Your interactive trading laboratory!
 • 
7 users online

Strange Trades in an RM without having any Buy-Signals
Q Merlin wrote 4 days ago, and edited 4 days ago ...

Just found a strange behavior using a rotational model on an indicator that I have developed myself (the Indicator says either 1 or 0. 1 shows a bar, 0 shows nothing). In the chart below, You'll see quite a few long positions although at the bottom in the signal panel, no buy signals have occurred. What's going on here?


Attachment

Cancel

Responses
Q Merlin posted 4 days ago

Another one

Another one

Q Glitch posted 4 days ago, and edited 4 days ago

The Rotation Model will always hold N symbols. It looks to me like possibly your indicator did not signal for ANY symbols for numerous bars. In this case the RM still sorts by the indicator values. If they are all zeroes you will wind up with essentially random signals. Try to use a weight indicator that provides some meaningful value on every bar.

The Rotation Model will always hold N symbols. It looks to me like possibly your indicator did not signal for ANY symbols for numerous bars. In this case the RM still sorts by the indicator values. If they are all zeroes you will wind up with essentially random signals. Try to use a weight indicator that provides some meaningful value on every bar.
Q Merlin posted 4 days ago, and edited 4 days ago

You're absolutely right, I did not think that far and of course, there are phases in which the whole universe is at 0. What does the backtester i'm just assign a value to Values if the signal is 1, else I do nothing?

if (signalValid == 1) { Values [idx] = signalValid }

You're absolutely right, I did not think that far and of course, there are phases in which the whole universe is at 0. What does the backtester i'm just assign a value to Values if the signal is 1, else I do nothing? if (signalValid == 1) { Values [idx] = signalValid }
Q Glitch posted 4 days ago

Thinking a bit more about this, I think a clean solution from Q’s end would sort by first weight indicator value, followed by a flag indicating if the symbol is already held. This way, symbols already held will be retained when using weight indicators that issue sporadic signals like yours.

Thinking a bit more about this, I think a clean solution from Q’s end would sort by first weight indicator value, followed by a flag indicating if the symbol is already held. This way, symbols already held will be retained when using weight indicators that issue sporadic signals like yours.
Q Merlin posted 4 days ago, and edited 4 days ago

Just realized it isn't that simple. Using a condition to either fill or not fill the indicators TimeSeries does not show any trading signals at all. It's a conditional indicator (I have about 6 of them) which is either true (1) or false (0), hence I don't see a way to convert this to an Oscillator. Any ideas how to solve this?

Just realized it isn't that simple. Using a condition to either fill or not fill the indicators TimeSeries does not show any trading signals at all. It's a conditional indicator (I have about 6 of them) which is either true (1) or false (0), hence I don't see a way to convert this to an Oscillator. Any ideas how to solve this?

Q Glitch posted 3 days ago

With an indicator that signals so sporadically, with either 1 or 0 values, are you sure a Rotation Model is the best expression of this idea? It strikes me that a standard model might utilize your indicator in a more natural way. You'd just need to establish the exit criteria.

But I think I will make a change in Q to make your indicator work. Basically, in cases where all of the weights are zero, I will prioritize symbols that already have a position. In this way, we won't see the random signals we're seeing now. Established positions will just remain open until they are usurped by new signals.

With an indicator that signals so sporadically, with either 1 or 0 values, are you sure a Rotation Model is the best expression of this idea? It strikes me that a standard model might utilize your indicator in a more natural way. You'd just need to establish the exit criteria. But I think I will make a change in Q to make your indicator work. Basically, in cases where all of the weights are zero, I will prioritize symbols that already have a position. In this way, we won't see the random signals we're seeing now. Established positions will just remain open until they are usurped by new signals.
Q Merlin posted 3 days ago

That's correct. Something like a checkbox saying "don't execute any trades if all symbols have the same value" would already solve the issue. I know that the weighting criteria is mainly driven by fluctuating oscillators but why not extend the RM for "business logic driven" buy signals?

That's correct. Something like a checkbox saying "don't execute any trades if all symbols have the same value" would already solve the issue. I know that the weighting criteria is mainly driven by fluctuating oscillators but why not extend the RM for "business logic driven" buy signals?
Forum Tips

Please sign in if you want to participate in our forum.

Our forum uses Markdown syntax to format posts.

To embed code snippets, enclose them in [CODE][/CODE] tags.