Hi I am using a similar strategy with different inputs on EURUSD and @YM on TS, however the optimisation speed of the @YM is much slower. I am based in Ireland anyone and idea why there is so much difference in speed. Is there a way to actually save historical data to PC. Or has anyone any idea regarding best way to run optimisation.
what are the chart settings for bar type and interval? Are you using look inside bar backtesting? If so, what is the resolution?
My guess is that it has simply something to do with the number of ticks processed during the optimization being different for each symbol. Therefore the questions above help to clarify this.
Thanks for your reply ABCTG. However the strange thing is I have each tick selected for EURUSD and selected seconds for @SP in the hope of speeding things up, no luck though. On the EURUSD I have the backtestiing resolution set to 1 tick and on the @SP it is set to 1 sec. I would have thought that there would be as much tick data for the EURUSD as the @SP as it is a liquid product. This has me puzzled ?? Hopefully someone can shed some light. Backtesting is also slow with the @YM. Also of interest is when I load the worksheet the @YM and @SP are much slower to load than the EURUSD. I do not understand how the data load process works in detail on TS. I was wondering if it hand anything to do with the symbol themselves. Can't find anything on the net about this issue.
This post has been selected as an answer to the original posters question
TS pulls the data and stores it locally on your disk. Assuming you are using the same start and end date try adding a simple counter to your strategy (or do an empty strategy with a counter) that counts each calculation cycle. This way you know if the amount is similar or way different.
That would be the first thing I'd do. If this shows the values being close enough the additional time might come from your code.
TS just got back to me and they have said that there is a lot more tick data on the @YM than the EURUSD, so that answers my question. The issue I was having with GV's was that the names of the GV's in the set and get functions were different. I thought this may have had an impact on speed of optimisation, but this was not the case. Looks like it is down to tick data.
The following user says Thank You to oreild3 for this post:
thanks for the explanation. You should be able to test this easily when you load one chart for each symbol with the same bar interval and same amount of bars. Then you run two optimizations with look inside bar backtesting disabled and the time it takes should be about the same.