I'm gonna try that mod on BB and will let you know about the results.
(I am not skilled at all in MQL4 but that should not be so difficult according to your previous explanation).

Alain

Alain, thank you for the report!

With a little change we can make everything on place. We can calculate "Jumps" by using the current bar opening price (instead High/Low).

Example for MA Crossover entry:

Rules:
We enter long when the market price touches a MA.
- If the price was below the MA before the touch, we enter long.
- If the price was above the MA before the touch, we enter short.   

Normal conditions:
The bar opens below the MA and goes up. When it touches MA, we enter long.
Jump upward:
  The bar opens below the MA and closes 3 pips below it without touch. The next (now the current) bar opens 2 pips above MA. We have a crossover but because of the discrete nature of MA we missed the exact entry. So we enter at Open, which is the first possible price after the crossover.

The code for covering special cases such jumps and gaps is:

for (int iBar = 2; iBar < Bars; iBar++)
{   // Covers the cases when the price can pass through the MA without a signal
    double dValue   = adMA[iBar - iPrvs];     // Current value
    double dValue1  = adMA[iBar - iPrvs - 1]; // Previous value
    double dTempVal = dValue;
    if ((dValue1 > High[iBar - 1] && dValue < Open[iBar])|| // The Open price jumps above the indicator
        (dValue1 < Low[iBar - 1]  && dValue > Open[iBar])|| // The Open price jumps below the indicator
        (Close[iBar - 1] < dValue && dValue < Open[iBar])|| // Positive gap
        (Close[iBar - 1] > dValue && dValue > Open[iBar]))  // Negative gap
        dTempVal = Open[iBar];
    Component[1].Value[iBar] = dTempVal; // Entry or exit value
}

This logic has to be applied at other "entry" indicators like: Bollinger Bands, Envelop, Keltner Channel, Starc Bands...

You quoted :

- the usage of code for "Jumping" bars highly improves the backtest.
no doubt, but backtest only !!!
that could be frustrating once used in real-time trading...
- in the real trade we have O=H=L=C at the bar opening and we can apply the formula.
yes, that sounds smart ... but how robust enough is this assumption on a long-term real-time strategy
- if the price goes in the direction of the Jump, the trade and the backtest are equal.
sure, but it goes back and forth ... the price may increase fisrt then reverse ... then close in the opposite direction, thus making the "open bar" entry strategy totally irrelevant !
- if the price goes in the opposite direction, the entry error is Open - Indicator value
arithmetically true ... but so treachery. It could make the backtest very promising but then a deceitful real-time trading.

I am not a top expert in trading algorithms, but don't you think that it would be more coherent to :
1- keep your 'previous bar' analysis that is essential and arithmetically 100%-proof (compared to the META-TRADER 'current bar' analysis that is evidently wrong)
2- then keep the bar that follows for testing purposes (what's the trend, is the "previous bar" entry touched, is it a gap, is it a jump ... )
3- and finally take the position on the THIRD bar : at previous(-2) bar value or at open according to the test made in 2-

I understand that this option might lead to some "LAGGING" on position opportunities but it would be perhaps a way to have BACKTESTING get "closer" to REAL-TIME reality ....

What do you think ?

Alain

To your first question:
1. Congratulations, the noted values and variables are correct.

2. The formula for backtesting  and for real trade is the same. My ideas for that are:

a) Indicator is used for entry
- the usage of code for "Jumping" bars highly improves the backtest.
- in the real trade we have O=H=L=C at the bar opening and we can apply the formula.
- if the price goes in the direction of the Jump, the trade and the backtest are equal.
- if the price goes in the opposite direction, the entry error is Open - Indicator value

b) The indicator is used for exit
- All points are the same as the previous except the last one. If the indicator determines a SL,
the exit will be executed at first possible price, which is Bar Open. So there is no error.

In case of pending orders - the execution can be either at the indicator value or at  bar open.

So I think if we have an error, it will be several pips at entry. In the opposite case we will miss many signals in the backtest as well as in the real trade.

I was surprised when I noticed this in a MT expert 5-6 years ago. The expert was set to open a position at a Moving Average. The price was well below it. Later I noticed that the price was crossed over the MA but no entry was executed. My further observations showed that the price was jumped over the MA with one pips.

To your second question:

Sorry to bother you for I think you have a lot of work with the FST new release, but I have another question concerning the "JUMP&GAP" theory ... (see pic).

Alain

I understand the "JUMP&GAP" principle much better now.

But I believe this theory can apply ONLY to FSB BACK-TESTING on Historical Data.

How do you handle this principle with a REAL-TIME feed through FST ?

You can NOT test in REAL-TIME High[iBar] (or Low[iBar]) and THEN place an entry (stop or limit) on Open[iBar] according to the test result ....
By the time the current bar (iBar) is tested, it's already over and it becomes a PAST bar and you can no longer place an entry on Open[iBar] !

Please see the attached pic for a concrete example.

Thanks.

Alain

The second case is a positive gap. The entry value is between the previous close and the current open. That's why the program opens at the first possible price - current open.

I am a bit lost with your figures because they now show 'Currrent values' instead of 'Previous Bar value" !!!

Is it possible for you to REDRAW them all with "Previous Bar value" ?

Before you do so, would you agree with these two pictures ?

Alain

The formula for fixing "jumping prices" doesn't prevent signals. It moves the entry level to the bar's opening when we have such cases.

The code is here (for Moving Average):

for (int iBar = 2; iBar < Bars; iBar++)
{   // Covers the cases when the price can pass through the MA without a signal
    double dValue   = adMA[iBar - iPrvs];     // Current value
    double dValue1  = adMA[iBar - iPrvs - 1]; // Previous value
    double dTempVal = dValue;
    if ((dValue1 > High[iBar - 1] && dValue < Low[iBar]) || // It jumps below the current bar
        (dValue1 < Low[iBar - 1]  && dValue > High[iBar])|| // It jumps above the current bar
        (Close[iBar - 1] < dValue && dValue < Open[iBar])|| // Positive gap
        (Close[iBar - 1] > dValue && dValue > Open[iBar]))  // Negative gap
        dTempVal = Open[iBar];
    Component[1].Value[iBar] = dTempVal;
}

In normal conditions the entry is at the indicator value:
    double dValue   = adMA[iBar - iPrvs];     // Current value
    double dTempVal = dValue;

In one of the special conditions the entry is at the bar open:
    dTempVal = Open[iBar];

The conditions are:

1. The bar jumps above the indicator without touching it:
        dValue1 > High[iBar - 1] && dValue < Low[iBar]
(The previous indicator is above the previous high and current indicator is below the current low)

2. The bar jumps below the indicator:
        dValue1 < Low[iBar - 1]  && dValue > High[iBar]
(The previous indicator is below the previous low and current indicator is above the current high)

3. The indicator value is in a positive gap:
        Close[iBar - 1] < dValue && dValue < Open[iBar]

4. The indicator is in a negative gap:
        Close[iBar - 1] > dValue && dValue > Open[iBar]

All the images show indicators without "Use previous bar value" (iPrvs = 0). If  "Use previous bar value" = true, iPrvs = 1.

I am back on this topic !!!

I am now 100% convinced that 'Previous Bar Value' is DEFINITELY the right way to do things ... (I don't use anymore your custom 'Current BB' indicator).

HOWEVER, I still have a few doubts that I would like you to raise :
Look at this bar #45136.
It should normally go LONG, why didn't it do so ? (see pic)

This is not the single instance that I picked in the chart, I could show you other doubtful instances.
But as I am sure that there is a RATIONAL explanation, I prefer to show you only one example for you to explain.

Thanks.

Alain

You can compare the results between the original BB and this custom BB. You can use both and you can make  custom indicators from almost all standard codes. They are free for download from the indicators section.

Anyway I think that the standard indicators are more safe and reliable because of these modes. All the sources are publicly available.