Public discussion of the formula for calculating the cost of resources in the MQL5 Cloud Network - page 29
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
Here is an unrepeatable example so far.
Values when tested on the network:
Profit values when tested on local agents with parameters from the corresponding pass
2514: -998.80
3517: -996.62
3591: -996.98
3537: 53468.81
3630: 28514.22
build 523
EURUSD M1
Initial deposit: 1000
Period 01.01.2011 - 16.10.2011
Open prices only
Divergence appears on large testing intervals.
Either it is the tester's error or some data on the instrument is updated between optimization and single testing. 5318.15
I suggest to go further and add a dummy input parameter (on which the results are independent) and run optimization using this parameter only...
I got the following results
On local agents only (100 runs):
-4681,85 27/100
-3514,45 73/100
On cloud only (100 runs):
-4659,74 41/100
-3514,45 33/100
-1154,93 9/100
0.00 1/100 (test failed)
795,44 16/100
Only on cloud, second run after cache reset (100 runs) :
-4659,74 51/100
-3514,45 8/100
-1154,93 23/100
0.00 8/100 (test failed)
795,44 10/100
Yes, this option makes it as easy as possible to identify the presence of errors in the tester.
Optimisation using only one and then a dummy parameter:
And there should only be a clean straight line without any dots.
Thanks for the idea with insignificant parameter - we will check it and look for error.
We have already fixed one problem with floating spread effect in the upcoming 525 build.
A few questions about the display of agents in the profile.
Am I correct in assuming that PR captures the maximum value reached on an agent?
When will it be possible to enter a description?
Is it possible to organize displaying of all agents on one page (instead of 20 like now)?
A few questions about the display of agents in the profile.
Am I correct in assuming that PR captures the maximum value reached on an agent?
Showing the last calculated PR - it's recalculated periodically.
When will it be possible to enter a description?
Can all agents be shown on one page (and not 20 like now)?
A number of improvements will be introduced: collapsing agents by computer and showing 100 agents per page.
Look at the new "Map" function in the agent list. It shows on a world map where agents are installed.
Renat
PR is the agent's performance estimate right? obviously this estimate is not true according to thetable provided by userAshes,
Intel Xeon 3.00GHz, 2047MB is clearly superior toIntel Pentium 4 3.00GHz, 2039MB andIntel Core i3-2100 @ 3.10GHz, 4008MB is superior toIntel Pentium G620 @ 2.60GHz, 1954MB,
Renat
PR is the agent's performance estimate right? obviously this estimate is not true according to thetable provided by userAshes,
Intel Xeon 3.00GHz, 2047MB is clearly superior toIntel Pentium 4 3.00GHz, 2039MB andIntel Core i3-2100 @ 3.10GHz, 4008MB is superior toIntel Pentium G620 @ 2.60GHz, 1954MB,
There are marketing comparisons and then there's the reality in work cases.
Our tests show the periodically calculated instantaneous performance of each agent. This mechanism allows us to adjust for real-world computer load and properly account for performance.
There are marketing comparisons and there is the reality in work cases.
Our tests show the periodically calculated instantaneous performance of each agent. This mechanism allows us to adjust to the actual computer load and properly account for performance.
Or is it based on the owner's current equipment load? then it's fine, it's competent.
...
Look at the new "Map" function in the agent list. It shows on a world map the places where agents are installed.