✨ Multi-TimeFrame usage + Automated zooming

[Rikj000]

The idea here is to:

• Make the buy/sell signals more robust due to more confirmation over more different candles
• Trade safer overall by doing more swing trades (Bigger candles give more certain predictions since they aren't so clouded by volatility)
• Make more profit, by going more and more near scalping mode when the market goes sideways (by detecting smaller trends and doing smaller shorter trades for smaller profits)

How this would be implemented:

• Use multiple candle data like we're doing now with TimeFrame-Zoom (TFZ), but even more:

  • 5m for original intra candle fluctuations during testing (purpose of TFZ)
  • Then for normal candle usage I was thinking about for example using:
    • 15m + 30m + 1h + 2h candles in a similar fashion as we combine candles in the dataframe now (so base timeframe will be 5m but we'll add the other data into it, this will make it rather heavy though), but being frugal in the amount of candles we'll use to not overload systems.

• We'd use all that candle data to detect which candles we'd like to use:

  • Base is 2hr, if this has an upwards/downwards trend detected then we use the buy/sell signals for it (upwards/downwards is currently enabled in the config)
  • If we use the signals from the 2hr timeframe, we zoom in, check if there the same trend is also detected, if so, add the buy/sell signals (2 candle sizes now confirm that it's a good moment to buy/sell)
  • We continue doing this till we don't detect a matching trend anymore (ideally we'd zoom in all the way to 5m, then all candle sizes give indication that it's a good time to buy/sell)

• If we detect sideways movement on 2hrs then we don't use that candle data for our buy/sell signals, but then:

  • We zoom in to 1hrs, to see if we can detect a trend here, if this is the case we start using the signals as buy/sell signals
  • We still continue zooming in if still no trend would have been found on the 1hrs

[Rikj000]

As @1337 pointed out on Discord, signals/indicators calculated using candles smaller then 15m are usually too clouded with volatility, leading to noisy/bad predictions. Best to not go under the 15m candle size when implementing this feature

[Rikj000]

Finally started on this feature! 😄 But it's a big & hard one to implement so it'll take some time.
The new settings section will most likely have the following format:

"timeframes": {
      "base_timeframe": "5m",
      "roi_timeframe": "4h",
      "unclogger_timeframe": "30m",
      "weighted_signal_timeframes": ["15m", "30m", "1h", "2h"],
      "core_trend_timeframe_multiplier": 4,
      "ignore_sideways_trends": false
}

I decided to split up the implementation into 4 parts:
☑️ ✨ Implement configurable core_trend_timeframe
☑️ ♻️ Convert core_trend_timeframe to core_trend_timeframe_multiplier + switch to timeframes settings section
🔳 ✨ Implement multiple weighted_signal_timeframes (:building_construction: In progress, very large changes... On-Hold, changes have been commited to a feature branch, but this lead to 19hr long HOs & no results found in it's current state)
🔳 ⚡ Implement ignore_sideways_trends

[Rikj000]

Decided to close this issue.
I deem it over-engineered after multiple tests where no best results could be found.

However the feature/multi-candle-usage branch won't be deleted.
So if any other developer still sees potential in this feature then they can still take a look at the work already done.