Extremes, pivots and statistics across languages

The built-in functions for highs, lows, pivots and statistics over a window of bars. Each row maps one concept from Lipi to the other languages; the word in a cell is the relation, explained in the language map overview.

Concept	Lipi	MQL4	MQL5	NinjaScript	Pine Script	thinkScript
Highest value	`ta.highest`	Differs	Differs	Caveat	Caveat	Caveat
Offset of the highest value	`ta.highestbars`	Caveat	Caveat	Caveat	Caveat	Caveat
Lowest value	`ta.lowest`	Differs	Differs	Caveat	Caveat	Caveat
Offset of the lowest value	`ta.lowestbars`	Caveat	Caveat	Caveat	Caveat	Caveat
Pivot high	`ta.pivothigh`	Caveat	Caveat	Caveat	Differs	Absent
Pivot low	`ta.pivotlow`	Caveat	Caveat	Caveat	Differs	Absent
Range of values	`ta.range`	Absent	Caveat	Differs	Caveat	Absent
Median	`ta.median`	Absent	Caveat	Caveat	Caveat	Caveat
Mode	`ta.mode`	Absent	Absent	Absent	Caveat	Absent
Percentile by linear interpolation	`ta.percentile_linear_interpolation`	Absent	Caveat	Absent	Caveat	Absent
Percentile by nearest rank	`ta.percentile_nearest_rank`	Absent	Absent	Absent	Caveat	Absent
Percent rank	`ta.percentrank`	Absent	Absent	Absent	Caveat	Absent
Correlation coefficient	`ta.correlation`	Absent	Caveat	Caveat	Caveat	Caveat

Each section below gives the Lipi construct first, then the construct in each language, with a note wherever behaviour differs. The language name links to the documentation the row was checked against.

Highest value

The highest value source reached in the last length bars. In Lipi: ta.highest. At this build its first valid bars can miss the true maximum.

Language	Construct	Relation	Note
MQL4	`iHighest`	Looks alike, behaves differently	Returns the bar offset of the maximum, not its value; read the searched MODE_* series at that offset, e.g. iHigh for MODE_HIGH. History indexing: result is a shift counted back from the current bar, reusable as the shift in iHigh.
MQL5	`iHighest`	Looks alike, behaves differently	Returns the bar shift of the highest value, not the value; read the price at that shift. The series comes from a mode enum. History indexing: Shift 0 is the current bar and counts toward older bars; -1 signals an error.
NinjaScript	`MAX`	Equivalent, with a caveat	At this build Lipi’s ta.highest can be wrong on its first valid bars and is na until bar length-1; values match after warm-up.
Pine Script	`ta.highest`	Equivalent, with a caveat	Pine also has a one-argument form that reads high; Lipi always needs the source, so expand ta.highest(n) to ta.highest(high, n).
thinkScript	`Highest`	Equivalent, with a caveat	Highest’s length defaults to 12, Lipi requires it; Lipi’s first valid bars can also be wrong, and it is na until bar length-1.

Offset of the highest value

Locates the highest value of source within the last length bars and returns its position rather than its value. In Lipi: ta.highestbars.

Language	Construct	Relation	Note
MQL4	`iHighest`	Equivalent, with a caveat	Same offset-style result. Source is a MODE_* series id, with count and start arguments; returns -1 on error. History indexing: shift counts bars back from the current bar; shift 0 is the still-forming bar, and any shift can be read in one call.
MQL5	`iHighest`	Equivalent, with a caveat	Returns the shift of the highest bar, 0 at the current bar, as Lipi’s ta.highestbars does; -1 on error. Series come from the mode enum. History indexing: Shift 0 is the current bar and counts toward older bars.
NinjaScript	`HighestBar`	Equivalent, with a caveat	Returns an int count of bars ago (0 = current bar); the page documents no value for a not-found case.
Pine Script	`ta.highestbars`	Equivalent, with a caveat	Pine returns 0 or a negative offset (-3 is three bars back). Lipi’s docs do not state the sign. Pine’s one-argument form reads high. History indexing: Pine offsets are zero or negative; confirm the sign Lipi returns before comparing with positive values.
thinkScript	`GetMaxValueOffset`	Equivalent, with a caveat	Counts bars back with 0 at the current bar, as Lipi’s ta.highestbars does. The length defaults to 25.

Lowest value

The lowest value source reached in the last length bars. In Lipi: ta.lowest. At this build it returns 0 on the first length bars after it first has a value.

Language	Construct	Relation	Note
MQL4	`iLowest`	Looks alike, behaves differently	Returns the bar offset of the minimum, not its value; read the value with iLow at that offset. Source is a MODE_* series id. History indexing: result is a shift counted back from the current bar, reusable as the shift in iLow.
MQL5	`iLowest`	Looks alike, behaves differently	Returns the bar shift of the lowest value, not the value; read the price at that shift. The series comes from a mode enum. History indexing: Shift 0 is the current bar and counts toward older bars; -1 signals an error.
NinjaScript	`MIN`	Equivalent, with a caveat	At this build Lipi’s ta.lowest returns 0 on its first length valid bars and is na before that; values match after warm-up.
Pine Script	`ta.lowest`	Equivalent, with a caveat	Pine also has a one-argument form that reads low; Lipi always needs the source, so expand ta.lowest(n) to ta.lowest(low, n).
thinkScript	`Lowest`	Equivalent, with a caveat	Lowest’s length defaults to 12, Lipi requires it; Lipi also returns 0 for its first length valid bars and is na before that.

Offset of the lowest value

Gives the position of the lowest value of source within the last length bars, not the value itself. In Lipi: ta.lowestbars.

Language	Construct	Relation	Note
MQL4	`iLowest`	Equivalent, with a caveat	Same offset-style result. Source is a MODE_* series id, with count and start arguments; returns -1 on error. History indexing: shift counts bars back from the current bar; shift 0 is the still-forming bar, and any shift can be read in one call.
MQL5	`iLowest`	Equivalent, with a caveat	Returns the shift of the lowest bar, 0 at the current bar, as Lipi’s ta.lowestbars does; -1 on error. Series come from the mode enum. History indexing: Shift 0 is the current bar and counts toward older bars.
NinjaScript	`LowestBar`	Equivalent, with a caveat	Returns an int count of bars ago (0 = current bar); the page documents no value for a not-found case.
Pine Script	`ta.lowestbars`	Equivalent, with a caveat	Reference states no sign for the offset; the docs’ highestbars example negates it, implying zero or negative values. Lipi’s sign is undocumented. History indexing: Pine offsets are probably zero or negative; confirm the sign Lipi returns before comparing with positive values.
thinkScript	`GetMinValueOffset`	Equivalent, with a caveat	Counts bars back with 0 at the current bar, as Lipi’s ta.lowestbars does, which returns -1 during warm-up. The length defaults to 25.

Pivot high

Looks for a pivot high in source, as this build implements it. In Lipi: ta.pivothigh.

Language	Construct	Relation	Note
MQL4	`iFractals`	Equivalent, with a caveat	Closest built-in: no leftBars or rightBars arguments, mode picks upper or lower, value read at a shift. History indexing: shift counts bars back from the current bar; shift 0 is the still-forming bar, and any shift can be read in one call. Other timeframes: symbol and timeframe arguments read another symbol or period directly; Lipi ta calls use the chart’s own bars.
MQL5	`iFractals`	Equivalent, with a caveat	Buffers are 0 upper and 1 lower. The page shows no left or right bar-count arguments and gives no window definition. State and side effects: Returns a handle; values are read later with CopyBuffer and are not ready right after creation. History indexing: CopyBuffer start_pos 0 is the current bar and counts toward the past; copied data lands oldest-first.
NinjaScript	`Swing`	Equivalent, with a caveat	One strength covers both sides. SwingHigh gives 0, not na, when no swing exists; SwingHighBar returns bars ago, or -1. Warm-up and missing values: Zero is returned until the bar count exceeds strength or a pivot is found. Bar timing: A swing is only definite after strength further bars, so earlier values are updated later.
Pine Script	`ta.pivothigh`	Looks alike, behaves differently	Pine returns the price of the pivot bar itself, or na. Lipi returns the current bar’s source when the window high sits leftBars back. Bar timing: Pine reports the pivot bar’s own price; Lipi’s value is the current bar’s source, not the pivot bar’s.
thinkScript		Not in thinkScript	No left/right-bar pivot function; the ZigZagHighLow study marks swing points filtered by a reversal threshold, and PivotPoints gives previous-period levels.

Pivot low

Looks for a pivot low in source, as this build implements it. In Lipi: ta.pivotlow.

Language	Construct	Relation	Note
MQL4	`iFractals`	Equivalent, with a caveat	Closest built-in: no leftBars or rightBars arguments, mode picks upper or lower, value read at a shift. History indexing: shift counts bars back from the current bar; shift 0 is the still-forming bar, and any shift can be read in one call. Other timeframes: symbol and timeframe arguments read another symbol or period directly; Lipi ta calls use the chart’s own bars.
MQL5	`iFractals`	Equivalent, with a caveat	Lower buffer (1) holds the low fractals. The page shows no left or right bar-count arguments and gives no window definition. State and side effects: Returns a handle; values are read later with CopyBuffer and are not ready right after creation. History indexing: CopyBuffer start_pos 0 is the current bar and counts toward the past; copied data lands oldest-first.
NinjaScript	`Swing`	Equivalent, with a caveat	One strength covers both sides. SwingLow gives 0, not na, when no swing exists; SwingLowBar returns bars ago, or -1. Warm-up and missing values: Zero is returned until the bar count exceeds strength or a pivot is found. Bar timing: A swing is only definite after strength further bars, so earlier values are updated later.
Pine Script	`ta.pivotlow`	Looks alike, behaves differently	Pine returns the price of the pivot bar itself, or na. Lipi returns the current bar’s source when the window low sits leftBars back. Bar timing: Pine reports the pivot bar’s own price; Lipi’s value is the current bar’s source, not the pivot bar’s.
thinkScript		Not in thinkScript	No left/right-bar pivot function; the ZigZagHighLow study marks swing points filtered by a reversal threshold, and PivotPoints gives previous-period levels.

Range of values

How much ground source covered over the last length bars, expressed as one figure. In Lipi: ta.range.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No range function; use iHigh/iHighest and iLow/iLowest over the window.
MQL5	`vector::Ptp`	Equivalent, with a caveat	Range of values of a whole vector; fill it with the window first. Lipi’s range is a per-bar windowed figure.
NinjaScript	`Range`	Looks alike, behaves differently	Range() is the current bar’s own range with no window length; Lipi’s takes a source and length. Use MAX minus MIN for a window.
Pine Script	`ta.range`	Equivalent, with a caveat	Equal after warm-up; Lipi’s first length valid bars are wrong because its lowest returns 0 there, so the range is overstated.
thinkScript		Not in thinkScript	No range function; subtract Lowest from Highest over the same length.

Median

The middle value of the last length values of source. In Lipi: ta.median.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No median function; copy to an array, ArraySort it and read the middle element.
MQL5	`MathMedian, vector::Median`	Equivalent, with a caveat	Takes an array you fill with the window first; it returns NaN on error. A Median method also exists on vectors.
NinjaScript	`GetMedian`	Equivalent, with a caveat	A plain method returning one double, not an indicator series. For an even look-back it averages the two middle values.
Pine Script	`ta.median`	Equivalent, with a caveat	Pine’s reference says na source values are ignored; Lipi’s ta.median returns na on a bar whose source is na.
thinkScript	`Median`	Equivalent, with a caveat	Median’s length defaults to 12, Lipi requires it; the middle-or-average-of-two rule matches.

Mode

Of the last length values of source, this reports the one that recurs most often. In Lipi: ta.mode.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No mode function; sort a copy and count runs in a loop.
MQL5		Not in MQL5	No mode function among the array statistics (mean, variance, skewness, kurtosis, median, deviation).
NinjaScript		Not in NinjaScript	No statistical-mode indicator or method in the system list; compute it by hand.
Pine Script	`ta.mode`	Equivalent, with a caveat	Pine ignores na source values; Lipi’s ta.mode returns na on a bar whose source is na, and float even for int sources.
thinkScript		Not in thinkScript	No mode function; Median is the nearest windowed statistic offered.

Percentile by linear interpolation

The requested percentile of source over length bars, interpolating between the two values that straddle the rank. In Lipi: ta.percentile_linear_interpolation.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No percentile function; sort a copy with ArraySort and interpolate by hand.
MQL5	`vector::Percentile`	Equivalent, with a caveat	A vector method; percent runs 0 to 100 and a linear algorithm is used. The page asks for sorted input. Fill the vector first.
NinjaScript		Not in NinjaScript	No percentile method in the system indicator or Analytical lists; compute it by hand.
Pine Script	`ta.percentile_linear_interpolation`	Equivalent, with a caveat	Same calculation; the third parameter is named percentage in Pine and percentile in Lipi, which matters for named arguments.
thinkScript		Not in thinkScript	No percentile function in the functions list or study library.

Percentile by nearest rank

Takes the requested percentile of source over length bars at the nearest rank, without interpolating. In Lipi: ta.percentile_nearest_rank.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No percentile function; sort a copy with ArraySort and pick the rank by hand.
MQL5		Not in MQL5	The documented Percentile method uses a linear algorithm only; no nearest-rank variant is described.
NinjaScript		Not in NinjaScript	No percentile method in the system indicator or Analytical lists; compute it by hand.
Pine Script	`ta.percentile_nearest_rank`	Equivalent, with a caveat	Same calculation; the third parameter is named percentage in Pine and percentile in Lipi, which matters for named arguments.
thinkScript		Not in thinkScript	No percentile function in the functions list or study library.

Percent rank

Ranks where source now stands against the last length bars of itself. In Lipi: ta.percentrank.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No percent-rank function among the indicator functions.
MQL5		Not in MQL5	No percent-rank method in the matrix and vector statistics list; count the values below the current one yourself.
NinjaScript		Not in NinjaScript	No percent-rank method in the system indicator or Analytical lists; compute it by hand.
Pine Script	`ta.percentrank`	Equivalent, with a caveat	Same ranking on clean data; Pine yields na when an na is in the window, Lipi keeps computing after the na bar.
thinkScript		Not in thinkScript	No percent-rank function in the functions list or study library.

Correlation coefficient

Correlation between two series, source1 and source2, measured over length bars. In Lipi: ta.correlation.

Language	Construct	Relation	Note
MQL4		Not in MQL4	No correlation function among the indicator functions or math group.
MQL5	`vector::CorrCoef`	Equivalent, with a caveat	A method on two vectors you fill first; it returns the Pearson coefficient. The bar window is whatever you copy into the vectors.
NinjaScript	`Correlation`	Equivalent, with a caveat	Correlates the chart series with another instrument given by name, not any two series.
Pine Script	`ta.correlation`	Equivalent, with a caveat	Pine’s reference says na source values are ignored; Lipi’s ta.correlation returns na on a bar whose source is na.
thinkScript	`Correlation`	Equivalent, with a caveat	Length defaults to 10. thinkScript can feed another symbol’s close as data2; Lipi reads only the chart symbol. Other timeframes: data2 may come from a secondary symbol, which Lipi cannot read.

Manual

Use Lipi with AI

Reference

Extremes, pivots and statistics across languages

Highest value

Offset of the highest value

Lowest value

Offset of the lowest value

Pivot high

Pivot low

Range of values

Median

Mode

Percentile by linear interpolation

Percentile by nearest rank

Percent rank

Correlation coefficient

​Highest value

​Offset of the highest value

​Lowest value

​Offset of the lowest value

​Pivot high

​Pivot low

​Range of values

​Median

​Mode

​Percentile by linear interpolation

​Percentile by nearest rank

​Percent rank

​Correlation coefficient

Highest value

Offset of the highest value

Lowest value

Offset of the lowest value

Pivot high

Pivot low

Range of values

Median

Mode

Percentile by linear interpolation

Percentile by nearest rank

Percent rank

Correlation coefficient