Prior to this latest desktop client update (2.22.2) visually back testing a strategy across 1 month of GU data at maximum speed (=x8, slider fully across to the right) used to take 3-4 minutes.

Now visually back testing the same strategy across the same period takes 18-24 HOURS! This around 400-500 times slower.

The speed of a back test without visual ticked appears to be the same as it was previously.

We tested it on GBPUSD all ticks of the last month and the VM on. It took a couple of minutes. Most probably it is related to somehting specific that your strategy is doing. Could you please provide the source code or piece of it that would show the problem?

I doubt it is the strategy - it was running as normal (fast) speed prior to the last API/ddesktop client update, and then very slowly afterwards, and back test without visual mode on is still as faster as it used to be.

Hi,

i'm having the exact same problem. Strategies that used to be fine are now incredibly slow to backtest.

I was able to reproduce it using the following "shell" strategy. From what i see, it seems either of those two lines make the backtest go nuts:

            context.getChart(selectedInstrument).addIndicator(indicators.getIndicator("HT_SINE"));
            context.getChart(selectedInstrument).addIndicator(indicators.getIndicator("HT_TRENDMODE"));

I'm using the following java version:

Java Web Start 10.10.2.18
Using JRE version 1.7.0_10-b18 Java HotSpot(TM) 64-Bit Server VM

If you need additional information, feel free to contact me.

Alex.

Hi,

i'm having the exact same problem. Strategies that used to be fine are now incredibly slow to backtest.

I was able to reproduce it using the following "shell" strategy. From what i see, it seems either of those two lines make the backtest go nuts:

            context.getChart(selectedInstrument).addIndicator(indicators.getIndicator("HT_SINE"));
            context.getChart(selectedInstrument).addIndicator(indicators.getIndicator("HT_TRENDMODE"));

I'm using the following java version:

Java Web Start 10.10.2.18
Using JRE version 1.7.0_10-b18 Java HotSpot(TM) 64-Bit Server VM

If you need additional information, feel free to contact me.

Alex.

Forgot to attach the strategy. Here it is.

We are investigating this issue.

When testing in the visual mode, indicators on chart are calculated on every tick, which is around 5000 times for each Hourly bar.
Hilbert Transform are the "heaviest" TA-Lib indicators with a large lookback.
Testing a strategy in visual mode with these indicators attached takes roughly the same time as calling their calculation on every tick from strategy with visual mode turned off.
The best solution for the moment is: add these indicators on chart while testing only when necessary or use the "recalculate on new candle only" option.

Yes, but previously this didn't happen - I have quite a few strategies which are indicator intensive, and they have all slowed down by 400 or 500 times during visual backtesting x8 playback as noted above. Some new issue has been introduced.

Please help, tell me how to do it in the code of the strategy??? indicator is added with the following code...

chart.addIndicator(indicators.getIndicator("SMA"), new Object[] { smaSignalTimePeriod }, new Color[]{new Color(89,144,32)},new DrawingStyle[]{DrawingStyle.LINE}, new int[]{1,2});
chart.addIndicator(indicators.getIndicator("MACD"), new Object[]{fastMACDPeriod, slowMACDPeriod, signalMACDPeriod});

IIndicator macd = indicators.getIndicator("MACD");

macd.getIndicatorInfo().setRecalculateOnNewCandleOnly(true);

chart.addIndicator(macd, new Object[] {fastMACDPeriod, slowMACDPeriod, signalMACDPeriod});