Pivots

[DaveSkender]

Community discussion about Pivots.

Note, this is not the traditional Pivot Points.

[DaveSkender]

The concept of pivots started as a discussion on how to visualize divergence, see #434 for the origination story of this indicator. It is meant to be a base to enable comparing point trends any two indicators or close price and an indicator and is a more customizable variation of Williams Fractal.

[kwcoffee1]

Hi Dave,

Can you recommend some practical values to test the Pivots indicator?
I tried: LeftSpan = 7, RightSpan = 3, MaxTotal = 30
Can you command on the purpose and the assigned values of these parameters?
Also the description of the 3 returned value, like HighLine?

Thanks

[DaveSkender]

Pivots are simply a more customizable version of a Williams Fractal. Based on his own analysis, Williams preferred these settings:

leftSpan = 2
rightSpan = 2
endType = high/low

To elaborate on the description in our documentation site, here's a bit more of an explanation and meaning.

the pivots

The pivots need leftSpan, rightSpan, and endType for calculation and represent a change in direction of price.

When a point pivot is detected, there will be a HighPoint or LowPoint value depicting the actual corresponding high or low point price.

Use larger span values to depict larger trends, but keep in mind that this is a retrospective analysis that will never identify a point until rightSpan periods have occurred, for confirmation.

the trend markers

To determine the trend markers (HH, LH, LL, HL) for an identified point, we look backwards maxTrendPeriods number of periods (bars) to find the last point. If the current point is a high point and is higher than the last high point, it's marked as a higher high (HH) in the HighTrend field. We do the same equivalent markings for lower high (LH) and for low point trends in the LowTrend field. We separate these low and high fields because a) there could be both a high/low point in the same bar in rare circumstances and 2) it's easier to consume for charting use cases.

the trend lines

HighLine and LowLine are simple values representing the line value between corresponding trend markers. Absence of a trendline value means the two adjacent trend markers were more than maxTrendPeriods away from each other and were not attributable to each other. This is also why we can't identify the marker of the first point in a contiguous trendline. Setting a very high maxTrendPeriods will produce more unbroken trendlines, but it comes at a cost of an exponentially slower computed analysis.

[kwcoffee1]

Thank you very much for the quick response.

I notice the default uses a maxTrendPeriods of 20 "periods". Is this the number of the 5-bar pivot evaluation periods?
That is what does maxTrendPeriods denotes?

After investigating the data furthermore, I believe maxTrendPeriods is the actual bar count and we should filter on the chances of the Trend and not on the individual bars as indicate return — your explanation sounds right.

[DaveSkender]

I used the term "periods" to mean the same thing as bars or candles. For example, if you're using daily quotes, 5 periods = 5 days. maxTrendPeriods is only used to limit the lookback period when identifying trends.

[Paroca72]

Hello,

I'm using 26 as period but the result are more that 1400 items.
I expected something like less 26 items as result.

What I'm missing?

Thanks for your support
Sam

[DaveSkender]

Are there actually 1400 items in candles?

[DaveSkender]

Yes. the other indicators work perfect with the same source of candles.

upload your raw candles and Period values to a new issue and I’ll try to repro.

[DaveSkender]

@Paroca72 I've added some suggestions for your code in this issue, and also updated docs to clarify what maxTrendPeriods means -- it's certainly confusing.

• GetPivot always returns a list of rows equal to the candles, not respecting the MaxTrendPeriod #1950

[Paroca72]

Yes..
the other indicators work perfect with the same source of candles.