Parabolic SAR robot Review

[david.jukl]

Batchboy, I am just testing that new version of Parabolic, so I can´t tell you the results. Personally I wouldn´t change settings on your mini-live-account until first results of optimized version. However as Pipcop says - EA proves it´s quality after longer period (weeks or better months...). I think you shoud wait at least one or two weeks

[Batchboy]

hi again Karl & David,
I was about to do this new version with optimization instead, but then I recalled my own wrestlings with what would be a solid scientifically created EA, so I hope I don't impose too heavily (??) if Karl...... you could look over this list from Invis threrad that I put together to see what other scientific type tidy-up could be done, otherwise, in the very least, live$ or demo-live may definitely not act like optimizations/back-tests, thanks buddies! When done to fruition I then partake of new EA in my $8_P.SAR thread, you won't be dissatisfied doing this:
http://www.pipcop.com/forums/2539-post11.html
jb

[1ka]

JB,

I'm not sure that we can do the 'scientific' analysis on this, primarily because the opening and closing of trades is done on a tick -by-tick basis, not when a new bar opens. We know that this means back-testing results may not be replicated when forward-testing.

What could be possible is to set up pending buy and sell trades that are used to trigger trades. These would be more 'repeatable', and could actually increase profit as we are opening a trade at a potentially 'higher' point in the trend , rather than at the bottom of the trend range . (Sorry if that makes no sense, I probably need to draw this )

I'll think about this over the next week and see if it's possible.

[Batchboy]

Originally Posted by 1ka

JB,

I'm not sure that we can do the 'scientific' analysis on this, primarily because the opening and closing of trades is done on a tick -by-tick basis, not when a new bar opens. We know that this means back-testing results may not be replicated when forward-testing.

What could be possible is to set up pending buy and sell trades that are used to trigger trades. These would be more 'repeatable', and could actually increase profit as we are opening a trade at a potentially 'higher' point in the trend , rather than at the bottom of the trend range . (Sorry if that makes no sense, I probably need to draw this )

I'll think about this over the next week and see if it's possible.

hi 1ka,
I notice you say that this is the way it is doing trades now, tick -by-tick , but that doesn't mean it couldn't be done logic-wise so that it is based on static C(-1) when C(0) is finished being formed??? You just lockk-in the decisuion of a trade or close a trade from dynamic tick -valued TAs and prior dynamic decisions from such, "hurrah scientific", LOL
It would be behooving and interesting to do this way so back-test optimization is correct and valuable (to make my appealing speech )
I know if Invis wasn't done this way it would've been all hopeless to find parameters "in the dark", inestimable value.
It also makes the optimization go faster thru wider range of parameters because "Opening of bar price" is selected.
jb

[1ka]

Originally Posted by Batchboy

hi 1ka,
I notice you say that this is the way it is doing trades now, tick -by-tick , but that doesn't mean it couldn't be done logic-wise so that it is based on static C(-1) when C(0) is finished being formed??? You just lockk-in the decisuion of a trade or close a trade from dynamic tick -valued TAs and prior dynamic decisions from such, "hurrah scientific", LOL
It would be behooving and interesting to do this way so back-test optimization is correct and valuable (to make my appealing speech )
I know if Invis wasn't done this way it would've been all hopeless to find parameters "in the dark", inestimable value.
It also makes the optimization go faster thru wider range of parameters because "Opening of bar price" is selected.
jb

Unfortunately when a bar opens, the price is not always at the right value to trade, there is a trigger point that is often not met at bar opening, but at some indeterminate point during the 30M bar. One possible way is to use the opening of 1M bars as a trigger to trade whilst still using 30M bars to calculate the parabolic SAR value. Hmm, that could work. I might tweak the EA to see what the impact of that is. Thanks

[Batchboy]

Originally Posted by 1ka

Unfortunately when a bar opens, the price is not always at the right value to trade, there is a trigger point that is often not met at bar opening, but at some indeterminate point during the 30M bar. One possible way is to use the opening of 1M bars as a trigger to trade whilst still using 30M bars to calculate the parabolic SAR value. Hmm, that could work. I might tweak the EA to see what the impact of that is. Thanks

1ka,
If you do 1min triggers, would that not mean the chart would have to be on 1min's, therefore recalculatiopn of what the parameter range would have to be to accomodate such? I forget what periodicity P.SAR is as I'm not looking at chart for it, is it 30Min? So the parameters would have to be in range of 30X larger, "stretched-out" you might say. That would work, I've looked at TAs when going to other periodicites and siimply multiplied up the parameters as I went "deeper" to look at it equivalently.
jb

[1ka]

Originally Posted by Batchboy

1ka,
If you do 1min triggers, would that not mean the chart would have to be on 1min's, therefore recalculation of what the parameter range would have to be to accommodate such?...
jb

The chart is a convenient place for the EA to run. It is not limited to the timescale, or even currency , of the chart.

[Batchboy]

Originally Posted by 1ka

The chart is a convenient place for the EA to run. It is not limited to the timescale, or even currency , of the chart.

ah true, coding -wise the case, but I go back to C(-1) end of closed-bar at 30th 1min or simply C(-1) of 30min bar, it's going to be the same. identical value, I don't think the TAs in any EA are meant to only trip on tick moments during forming bar, TAs have their dynamuic value available whenever a bar is finished being formed. Granted you'd like 1min bar C(-1)'s to yield tick -like interval checks for trips during the 30min forming bar to avail replicatability also during back-tests (more 1min bars than tick -data), ok, so be it, but have a true/false that has EA deal with 30min C(-1) or 1min C(-1) so you can see if one over the others back-test/optimization will be anything more desirable one way or the other. Or not if you prefer and rather just go straigtht with 1min resolutuion results. I'm not voting here for one or the other to prove any point , because I don't think any of us know if one way or the other will be more desirable than the other, main point is you'll be "going scientific" in doing it at C(-1) (or O(0) [Open price, which is what selection on tester actually is, most of the time close values are same as opens, but not always on a percentage basis]) and one can thus find with true/false switch (if you're curious lijke me ) if 1min resolution yields any more desirable results than the other. In meantime, it be all replicatable by way of "end of bar" Statics . I have a hunch that just straight 30min C(-1) will be just as good as any results for 1min resolution, but not so strong a hunch that I vote for such .
jb

[Batchboy]

but I do see how for SLs 30min C(-1) values could half the time go over the trip value, of course, buuuut, "half the time" quotes be going under not tripping and continuing a trade would compensate to a large degree if not half as much, thus it would be interesting to run long enough back-tests/optimizations to have true/false switch for 1min bar action or just 30min bar action. If no, then doing what you all (David & 1ka) plan for 30min bar chart action checked within by 1min bar action seems good and back-test replicatable, as far back as 1min data exists in an MT4, which often can be a problem (think I have one with 4mo of 1min data on mini-demo).
jb

[portal]

Originally Posted by david.jukl

Batchboy, I am just testing that new version of Parabolic, so I can´t tell you the results. Personally I wouldn´t change settings on your mini-live-account until first results of optimized version. However as Pipcop says - EA proves it´s quality after longer period (weeks or better months...). I think you shoud wait at least one or two weeks

Is that sample 3 you are testing? Can you keep us updated with your results pls.
Thanks