Questions from Beginners MQL5 MT5 MetaTrader 5 - page 1528
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Well, what do you do in that case?
Reduce the number of combinations, use genetics.
Or break the optimisation period into parts and keep checking successful settings sequentially on new parts of the history. This can improve the stability of the system.
You can optimise on OHLC M1, and forward on ticks.
It is possible to optimise one parameter in the beginning, then save the best settings and optimise the next ones, selecting only the best variants from the previous optimisations.
In general, there are variants.
It makes sense to count 1000 indicator settings on the GPU at once, which I understand you don't need. But you can think about virtual positions - then it will make sense to test 1000 strategies at once, but this is a more complicated approach.
or is cloud computing more profitable for them?
Of course it is favourable, but often slow calculations are related to an error in the code logic. Do you really need to recalculate the indicator value on every tick, maybe it is enough to count at the opening price?
I don't know your code and strategy. Why moving stops on every tick is necessary - close very close to the market? Then the DC is not a flipper to set them.
In my experience, if your patterns are not on ticks, it makes sense to check the logic of the EA and move stops once a minute, then the result from changing the testing mode will be comparable. Check the final result on ticks, and the intermediate result can be optimised using OHLC.
I don't know your code and strategy. Why do you need to move stops on every tick - close very close to the market? Then DC is not a flipper to set them.
In my experience, if your patterns are not on ticks, it makes sense to check the logic of the EA and move stops once a minute, then the result of changing the testing mode will be comparable. Check the final result on ticks, and the intermediate result can be optimised using OHLC.
I have a moving stop loss and it approaches every N pips to the price if it goes in the losing direction, how can it be profitable in this case to test on OHCL?
it will simply react incorrectly in case the candle closes and the stop loss has not approached the required number of N'th pips, so it will increase the loss.
and if I close the loss on the fact of candlestick closing by subtracting from it the number of bullish or bearish or high or low points, but it will not be a moving stop loss, but a drawdown squared.
I have a movable stop loss and it approaches every N pips to the price if it goes to the losing side, how can it be profitable in this case to test on OHCL?
It will simply react incorrectly when the candle closes and the stop loss has not approached the required number of N'th pips, so it will increase the loss.So what prevents to calculate stop loss at once, which will be maximum? Changing the stop once a minute seems quite reasonable, it reduces the probable overfitting when evaluating the result.
So what prevents you from calculating the maximum stop loss right away? Changing the stop once a minute seems quite reasonable, it reduces the probable overfitting when evaluating the result.
And if the volatility is high? Will the stop not be knocked out?
What if the volatility is high? Will it hit the stop?
Do you count stops from the high stop - when buying and low - when selling? Well, then you can perform your entire algorithm not once a minute, but twice - checking extrema and if the price is higher or lower than the opening, then make the calculation.
However, most likely, close stops will not allow to set DC!
That's the point, they are far and it is necessary to move them closer to the market when the market goes into negative, the more into negative the less stop loss.