General Indicator Miscalculations

[iocron]

Calculating the SMA20 of a dataset/series like 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30 will result in 21.4 (previous last 20.4). But the real correct result should have been 20.5, if we consider the official mathematical formula of calculating a SMA (verified by investopedia, tradingview, other indicators, ..). The same problems happens on the EMA calculation (with a much worse result) the last time I checked (this might apply to other indicators as well).

[BusinessDuck]

Described problem does not exists. May be you can get some wrong values, but is not worked as you described. We have miscalculations but it's very very very small, that the machine precision error. It's especially visible on big integer numbers. It still ignored by me. Because does not produce visible deviations for strategies. Especially if we need to have high boost for performance, its a pay for highload.

1. Let me create a unit test that will test the case that described by you.
2. I will create a branch that i can use for experiments with machine precision error, i know the good way to improve and fix that problem.

[BusinessDuck]

You can find more details here

[BusinessDuck]

I did a PR for demonstrate the calculations of SMA.

The result is 20.5. As you described as ideal.

[BusinessDuck]

Check the test here please https://github.com/debut-js/Indicators/pull/53/files

@iocron

[BusinessDuck]

I have added the test for the difference between BigFloat (floating error fixed) and with native number SMA calculations. Here is the difference for 900_000 calculations.

I create a new issue with that #54

[BusinessDuck]

I will close that issue because no activity