JSParticipant
Veteran
The code runs on a 1-minute timeframe, and I do not use StopLoss, TakeProfit, or TrailingStops.
What I do use is a kind of emergency stop set at 2% (based on PositionPerf).
The decomposition and standard deviation I use have a period of 100 minutes, meaning it looks back 100 minutes.
Every minute, the code checks whether a characteristic of the signal crosses a certain threshold.
In earlier versions, these thresholds were static, but in the latest version, they are dynamic, based on the aforementioned standard deviation.
Only these threshold values have been optimized, and that optimization was done over a dataset of 200k units.
JSParticipant
Veteran
I also apply money management, where the system starts with a minimum position size of 0.2 contracts.
All profits are reinvested, with a maximum position size capped at 10 contracts (for the sake of my blood pressure 😅).
This is gold.
Just had to unload my brain onto a whiteboard before commencing studies. This may take me all year and hopefully be done with indicators. Godspeed to you too @Grahal and anybody else pursuing this direction.
JS are you using Indicators to quantify any or even all of the ‘9 dominant characteristics of the signal’?
Is High, Low, Open and Close … 4 of the 9 characteristics or does decomposition centre around ‘Close’ only?
JSParticipant
Veteran
No, I don’t use indicators only simple arithmetic…
It’s only about the decomposition of the “Close”…
JS is your System we are talking about here based on below, and also with your recent addition of standard deviation?
Big Thanks in Anticipation
JSParticipant
Veteran
Yes, that’s part of it, along with an equivalent formula using the standard deviation…
I was good at math… but I still don’t understand this.
JSParticipant
Veteran
It is a similar type of formula to the one used for an EMA, for example:
EMA[n] = α ⋅ x[n] + (1 − α) ⋅ EMA[n−1]
Both formulas are recursive, meaning that previously calculated values (of the EMA) are used in the current calculation…
JS, have you read his last book ? And are you using MTF ?
JSParticipant
Veteran
His last book?
No use of MTF… (only 1 minute)
Cycle Analytics for Traders, last J Ehlers’ book
JSParticipant
Veteran
No, I haven’t read it, although I find the work of John Ehlers very interesting…
Most of the things you describe of your system match perfectly with Ehlers’s works : cycles, DSP, Recursive and non Recursive Filters, even how he uses the stop loss…
“My experience is that a stop
loss will decimate the
robustness of a trading
strategy if it is built into the
strategy and becomes an
integral part of it. Rather, a
stop loss is best left only as a
guard against extremely large
losses. Using a stop loss this
way will maintain the
robustness of the core
strategy you have built. There
are a large number of ways to
implement a stop loss rule.
The simple rule that works
for me is to let the stop value
just be a percentage of the
entry price”
If i understand well what you told us about your system, i think it might look like something like this :
Maybe this link can help you to improve your system. This guy has also embeded the volatility in the Sinewave indicator…
JSParticipant
Veteran
Correct, that matches my experience with a “stop loss” as well — not that I have anything against using a “stop loss”, “take profit”, or “trailing stop”, but it disrupted my system…
The most important and commonly used domains in DSP are the “time domain” and the “frequency domain”.
While the “frequency domain” uses the analysis of sinusoids (sine and cosine waves), the time domain deals with changes over time, for example in amplitude (price)…
The “frequency domain” is well suited for analyzing sinusoids, such as in sound/music or anywhere the source of the signal consists of sinusoids. However, I believe the origin of our signal does not come from sinusoids, but from “tick data”…
This “tick data” is first made “discrete” (by using timeframes) and can then be analyzed in the “time domain”…
Another factor is that frequency analysis is considerably more difficult and complicated to perform…
So the first choice to make is: do I stay in the “time domain”, or do I conduct my analysis in the “frequency domain”?
Where John Ehlers has clearly chosen to perform analysis in the “frequency domain”, my analysis remains “limited” to the “time domain”, meaning no use of sinusoids or spectrum analysis, for example…
