Number of patterns which are looked for
  • Number of all symbols loaded into CandleScanner (across all symbol groups)

The maximum number of candlesticks (quotes) for one symbol in CandleScanner cannot exceed 500,000 (half million).

Having, however, such large data sets is not recommended due to performance issues should all quotes be plotted on the chart or used for statistics calculations. Smaller data sets should be sufficient in most cases. For example, a data set containing 100,000 quotes is equivalent to some 500 years of daily prices or 2 years of hourly prices.

The currently used time interval is important as it comes to the speed of scanning. The factor which does not always matter is the total number of quotes (candlesticks in the base interval). For example, if there is a symbol with 500,000 1-minute quotes (base interval), displayed in a daily mode (c.a. 347 days in this case), it will be faster than the ten times smaller data set made up of 50,000 1-minute quotes (base interval) displayed in 1-minute mode (c.a. 34 days). Simply, in the first case, although the a data set consists of 500,000 quotes it is scanned as if there were only 347 quotes because it is presented in daily time intervals.

The base time interval of the symbol is the minimal time interval which can be used for the given symbol. For example, having imported the quotes for EUR/USD symbol expressed in 15-minute time interval, allows one to plot the chart as a 15-minute chart, or any multiple of 15-minutes (e.g. 30-minutes, hourly, daily). With the base time interval equal to 15-minutes, you cannot however plot the chart made of 20-minutes candlesticks.

In order to speed up your scanning, you should store only the data set which is really needed and relevant to your trading. For example, if you are trading stocks using daily candlestick charts it is far better to use the data set with daily quotes rather than an intraday data set plotted to a daily chart. In this case, even a data set made up of 10,000 daily quotes allows the plotting of charts for the last 40-50 years, which is most probably even more than you need. Even so, such data sets are still relatively very small for CandleScanner, and can be quickly processed.

Another factor which has an impact on scanning speed is the number of patterns which you are looking for. CandleScanner supports 86 patterns (excluding the basic candles). But, most probably, a greater number of them are not that interesting due to their very low frequency of occurrence and hence low statistical significance. It is recommended to exclude from scanning those patterns which are not interesting. As an example of a good filter for patterns to scan, you can use the following approach:

  1. Make sure that all patterns are included in scanning process (FILE » Settings » Candle Patterns).
  2. Calculate the statistics for the whole symbol group you are going to trade (make sure to specify the right time interval).
  3. Open the statistics and check the patterns frequency. Write down the 10 most frequent patterns as the most interesting (i.e. those having a higher statistical significance than the very rare patterns).
  4. Exclude the less frequent patterns from the scanning process using Options.

From now on CandleScanner will be looking only for the 10 most frequent patterns, thus running significantly faster than when all 86 patterns are activated for the scanning process. Make sure, however, that this is undertaken at the individual symbol level as there can be considerable differences in pattern profiles.

Comments are closed.