ig-dev
commented
6 years ago Let me given an example to show that this library cannot even calculate an Exponential Moving Average - one of the most basic averages.
import com.tictactec.ta.lib.Core;
import com.tictactec.ta.lib.MInteger;
import java.util.Arrays;
public class Example {
private static final int LOOKBACK = 3;
private static final double EXTRA_VALUE = 10;
private static Core core = new Core();
private static MInteger outBeginIndex = new MInteger();
private static MInteger outLength = new MInteger();
private static double[] values;
private static double[] result;
public static void main(String[] args) {
values = new double[] { 1, 2, 3, 4, 5 };
result = new double[values.length];
core.ema(0, values.length - 1, values, 3, outBeginIndex, outLength, result);
System.out.println(Arrays.toString(result));
values = new double[] { EXTRA_VALUE, 1, 2, 3, 4, 5 };
result = new double[values.length];
core.ema(0, values.length - 1, values, 3, outBeginIndex, outLength, result);
System.out.println(Arrays.toString(result));
}
}This prints:
[2.0, 3.0, 4.0, 0.0, 0.0]
[4.333333333333333, 3.6666666666666665, 3.833333333333333, 4.416666666666666, 0.0, 0.0]In the first result we take the 3-period EMA over 5, 4, 3, 2, 1 and get [2.0, 3.0, 4.0, 0.0, 0.0], or rather
[0.0, 0.0, 4.0, 3.0, 2.0 ] , reversed, because TA-Lib iterates from the end to the beginning. The first two values are zero because we cannot make a 3-period average using only 2 values. Then the 4 is the result of something to the effect of (5+4+3) / 3 etc. So far so good.
Now we take the EMA again, but this time we prepend a 10 value to the list. We get:
[4.333333333333333, 3.6666666666666665, 3.833333333333333, 4.416666666666666, 0.0, 0.0]
Now the first two values are still zero, but then all values suddenly change. Since the first entry in the list represents the newest entry, only the very last, new value should have changed.
Now if you think that TA-Lib may iterate the other way, simply replace ema with sma. If we do this we get the following result:
[2.0, 3.0, 4.0, 0.0, 0.0]
[4.333333333333333, 2.0, 3.0, 4.0, 0.0, 0.0]Only one new value 4.333333333333333 and the rest stays the same, as we expected. This is not the behaviour we see for ema. Either the EMA is computed wrong, or the SMA, but no matter how much you twist it, the library can either not compute a simple SMA or EMA. And all the other indicators, such as Bollinger Bands, MACD, and others that depend on these averages, are also wrong. So are the stochastic indicators, and the list goes on and on and on.
This is quite significant because there is a ton of Java-based projects that use this library here. It's the first go-to library for a lot of projects and trading interfaces. I'm surprised nobody else noticed or made an issue.
sebbie
yul
shevkoplyas
Edit: This was previously a false warning to the community not to use this library - I apologise for the false alarm because of my own unflattering misunderstanding. Turns out that I had made wrong assumptions about how the library processes input and returns values.
As long as you understand that
You should be fine :slightly_smiling_face: