A very simple question for programmers

[pstrusi]

I apologize for my programming ignorance but I'm stuck in something that it shouldn't happen

I'm testing a very simple condition:

if ( (Close[0]-(Close[0]+Open[0])/2) < -0.000175)
{
EnterShort(100000, "SELL");

}

I work with Range bars, onBarUpdate=true....ect etc

The problem comes whenever I set "==" instead of ", or >" , there's not any result, when I've previously checked that actually there are a lot of results with that condition...so when I use < or > no problem, but if I use == I got NONE results....

how is this possible?
How could I fix it?

Thanks in advance

[koganam]

I apologize for my programming ignorance but I'm stuck in something that it shouldn't happen

I'm testing a very simple condition:

if ( (Close[0]-(Close[0]+Open[0])/2) < -0.000175)
{
EnterShort(100000, "SELL");

}

I work with Range bars, onBarUpdate=true....ect etc

The problem comes whenever I set "==" instead of ", or >" , there's not any result, when I've previously checked that actually there are a lot of results with that condition...so when I use < or > no problem, but if I use == I got NONE results....

how is this possible?
How could I fix it?

Thanks in advance

Try rounding your calculation off; either mathematically, or to tickSize. ref: http://www.ninjatrader.com/support/h...trument_ro.htm

[sledge]

Turn on output window from tools menu item.

Then use a print statement before your if condition.

Print("result="+(Close[0]-(Close[0]+Open[0])/2));

You will see if you ever get -0.000175.

[pstrusi]

Thanks Koganan but I'd need to work with those values, so what others ways are there?

[koganam]

Thanks Koganan but I'd need to work with those values, so what others ways are there?

I do not understand what you just wrote. Use any values you want; just round off before/when you do the comparison. Comparison of doubles requires exactitude. IOW, leave the values intact; just round off the result of your calculation at the time that you do the comparison. Your values remain. So what is the problem again?

[pstrusi]

I meant those format values ( 6 digits after 0 ), I didn't understand how the round off will work out with it

Thanks Koganan, I'll try it that way; this is embarrassing cause it should be pretty easy to solve

[koganam]

I meant those format values ( 6 digits after 0 ), I didn't understand how the round off will work out with it

Thanks Koganan, I'll try it that way; this is embarrassing cause it should be pretty easy to solve

 
double ComparisonValue = Close[0]-(Close[0]+Open[0])/2;
if (Instrument.MasterInstrument.Round2TickSize(ComparisonValue) == 0.000175)
{
//MakeMegaBucks();
}

That, of course assumes that your tickSize accuracy matches the value that you are comparing. I do not know any instrument that has a tickSize that goes to 6 digits. Therefore, given that you are using Close and Open, which are both exactly sized to tickSize, there is only one very limited exact condition under which your equation will be satisfied.

It seems to me that you may be better off multiplying both sides by a suitable factor, such as 1000, so that you are comparing values of like accuracy, or else mathematically rounding to 6 decimal places, which will allow a match if, and only if, the instrument has a tickSize to 5 decimal places, and the range of the candle is exactly 0.00035.

Edit: As a matter of fact, writing

 
if ((Close[0] - Open[0]) == 0.00035) ...

might be the better condition to write, as on an instrument with a 5 decimal place tickSize, that equation is already exactly matched to ticSize, and so avoids all rounding errors.

[pstrusi]

Koganan in my backtesting I'd need to compare 2 values like differences between average, therefore I have two questions in order to find the best solution:

1. The best way could be using for example:
if (Instrument.MasterInstrument.Compare(Price1, Price2) == 1)

2. Or can I format double values to a fix format, for example: round to 0.000000 format but not necessarily divisible exactly by ticksize ?

Thanks for all inputs my friends

[sledge]

 
double ComparisonValue = Close[0]-(Close[0]+Open[0])/2;
if (Instrument.MasterInstrument.Round2TickSize(ComparisonValue) == 0.000175)
{
//MakeMegaBucks();
}

This isn't the first time Koganam has left out the MakeMegaBucks function.

One day there will be a slip up.

[pstrusi]

 
double ComparisonValue = Close[0]-(Close[0]+Open[0])/2;
if (Instrument.MasterInstrument.Round2TickSize(ComparisonValue) == 0.000175)
{
//MakeMegaBucks();
}

That, of course assumes that your tickSize accuracy matches the value that you are comparing. I do not know any instrument that has a tickSize that goes to 6 digits. Therefore, given that you are using Close and Open, which are both exactly sized to tickSize, there is only one very limited exact condition under which your equation will be satisfied.

It seems to me that you may be better off multiplying both sides by a suitable factor, such as 1000, so that you are comparing values of like accuracy, or else mathematically rounding to 6 decimal places, which will allow a match if, and only if, the instrument has a tickSize to 5 decimal places, and the range of the candle is exactly 0.00035.

Edit: As a matter of fact, writing

 
if ((Close[0] - Open[0]) == 0.00035) ...

might be the better condition to write, as on an instrument with a 5 decimal place tickSize, that equation is already exactly matched to ticSize, and so avoids all rounding errors.

How could I round to 6 decimal places?

[koganam]

Koganan in my backtesting I'd need to compare 2 values like differences between average, therefore I have two questions in order to find the best solution:

1. The best way could be using for example:
if (Instrument.MasterInstrument.Compare(Price1, Price2) == 1)

2. Or can I format double values to a fix format, for example: round to 0.000000 format but not necessarily divisible exactly by ticksize ?

Thanks for all inputs my friends

To start off with, anything that attempts to calculate an output to a greater accuracy than its inputs is mathematically meaningless. Hence my use of Round2TickSize().

That mathematical puritanism aside, yes you can round to any value you want. There is the C# Math.Round() function. ref: http://msdn.microsoft.com/en-us/library/75ks3aby.aspx

[pstrusi]

Thanks a lot for your help. best regards