HI @ all I have no satisfactory reply found in the forum for my problem: When I am using the advanced system tester I can have two scenarios: First: The test is fast and I get my results. Second: The test is awesome slow and take hours. I would understand that if attrebutes need more processor capacity than others. But if I am using special stops, simulation attrebutes like "simulation.currentpositionage" or maybe "prev" the simulation slows down. I am still doing my simulations overnight, because some of them are taken hours... And I am not optimizing, only taking one security. For the same chart but a bit different coding the test time drops down to less 2 minutes... Sadly I can not code everything without the "critical" attrebutes. I am wondering very much why so much time I needed, because my computer running in idle. Beside: I do have an dual 2.80 Ghz, 4BG Ram, superfast solid state drive, 64bit system... The computer is very clean and optimzed to performance... The ST_Data file size: 19MB... Where is my fault, or is it my fault that the systemtester is slower than a snail??? I am using MS 11 Realtime Pro with QuoteCenter. Thanks for your help. Scaleo

Hi Scaleo

PREV and Simulaion functions both place heavy loads on the EST. Although I haven't proved it my strong belief is that the Simulation function is even slower than PREV, and multiples of it even worse than multiple PREVs. When either of these are used it's important to limit the amount of data loaded to only that necessary for the task. Doubling the amount of data will tend to quadruple the time taken to process it. This is certainly the case with PREV and explorations, and I see no reason to expect the EST to be different.

After flirting with Simulation functions in the early days of the EST I now always resort to PREV code and never bother with Simulation. It's true that for most people Simulation is easier to use and offers more obvious flexibility than PREV, but in almost every other respect it's a disaster. At least with PREV you can code a formula and know that it will work as well in an exploration or indicator as it will in a system test.

I think as MetaStock users we have to realize that Equis has little desire and no intention of fixing problems such as those posed by PREV and Simulation - these will most likely continue to be a source of frustration for several more "updates".

To quantify your problem I'd suggest that you build two simple identical systems, one using PREVs and one using Simulations, and run some timing tests on a small portfolio. This will at least help you decide which function is creating the biggest bottleneck.

Code generally executes more slowly as more PREVs are added. I know that there are some ways of using PREV that create a bigger bottleneck (slower execution), but I also know that there are some situations where increasing the number of PREVs can actually improve execution speed. A case in point is "X+PREV-PREV" as used in the adaptive indicators of MS11, versus "X+PREV*0". The latter expression only has one PREV but code using it runs marginally slower than the two-PREV version. So the problems are not quite as straightforward as one might wish them to be.

One thing that I suggest you check is that no plots on the active chart are marked when you activate the EST. A marked plot can seriously slow down an exploration and the same might be true of a system test (I haven't checkd out that possibility). A marked chart calls the P (INDICATOR) function into use and who knows what overheads it could add to a system test.

Just a few random thoughts.

Roy

Two further questions: 1. it seems to be impossible to set a positive leverage at the short side in the adv System Tester. Is there a possebility to set the same leverage like at the long side (for example short margin 25%)? It would be very useful for CFDs, FFDs, ETFs and much more... 2. Is in the systemtester automatical positionsizing possible? Or in other words Moneymanagement, to have ever the same risk at every trade by variating the number of shares per Trade in dependency of the stoploss? E.G. long position: Number of shares = Overall Risk per Trade /(Position Open - Position stoploss) I coded something like below but nothing worked. first try: 1000/(O-ref(L,-1)) second try: 1000/(ref(O,1)-L) I hoped that this could would take effekt in the "Entire síze:" --> "Use default size" box would take effect, but I was wrong... Has someone an idea? If moneymanagement in tests would be possible, nice dynamic simulations would be necessary. Have a nice weekend Scaleo