Zig Zag

[DaveSkender]

Community discussion about the ZigZag indicator

[codebeaulieu]

I'm not sure how to interpret the output from this indicator. I'm not using any sort of drawing library to visualize the output, it looks like the output is useful for actually drawing. I write unit tests against the output results, but I cannot tell from the output, what the most recent zigZag high and low pivots are.

I'd expect to see something like your SuperTrend indicator where it's easy to tell based on the date and whether the upper band or lower band is null or not.

I understand algorithms differ across libraries, so I expect the output to be somewhat different. But if the inputs are similar, the values should be close. The way I work with indicators is by saving candle data to CSV files and then writing unit tests against them. So, for most indicators I can get very close to trading view charts.

I've tried three different input settings for the % change, 3.5, 5 and 10 and the output didn't appear to come remotely close to Trading View, or at least I couldn't interpret it correctly.

Here's the output for daily time interval, EndType.Close and percentChange of 5. The TradingView chart above had a deviation % of 5 and a depth of 10. Data source is polygon.io

---
retrace high: 1186.4142857142857142857142857
retrace low: 997.6476923076923076923076923
date: 01/04/2022 00:00:00
zigZag: 1156.575
---
retrace high: 1173.0485714285714285714285714
retrace low: 1007.4184615384615384615384615
date: 01/05/2022 00:00:00
zigZag: 1113.370
---
retrace high: 1159.6828571428571428571428571
retrace low: 1017.1892307692307692307692308
date: 01/06/2022 00:00:00
zigZag: 1070.165
---
retrace high: 1146.3171428571428571428571429
retrace low: 1026.96
date: 01/07/2022 00:00:00
zigZag: 1026.960
---
retrace high: 1132.9514285714285714285714286
retrace low: 1014.0592857142857142857142857
date: 01/10/2022 00:00:00
zigZag: 1053.38
---
retrace high: 1119.5857142857142857142857143
retrace low: 1001.1585714285714285714285714
date: 01/11/2022 00:00:00
zigZag: 1079.80
---
retrace high: 1106.2200000000000000000000000
retrace low: 988.2578571428571428571428571
date: 01/12/2022 00:00:00
zigZag: 1106.22
---
retrace high: 
retrace low: 975.3571428571428571428571429
date: 01/13/2022 00:00:00
zigZag: 1082.5954545454545454545454545
---
retrace high: 
retrace low: 962.4564285714285714285714286
date: 01/14/2022 00:00:00
zigZag: 1058.9709090909090909090909091
---
retrace high: 
retrace low: 949.5557142857142857142857143
date: 01/18/2022 00:00:00
zigZag: 1035.3463636363636363636363636
---
retrace high: 
retrace low: 936.6550000000000000000000000
date: 01/19/2022 00:00:00
zigZag: 1011.7218181818181818181818182
---
retrace high: 
retrace low: 923.7542857142857142857142857
date: 01/20/2022 00:00:00
zigZag: 988.0972727272727272727272727
---
retrace high: 
retrace low: 910.8535714285714285714285714
date: 01/21/2022 00:00:00
zigZag: 964.4727272727272727272727273
---
retrace high: 
retrace low: 897.9528571428571428571428571
date: 01/24/2022 00:00:00
zigZag: 940.8481818181818181818181818
---
retrace high: 
retrace low: 885.0521428571428571428571428
date: 01/25/2022 00:00:00
zigZag: 917.2236363636363636363636364
---
retrace high: 
retrace low: 872.1514285714285714285714286
date: 01/26/2022 00:00:00
zigZag: 893.5990909090909090909090909
---
retrace high: 
retrace low: 859.2507142857142857142857143
date: 01/27/2022 00:00:00
zigZag: 869.9745454545454545454545454
---
retrace high: 
retrace low: 846.3500000000000000000000000
date: 01/28/2022 00:00:00
zigZag: 846.3500000000000000000000000```

### Questions

Could you please help me to understand the output. 
Where are the pivot points from this indicator? If the zigZag value represents the pivot points, there seems to be too many, even if I have the percent change set to 10 there are pivots nearly every day.

Let me know if you need clarifications or any more info from me.

[DaveSkender]

The first thing I'm seeing here is that the chart is pivoting on High/Low, not Close, so to be the same, you'd have to use EndType.HighLow as your parameter.

The first span is always unreliable too since it has to find the first pivot, so you should ignore that section. There are some things to also consider if the underlying data is different too.

[DaveSkender]

Another thing to note, the "zigzag" value in your results is the actual line between points; so to isolate the actual Pivots, you'd have to filter it to only include PivotType is H or L, or simply not null.

[codebeaulieu]

@DaveSkender Relevant issue for the Stock.Charts library. facioquo/stock-charts#180

[DaveSkender]

Yep, I'm actively working towards getting the demo site updated to shows more current chart data for a few symbols and for all indicators in the library, to help people visualize what these actually look like in a live-action scenario.

I suppose it could also be used to compare with other charting sites' output; though, you have to also consider all the factors to understand why they might be different -- including that the other site might be wrong! 😃

[DaveSkender]

You can see facioquo/stock-charts#181 as an example of how I added ZigZag to our demo chart.

[M22arius494]

Hello Dave,

I'm also observing that the results of the ZigZag are rather different than what TradingView is presenting/calculating. Yours is picking up additional highs and lows which TradingView is ignoring. TradingView is using a depth which isn't in you r sources (not sure if it should be).

For example the AXSBTC@Binance. I'm using the 2h candles from 1 december 21 up to 3 march 22 (746 candles should be more than enough) and start comparing from 1 jan 22. The first 2 are perfect but the 3 and 3th are weird (the two marked with the question marks). Because of this extra h/l the ZigZag is unusable for my needs (the results from this example isn't to bad though). In other cases it will pick up additional lower lows and lower highs while in fact it is properly trending on TV.

date=01-01-2022 11:00:00 zigzag=0.00194800 PointType=L RetraceHigh= RetraceLow=0.00194800
date=03-01-2022 15:00:00 zigzag=0.00207400 PointType=H RetraceHigh=0.00207400 RetraceLow=0.00186230

date=05-01-2022 23:00:00 zigzag=0.00177000 PointType=L RetraceHigh=0.00191853 RetraceLow=0.00177000 ???
date=06-01-2022 19:00:00 zigzag=0.00186300 PointType=H RetraceHigh=0.00186300 RetraceLow=0.00172467 ???

date=07-01-2022 05:00:00 zigzag=0.00170200 PointType=L RetraceHigh=0.00184682 RetraceLow=0.00170200
date=08-01-2022 05:00:00 zigzag=0.00180800 PointType=H RetraceHigh=0.00180800 RetraceLow=0.00168165
date=11-01-2022 01:00:00 zigzag=0.00162400 PointType=L RetraceHigh=0.00180615 RetraceLow=0.00162400
date=12-01-2022 19:00:00 zigzag=0.00180500 PointType=H RetraceHigh=0.00180500 RetraceLow=0.00165741

@code
decimal zigZagPercentage = 5;
List resultZigZag = (List)Indicator.GetZigZag(candles.Values.ToList(), EndType.HighLow, zigZagPercentage);
//^The editor is removing some generics in above line

for (int i = 0; i < resultZigZag.Count; i++)
{
ZigZagResult zigZagResult = resultZigZag[i];

string s = string.Format("date={0} zigzag={1:N8} PointType={2:N8} RetraceHigh={3:N8} RetraceLow={4:N8}", zigZagResult.Date.ToLocalTime(), zigZagResult.ZigZag, zigZagResult.PointType, zigZagResult.RetraceHigh, zigZagResult.RetraceLow);
AddTextToLogTab(s);
}

[DaveSkender]

Thanks for your feedback, Marius. I think adding this new depth parameter can help users skip short-term reversals. This is already in our backlog. See #694. Currently, if you use a Depth value of 1 in TradingView, it should look the same, assuming their base approach is the same.

Our current formula is the same one used by StockCharts.com, Fidelity, and most other platforms.

Just as an FYI, as a rule, we don't source our formulas from TradingView or any other third-party source code because it's impossible to validate them against reputable original sources of formulas. In other words, we can’t know whether they are implemented correctly. Sometimes they are not. This is especially true for community-contributed indicators as there is little or no quality control on them.

[Gadirino]

Hello @DaveSkender
I used ((( indicators.get_zig_zag ))) in my python codes and the outputs were very irregular.
What do you think is the problem?
There are tow different chart data and zigzag outputs:

[DaveSkender]

It's hard to say. It almost looks like you're overlaying results from a different ticker symbol. @LeeDongGeon1996 any thoughts?

[LeeDongGeon1996]

Are those calculated from same data? and got different results? I get it.
Umm.. because of passing too small percent_change value, that is unable to ignore small percent movement?

[Gadirino]

Thanks a lot for your answers
I resolved that.
My input data have some unknown problems in its date column. I changed data to another quote and the problem solved.

[Abdulhadi77]

How to draw the zigzag in python?

[DaveSkender]

I don't have any specific ideas for Python, but checkout our list of charting and visualization options

[LeeDongGeon1996]

I recommend Plotly for Python.

[adealkdke]

How do I get my zigzag indicator with pivots to look like yours from your website stock indicators.dev specifically referencing the zigzag indicator. I have tried multiple times to replicate your chart indicator but have not been able to get straight lines like yours from the pivot reversals and the zigzag lines.

[DaveSkender]

Start with a standard 5% percentChange deviation with high/low endType reversals. If your prices have typical volatility it should look something like this.

// C# usage 
var results = quotes.GetZigZag(
  endType: EndType.HighLow,
  percentChange: 5);

If you’re seeing too many or too few reversals, increase or decrease the change threshold. You can play around with different values to see what they do.

If you’re plotting on your own chart, and are seeing flat, jagged, or missing spots in the lines between pivots, ensure your axis scale isn’t inserting extra days that don’t correspond to your quotes — some time scales will do that. Most market prices won’t include weekends.

[DaveSkender]

Also, check the dates on your quotes. Like the example above, bad or incorrectly sorted dates can cause problems. We have a quotes.Validate() utility that can check those.

[ydqsbfsyi]

I'm getting this error: No method matches given arguments for List`1.Add: (<class 'str'>)
I checked, "date" dtype is datetime64[ns], like yy-mm-dd hh:mm:ss, all others are float64.
I tried "quotes.Validate()", but I got AttributeError: 'DataFrame' object has no attribute 'Validate'.
Could anybody please give me a hint?

[DaveSkender]

This doesn't sound like an issue with ZigZag, but rather a more basic issue with how you're composing your quotes from another Dataframe object. There are examples on composing quotes in our Quickstart guide that may help.

[ydqsbfsyi]

This doesn't sound like an issue with ZigZag, but rather a more basic issue with how you're composing your quotes from another Dataframe object. There are examples on composing quotes in our Quickstart guide that may help.

Thank you very much for your help. I'll look into it.