Error in Calculations

[xbaha]

hi, i am trying simple print:

hr=High[0]-Low[0];
Print(hr);

while the stock is a 2 fraction stock
i am getting output like this:

0.0800000000000125
0.0699999999999932
0.0499999999999829
0.0600000000000023
0.0999999999999943
0.0500000000000114
0.0800000000000125
0.0500000000000114
0.0600000000000023
0.039999999999992
0.0600000000000023
0.0800000000000125

Why???
if the h[0]=12.1
and l[0]=12
i must get aprinted value of 0.1
not 0.099999999999999

why is this annoying error??

[NinjaTrader_Bertrand]

You're working with true double values here, so floating point arithmetic issues need to be considered as explained here - http://www.ninjatrader.com/support/f...ead.php?t=3929

[xbaha]

You're working with true double values here, so floating point arithmetic issues need to be considered as explained here - http://www.ninjatrader.com/support/f...ead.php?t=3929

you referred to a thread about comparing double to double.
but what if i have more complex calculations?? like i am converting my system from TS to NT, i am getting totally different signals because of this small difference, why? because i keep doing more & more complex calculations until the double variable have a big difference from its theoretical value...
any other suggestion?

[NinjaTrader_Bertrand]

Correct working strongly with doubles and the precision they offer can result in 'blowing up' calculations by not properly debugging scenarios created by the code. You could for example use less precision (it's seldom needed in the final results) which would also use less memory then (floats for example).

[koganam]

you referred to a thread about comparing double to double.
but what if i have more complex calculations?? like i am converting my system from TS to NT, i am getting totally different signals because of this small difference, why? because i keep doing more & more complex calculations until the double variable have a big difference from its theoretical value...
any other suggestion?

If you are working with price data, just round all your results to TickSize, and that should handle it. Yes, that means intermediate results too. Calculating anything to 15 decimal points when the input data is to 2 decimal points does not increase precision. Actually, it could very well decrease it, as the so-called extra precision is really noise. And that even seems to be what you are seeing in practical terms.