Alan please be more precise and write in numbers what the entry price is and what entry price you expect.

According to your chart the error is less then a pip.

I'm sure there is no error in the calculations.

No. You don't need crossing of the current band.

Listen.

Bollinger Band is based on the closing price right? UpBB = MA(Close, 20) + 2* Std. Dev.

Imagine the following case:

Trading on a daily chart at the time is 2pm. MetaTrader plots the current bar and announces that the current price (at 2pm) is the closing price. Respectively it plots also the B Bands. But this is the closing price and this is not the B Band. These are some Current price and bands. They will change till the end of the bar. The platforms shows this values but they are not Final. You and every one who pretends for correct backtesting cannot use the current values. Doing so you'll receive wrong backtest results. To prevent this FST automatically decides whether  to use current values or previous values for backtesting.

The cases are:
- When the indicator is based on Close and the position opens on price different from Bar Open, you have to use the previous values.

- When the indicator is based on Close and position opens at Bar Close - current value.

- Indicator based on Open - current value.

Your case is the first one. The opening doesn't depend on the current B Band. The logic rule "Enter long at the upper Band" together with "Use previous bar value" means that you have to open long when the market price reaches the previous UpBand. You have nothing to deal with the current BB. You even don't know how much is it in the real live. The entry is at the indicator value plotted at the middle of the previous bar.

When you see an entry arrow, look at that level one bar ahead. That's the entry indicator value is. The program rises an entry signal when the price reaches the entry value +- half pips.
The prices are normalized from the broker to 4th  or 5th digits. The indicators are calculated with full precision. The position price can be fractional because of averaging.

These differences of up to half pips after zooming looks like gaps in the chart. That's the reason all the information is shown in the right side info bar in digits. You can see there the Current indicator value, the entry value and the position's price.

If I'm not enough clear, please write in your posts the values or when you make screen shots put the cross at the desired bar.  Make the screen shot showing the info bar like this:

Alain this is a special case when the price jumps over an indicator without touching it. Fortunately Forex Strategy Builder and Forex Strategy Trader manage this situation properly by opening position at Bar Open price.

See the attached pictures.

Without this features the signal would be missed.

Dear Alain,
I made a special BB for you. It is without "Use previous bar value" and "price jumps".
It has to be easier for understanding.

Put it in "Custom Indicators".

This custom indicator is for testing the different logic rules. The entry points will be much clear because they are exactly on the BB bands. This indicator will act exactly like the MetaTrader BB indicator. 90% of the experts I've seen rise signals at the current indicator's value and do not use modifications for "jumping" prices.

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.

There is HUGE difference in the reliability of back test.

There is nothing special in this bar. There is no entry there because the bar doesn't touch the entry price.

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.

In the first case we have a jump.

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.

Excellent. Three notes only.

Popov,

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

Popov,

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