Conventions in mpltern¶
Axes and Ticks¶
In a ternary plot, three variables which sum to a constant
t + l + r = ternary_scale
(= 1 in mpltern by default) are projected onto
a twodimensional triangle.
Each variable is associated with each corner of the triangle, and the value is
represented by the scaled distance to the corner from its opposite side.
There may be two kinds of perspectives to read a ternary plot; the
“cornerbased” and the “sidebased” perspectives.
Mpltern adopts the “cornerbased” perspective.
In this perspective, each of the three variables is associated with a corner of
the triangle, and the position in the triangle is given as the scaled distance
to the corner from its opposite side, as already written above.
In mpltern, the order of the variables is T (top) → L (left) → R (right)
(counterclockwise).
In mpltern, by default, the ticks are shown to the right side of the triangle
with seeing the corresponding corner upward.
You can put the ticks to the opposite sides by ax.opposite_ticks(True)
.
Notice that, although the tick positions are changed, still a point in the
triangle corresponds to the same composition (see the red point).
Some people read a ternary plot with the “sidebased” perspective like the figure below. In this perspective, we must also specify if the ticks proceed in a clockwise or a counterclockwise manner. Be careful that, in the “sidebased” perspective, a point in the triangle corresponds to different compositions between the clockwise and the counterclockwise tick progresses, which may be confusing (see the red point). This also means that a position for a side is not immediately associated with the value of the corresponding variable in the “sidebased” perspective. To avoid such confusions, in mpltern, the “cornerbased” perspective is adopted.
The discussions here and here may be also helpful to understand.
Convention in Other Software¶
In the cornerbased perspective, existing codes for ternary plots give the following orders by default:
Code 
Order of triangle corners 
Ticks 

Plotly 

CW 
pythonternary 

CCW 
ggtern 

CCW 
Ternary (R) 

CW 
d3ternary 

CW 
PGFPlots 

CCW 
Veusz 

CCW 
ternaryplot.com 

CW 
Delta Plotware 

CCW 
Dplot 

CW 
Grapher 

CCW 
JMP 

CW 
Origin 

CCW 
Statgraphics 

CCW 
In mpltern by default:
T → L → R
(CCW) for the order of triangle cornersCCW for the ticks progress
Normalization in Ternary Plots¶
In most plotting methods in mpltern, the given three variables are
automatically normalized by ternary_scale
.
By this convention, the three variables can be treated on an equal footing.
The exceptions are the spanplots like ax.axtline
.
Since in these methods only one of the variables is given, in principle it is
not possible whether the given value is already normalized or not.
To avoid confusions, it is strongly suggested to normalize the three
variables beforehand outside mpltern.