Dangers of Tick Simulation on lower timeframes
10件の返信
Just a quick reminder of why tick simulation is uses on lower timeframes.
Here is a strategy build using M1 bars from real tick data, using tick simulation, spread fixed.
Here is the exact same strategy, exact same fixed spread, only with real tick precision. Note it’s the same data as the M1 above, only the precision is real tick, not simulated this time. Same spread.
Somewhat different.
So, remember, if you are on the lower timeframes, you need real tick and not tick simulation.
0
Any success with m1 strategies?
0
Any success with m1 strategies?
No because to find them you really need to build on real tick with real tick precision. In SQ3 the real tick data precision is read constantly from disk it is not held in memory. Even with a fast SSD it is very slow. If you build using tick simulation the results are meaningless at this timeframe.
0
Using virtual hard disk as a solution.
RamDisk Benchmark
http://fiehnlab.ucdavis.edu/staff/kind/Collector/Benchmark/RamDisk/ramdisk-benchmarks.pdf
http://winramtech.hostei.com/RAMDriv/
0
Using virtual hard disk as a solution.
RamDisk Benchmark
http://fiehnlab.ucdavis.edu/staff/kind/Collector/Benchmark/RamDisk/ramdisk-benchmarks.pdf
You would need a lot of RAM to hold the tick data in that way.
0
You would need a lot of RAM to hold the tick data in that way.
Only as much as the tick file, correct? or much more?
0
Only as much as the tick file, correct? or much more?
Enough to hold the .dat file plus some. Actually looking at the files (the .dat files in SQ3) it might work. They are not as big as I imagined. About 4gb per file.
0
With any standard server or gaming PC/higher end laptop it should work.
0
With any standard server or gaming PC/higher end laptop it should work.
Problem is I think SQ3 expects the .dat files to be in bin folder under installation folder. Would have to put all the SQ3 into RAM disk, unless there is a setting for SQ3 that can be used to tell SQ3 that a particular symbol is on a different (RAM) drive to main installation.
0
Have you tried building with M1 resolution bars (not sim) then retest on tick your group of strategies produced? With MT4 I optimize on bar open then retest on real tick. Some suck and the results are way off, but after going through a few of them I can usually find one that’s good. MT4 tick takes about 45 days to optimize my EA Wizard strategies on vs less than a day with bar open so this saves a lot of time. I think that would be better all around because you’ll be able to throw a lot more variations at the data in a shorter time span then all you need to do is just retest and pick the ones that actually pass on real tick. This means you won’t even have to do a RAM virtual drive sim.
0
Problem at this timeframe, is you will find 1000 fantastic strategies using tick simulation, spend ages retesting on real tick, to find all of them are worthless.
0
10件の返信を表示中 - 1 - 10件目 (全10件中)