Notes about possible strategies the trading bot is/might be implementing.

My personal preference is long term investing, hence the bot will be identifying only long entries (no sell signals for now).

We need to manually pick only world class companies. This way we have high chance of profit even at suboptimal entry points (in long term).

Signals will be generated based on following criteria:

• daily RSI below 30 (rather recovering from being below 30)
• weekly stochastic RSI recovering from below 20 or so (daily stochastic tends to be rather noisy), weekly stochastic is giving rather strong signals
• another signal from weekly stochastic RSI - if below threshold, check where crossover of those 2 lines happens, this will be likely good long signal
• closing price near significant support (exponencial) moving average (50, 100, 150, 200)
• Heikin Ashi candle sticks changing from red to green
• moving average crossing could be useful signal
• monitor also slope of moving averages when they are crossing, this will be indicating momentum
• detect double bottom for multiple time intervals (years, months) - use SavGol filter as a fit
• Bollinger bands - checking if price in vicinity of the upper/lower band
• ...
• (maybe even use machine learning - combine LSTM RNN network with decision tree or something like that)
• examples on Iris dataset turn out to follow similar logic as trading according to technical indicators

Observation about RSI and MACD: often I see that RSI goes on daily highly to overbought, then drops to vicinity of overbought and then MACD signals sell on daily, likely similar will work for buy signals.
Also just crossover on MACD is not sufficient, the corossover has to be the highest/lowest crossing poin for last n days to be actionable signal. Additionally we can check if histogram is changing directions, can serve as early warning for buy signal. MACD is in general a bit lagging signal.

Supertrend indicator looks promising.

Use confluence of signals for higher success probability.