TradingView-RU
06/23 14:00
Modern Adaptive Supertrend [GBB]
Supertrend is probably the most copied trend indicator on TradingView. Most of the copies inherit the same handful of weaknesses so this is my attempt to fix the ones worth fixing, test the result properly, and be straight about the fix that didn't pan out.
If you have used the classic Supertrend you already know the idea. It is a volatility-based trailing line. It sits below price in an uptrend, above price in a downtrend, and it flips when price closes through it. Simple, fast to read, works on any chart. That simplicity is why it spread everywhere, and it is also where the trouble starts.
WHAT GOES WRONG WITH THE CLASSIC VERSION
Three things. The distance from price never adapts, so a clean trend and a sideways mess get treated exactly the same. The lookback never changes speed. And it flips on a single touch, so one wick through the line flips your bias and the next bar flips it back. That last one is the whipsaw everyone complains about. Run the classic as an always-in system on the 1h or below and in my testing it bleeds out. It only really holds up at 4h and above, where the trends are big enough to pay for all the flipping.
WHAT THIS VERSION DOES DIFFERENTLY
Two changes that work (and one that didn't but I left in place for you to experiment with)
The commit filter . This is the one that did the real work. Instead of flipping the moment price tags the line, price now has to close past it by a margin, about half an ATR by default, and optionally hold for a bar before the flip counts. That single change cut false flips by roughly 60% across the timeframes I tested. Same trend read, far less noise.
The adaptive distance . The band widens when the market is trending cleanly so it holds the move instead of shaking you out, and it also widens in chop, because a tight band in a range is a whipsaw machine. The only place it tightens is the transition, when a new move is just starting and you actually want it responsive. It judges trend versus chop against the market's own recent behaviour rather than a fixed threshold, because fixed thresholds do not travel well on crypto. I found that out the hard way.
The one that didn't work . I also built an adaptive lookback that was meant to let the indicator change speed with the market's rhythm. It added nothing. On crypto the rhythm turned out stable enough that adaptive just gave me a slower copy of the same line. It is still in the settings as an experimental toggle, switched off by default. Maybe you can find a tweak to make this feature work? If yes send me a DM :-)
One thing to be clear about before the settings. This is a trend filter, not a trading system. On raw direction it is right about 48% of the time, which is a coin flip, because Supertrend follows trend, it does not predict it. The value is in a cleaner trend read and lower drawdown, not in calling tops and bottoms. More on that at the bottom.
SETTINGS
Preset
Classic, Modern, or Custom. Classic is the textbook indicator, unchanged, so you can put it side by side. Modern is the default and what I recommend. It is timeframe aware: on 1h and below it runs both the adaptive distance and the commit filter, and above 1h it runs the commit filter only, because the adaptive distance did not add anything at higher timeframes in testing. Custom lets you toggle the layers yourself.
Commit filter (hysteresis)
- Commit buffer: how far past the line price has to close before a flip counts, measured in ATR. Default 0.5. Higher means fewer flips and slower turns.
- Persistence: how many closed bars the break has to hold before it flips. Default 1.
Adaptive distance (regime multiplier)
- Trend gain: how much wider the band gets in a clean trend. Default 0.8.
- Chop gain: how much wider it gets in a range. Default 0.5.
- Percentile window: how much recent history it uses to judge trend versus chop. Default 500 bars. Shorter reacts faster but is jumpier.
Adaptive period (experimental, off by default)
Turns on the adaptive lookback described above, with min and max clamps on the range. The defaults are in there, but as I said, it did not improve results. Leave it off unless you want to poke at it yourself.
Classic settings
ATR period and multiplier. Used by the Classic preset and as the base the other layers build on. Defaults 10 and 3.
Display
Direction-coloured line, an optional fill between price and the line, and flip markers. There is also an optional regime readout that shows whether the engine currently sees trend or chop. Off by default to keep the chart clean. Turn it on if you want to see what it is "thinking".
Alerts
Separate flip-up and flip-down conditions, plus an optional JSON payload if you want to wire it into a webhook.
HOW TO USE IT
As one input, not the whole decision. It answers a single question, which way is the trend leaning right now, with less noise than the classic gives you. It does not tell you where to enter or where price goes next. Use it to set your bias, then bring your own levels, structure, or entry trigger for the rest. Treating a trend filter as an entry system is how people lose money with a tool that is working exactly as designed.
ThíchChia sẻ
Tất cả bình luận0Mới nhấtPhổ biến