In this video is shown the main ideas of the best system designed by me:
https://drive.google.com/file/d/1PNn2umMgUX7CX-gAVnlbzF2kn4iBGBCn/view?usp=drive_link
OUR SYSTEM: REGRESSION LINE CROSSES NOISE
The most fundamental aspect of the system is to consider the “true price” as a linear regression of the price, represented as a curve.
Market selection, timeframe and direction:
It is valid for any type of market. The backtests have been carried out on individual American tickers, while the real-time simulated walkforward has been executed on Forex, metals and cryptocurrencies. They are fractal patterns, valid in any time space. The backtest has been in daily bars, while the walkforward in 1 and 3 minute bars. Both long and short.
Patterns:
There are many possible patterns, but the easiest to operate and the one that will be studied in this dossier is the trend continuation pattern, sometimes known as “Turtle Soup” (Linda Raschke), regardless of whether it is bullish or bearish. An unequivocal prior movement in the direction is necessary, a clear trend, indicated by the curve of the regression line (true price), and if possible, with a certain order (an easy-to-manage volatility). A halt in that trend and an indication that the trend is resumed by a new higher “belly” of the regression line (long) or a new lower “hump” of the regression line (short). Obviously, it has more details that are difficult to convey in writing. And one of the main issues is to reduce the probability of lateral movement or “chop”.
Entry:
Given the conditions of the pattern. The standard entry consists of synchronizing the timing of the entry with the passage of the regression line (true price) through the noise (expressed as +/- 0.5 standard deviations in color beige) and signaled as a flip of the PSAR trend indicator (the regression line changes the PSAR). Timing is crucial. We go with the trend, we look to go long at a higher price, and we look to go short at a lower price.
Initial stop:
The stop is defined by the “belly” of the regression line in long, and the “hump” in short. In both cases, if the price reached that place, the idea of the trend continuing would be mathematically invalidated. Such a close stop allows us to get an idea of low risk in relation to the possible profit.
Exits:
In the backtest, the simultaneous interaction of three outputs. A first exit that serves as a worst-case stop loss and that calculates the value of 1R. A trailing stop as a function of a moving average. An exit can be added if the trade reaches a target of R-Multiple. In the manual walkforward, exits have been purely discretionary, emphasizing taking risk out of the trade as soon as possible. The application of indicators, especially the PSAR, is the most advisable.
Recap:
It seems much simpler and more profitable to operate these systems manually and discretionally than fully automatically. There are many very complicated details to code for a robot.