OPEN-SOURCE SCRIPT
Swing Persistence Profile - Pivot Width Survival
Swing Persistence Profile ranks every swing high and swing low on the chart by the widest pivot
window it survives. It is also explicit about the exact bar on which that rank becomes knowable,
and it shades the region where it is not knowable yet.
THE IDEA
A pivot is not a property of a bar on its own. It is a property of a bar together with the width
of the window it was compared against. The same high is a swing high when you look two bars either
side, and is not a swing high when you look thirteen. Tools that ask you to pick one width and
then draw the result bury that choice inside a single number.
This script evaluates up to five widths at once - 2, 3, 5, 8 and 13 bars each side by default -
and records, for every bar, the widest width at which that bar is still the extreme of its window.
That number is the bar's persistence depth.
Because widening a window can only add constraints, the widths are strictly nested: a bar that is
the highest of its 13-bar neighbourhood is necessarily also the highest of its 8, 5, 3 and 2-bar
neighbourhoods. Depth is therefore a single ordered rank rather than five separate opinions that
have to be reconciled. A depth-13 swing is a depth-3 swing that kept going.
TIMING
A pivot of width w cannot be identified until w bars have closed after it. That is arithmetic, not
an implementation detail, and it means a bar's persistence depth is not final until the deepest
active width has passed.
There is one further subtlety that most pivot scripts leave in place. The w-th bar after the
candidate is itself still forming while it is the live bar, and its own high or low can still
invalidate the pivot. A script that reads the verdict during that bar will show a marker that
changes, or vanishes, before the bar closes.
This one waits one bar longer. Every verdict it reads comes from a bar that has already closed, so
the confirmation lag is the deepest active width plus one bar. A marker is created once, on the
bar it describes, and is then left alone: it is not moved, recoloured or deleted on later bars,
and it does not flicker intrabar either.
The bars still inside that delay are shaded, so the region where the answer can still change is
visible on the chart instead of being left to inference. Nothing inside the shaded region has been
ranked yet, so an empty shaded region means "not decided," not "no swings here."
WHAT IS DRAWN
A marker on each qualifying swing, placed on the bar that made the extreme, coloured and sized by
depth, optionally printing the depth as a number.
An optional structure line joining swings at or above one chosen depth, alternating between highs
and lows. The most recent segment extends if a later bar sets a more extreme value in the same
direction; once a swing in the opposite direction closes a segment, that segment is fixed.
An optional higher-timeframe reference: the price of the most recent confirmed swing high and
swing low measured on a higher timeframe, drawn as a stepped level. The requested expression is
offset by one bar and the request uses lookahead, which is the pair the Pine Script documentation
specifies for confirmed higher-timeframe values, so historical and realtime bars receive the same
series. The level therefore lags by the reference width plus one bar of the reference timeframe,
and that lag is the point rather than a defect.
Optional alerts on a swing reaching a chosen depth. The alert fires as many bars after the swing
as the deepest active width plus one; the swing itself is on the earlier bar. Because every
verdict is read from a closed bar, "Once Per Bar" and "Once Per Bar Close" fire on the same bar.
THE TABLE
"Swings" counts how many swings reached at least that depth across loaded history, highs and lows
together. "Kept" is the share of the previous row's swings that survived into this row. "Bars ea"
is the average number of bars per swing at that depth. The footer reports the confirmation lag in
bars and how many bars have been rated.
The survival ratios describe the instrument and timeframe you are looking at, not any particular
trade. If most depth-2 swings die at depth 3, the chart is producing many local extremes that do
not extend. If they survive, the same series is producing fewer and longer swings. Comparing that
profile across timeframes or across symbols is a way of choosing a swing width that matches how a
market actually moves, rather than choosing 5 because 5 is a common default.
Counts cover every bar of history the chart has loaded, so they are comparable between symbols
only when the loaded history is comparable.
SETTINGS
Depths in use activates between two and five of the width slots. Reducing it also shortens the
undecided region at the right edge, because that region is as long as the deepest active width
plus one settling bar.
Every depth threshold in the script - minimum depth to mark, minimum depth for the structure line,
and the alert depth - is stated in BARS EACH SIDE, not as a slot number. With the default widths,
"minimum depth 3" means the second of the five slots. A threshold set above the deepest active
width produces nothing, by design.
Enter the widths in ascending order. The ranking itself does not require it, but the table only
reads as a survival sequence when they ascend, and "Kept" can exceed 100% if they do not.
The higher-timeframe reference width is counted in bars of that timeframe: a width of 5 on a daily
reference means five daily bars each side.
WHO IT IS FOR
Anyone who uses swing points as an input to something else - reading structure, choosing the
lookback for a range or a channel, deciding how far back a level should be considered relevant -
and would rather see how sensitive those points are to window width than settle it by habit. It is
also usable as a calibration step before fixing a pivot length in another tool.
WHAT IT DOES NOT DO
It produces no entries, exits, targets, direction or bias, and nothing in it should be read as
one. It measures the geometry of the price series and stops there. Persistence depth describes
what has already happened; a deep swing is not evidence about what happens next.
NOTES
Highs and lows come from the chart series. On non-standard chart types - Heikin Ashi, Renko, Kagi,
Point and Figure, Range - those values are synthetic, and the depths will describe the synthetic
series rather than the market. The script declares alert conditions, so it should be used and
presented on a standard chart type.
TradingView permits 500 labels and 500 lines per script. On long histories the platform drops the
oldest drawings as newer ones are created, so the left edge of a deep history may be bare. This
does not affect the table, which counts every rated bar.
Persistence depth for highs and lows, and the bar index each verdict refers to, are available in
the Data Window. Those values describe the bar named by "Swing bar index," not the bar the cursor
is on.
One data request is made for the higher-timeframe reference whether or not it is displayed, so
enabling it costs nothing extra.
ORIGINALITY
The built-in ta.pivothigh() and ta.pivotlow() functions are used as the primitive, as they are in
a great many scripts. Everything above them is written for this script: running several widths in
parallel, aligning their verdicts onto a common bar so that they can be compared, reducing them to
one nested rank, deferring every drawing until that rank is final on closed data, marking the
region where it is not yet final, and reporting the survival ratios between widths.
Open source under the Mozilla Public License 2.0. The logic is described above in enough detail to
be checked line by line against the source.
window it survives. It is also explicit about the exact bar on which that rank becomes knowable,
and it shades the region where it is not knowable yet.
THE IDEA
A pivot is not a property of a bar on its own. It is a property of a bar together with the width
of the window it was compared against. The same high is a swing high when you look two bars either
side, and is not a swing high when you look thirteen. Tools that ask you to pick one width and
then draw the result bury that choice inside a single number.
This script evaluates up to five widths at once - 2, 3, 5, 8 and 13 bars each side by default -
and records, for every bar, the widest width at which that bar is still the extreme of its window.
That number is the bar's persistence depth.
Because widening a window can only add constraints, the widths are strictly nested: a bar that is
the highest of its 13-bar neighbourhood is necessarily also the highest of its 8, 5, 3 and 2-bar
neighbourhoods. Depth is therefore a single ordered rank rather than five separate opinions that
have to be reconciled. A depth-13 swing is a depth-3 swing that kept going.
TIMING
A pivot of width w cannot be identified until w bars have closed after it. That is arithmetic, not
an implementation detail, and it means a bar's persistence depth is not final until the deepest
active width has passed.
There is one further subtlety that most pivot scripts leave in place. The w-th bar after the
candidate is itself still forming while it is the live bar, and its own high or low can still
invalidate the pivot. A script that reads the verdict during that bar will show a marker that
changes, or vanishes, before the bar closes.
This one waits one bar longer. Every verdict it reads comes from a bar that has already closed, so
the confirmation lag is the deepest active width plus one bar. A marker is created once, on the
bar it describes, and is then left alone: it is not moved, recoloured or deleted on later bars,
and it does not flicker intrabar either.
The bars still inside that delay are shaded, so the region where the answer can still change is
visible on the chart instead of being left to inference. Nothing inside the shaded region has been
ranked yet, so an empty shaded region means "not decided," not "no swings here."
WHAT IS DRAWN
A marker on each qualifying swing, placed on the bar that made the extreme, coloured and sized by
depth, optionally printing the depth as a number.
An optional structure line joining swings at or above one chosen depth, alternating between highs
and lows. The most recent segment extends if a later bar sets a more extreme value in the same
direction; once a swing in the opposite direction closes a segment, that segment is fixed.
An optional higher-timeframe reference: the price of the most recent confirmed swing high and
swing low measured on a higher timeframe, drawn as a stepped level. The requested expression is
offset by one bar and the request uses lookahead, which is the pair the Pine Script documentation
specifies for confirmed higher-timeframe values, so historical and realtime bars receive the same
series. The level therefore lags by the reference width plus one bar of the reference timeframe,
and that lag is the point rather than a defect.
Optional alerts on a swing reaching a chosen depth. The alert fires as many bars after the swing
as the deepest active width plus one; the swing itself is on the earlier bar. Because every
verdict is read from a closed bar, "Once Per Bar" and "Once Per Bar Close" fire on the same bar.
THE TABLE
"Swings" counts how many swings reached at least that depth across loaded history, highs and lows
together. "Kept" is the share of the previous row's swings that survived into this row. "Bars ea"
is the average number of bars per swing at that depth. The footer reports the confirmation lag in
bars and how many bars have been rated.
The survival ratios describe the instrument and timeframe you are looking at, not any particular
trade. If most depth-2 swings die at depth 3, the chart is producing many local extremes that do
not extend. If they survive, the same series is producing fewer and longer swings. Comparing that
profile across timeframes or across symbols is a way of choosing a swing width that matches how a
market actually moves, rather than choosing 5 because 5 is a common default.
Counts cover every bar of history the chart has loaded, so they are comparable between symbols
only when the loaded history is comparable.
SETTINGS
Depths in use activates between two and five of the width slots. Reducing it also shortens the
undecided region at the right edge, because that region is as long as the deepest active width
plus one settling bar.
Every depth threshold in the script - minimum depth to mark, minimum depth for the structure line,
and the alert depth - is stated in BARS EACH SIDE, not as a slot number. With the default widths,
"minimum depth 3" means the second of the five slots. A threshold set above the deepest active
width produces nothing, by design.
Enter the widths in ascending order. The ranking itself does not require it, but the table only
reads as a survival sequence when they ascend, and "Kept" can exceed 100% if they do not.
The higher-timeframe reference width is counted in bars of that timeframe: a width of 5 on a daily
reference means five daily bars each side.
WHO IT IS FOR
Anyone who uses swing points as an input to something else - reading structure, choosing the
lookback for a range or a channel, deciding how far back a level should be considered relevant -
and would rather see how sensitive those points are to window width than settle it by habit. It is
also usable as a calibration step before fixing a pivot length in another tool.
WHAT IT DOES NOT DO
It produces no entries, exits, targets, direction or bias, and nothing in it should be read as
one. It measures the geometry of the price series and stops there. Persistence depth describes
what has already happened; a deep swing is not evidence about what happens next.
NOTES
Highs and lows come from the chart series. On non-standard chart types - Heikin Ashi, Renko, Kagi,
Point and Figure, Range - those values are synthetic, and the depths will describe the synthetic
series rather than the market. The script declares alert conditions, so it should be used and
presented on a standard chart type.
TradingView permits 500 labels and 500 lines per script. On long histories the platform drops the
oldest drawings as newer ones are created, so the left edge of a deep history may be bare. This
does not affect the table, which counts every rated bar.
Persistence depth for highs and lows, and the bar index each verdict refers to, are available in
the Data Window. Those values describe the bar named by "Swing bar index," not the bar the cursor
is on.
One data request is made for the higher-timeframe reference whether or not it is displayed, so
enabling it costs nothing extra.
ORIGINALITY
The built-in ta.pivothigh() and ta.pivotlow() functions are used as the primitive, as they are in
a great many scripts. Everything above them is written for this script: running several widths in
parallel, aligning their verdicts onto a common bar so that they can be compared, reducing them to
one nested rank, deferring every drawing until that rank is final on closed data, marking the
region where it is not yet final, and reporting the survival ratios between widths.
Open source under the Mozilla Public License 2.0. The logic is described above in enough detail to
be checked line by line against the source.
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
Ten checks before you believe a backtest — the free one-page checklist: quanticator.app/tv · No-repaint HTF Template Pack (Pine v6, $39): jordonharrell.gumroad.com/l/norepaint-htf-template-pack
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
Ten checks before you believe a backtest — the free one-page checklist: quanticator.app/tv · No-repaint HTF Template Pack (Pine v6, $39): jordonharrell.gumroad.com/l/norepaint-htf-template-pack
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.