\documentclass[a4paper]{article}
\usepackage{amsmath}
\begin{document}

{%\obeylines
%\let\par=\cr
\halign{\hfil$#$ &\hfil$#$ &\hfil$#$
 &\hfil$#$ &\hfil$_{#}$ &\hfil$#$ &\hfil$#$
  &\hfil$#$ 
&\hfil$#$
&\hfil$#$&$#$\hfil
&\hfil$#$ &\hfil$#$&$#$ \hfil
&\hfil$#$&$#$\hfil
\cr
n&C&A&D&\omit$D$\hfill&T_{\min}&T_{\max}&ST_{\min}&E_{\min}&\omit\span ST_{\mathrm{avg}}&ST_{\max}&
U&&R\cr
1&1&1&1&0,0,1&1&1&1&1&1&&1&1&&1&\cr
2&2&2&2&0,1,1&2&2&2&2&2&&2&2&&2&\cr
3&5&12&6&1,1,1&4&8&3&3&4&.667&6&4&&4&.5\cr
4&14&55&18&1,1,2&11&30&7&7&11&.510&18& 8&.666667&10&.6\cr
5&42&273&60&1,2,2&32&150&14&14& 29&.313 &60&19&.333333&25&.6\cr
6&132&1428&222&2,2,2&96&780&27&27& 75&.799 &222&44&.200000&62&.828571\cr
7&429&7752&794&2,2,3&305&4550&58&58& 197&.617 &794&102&.733333&155&.958929\cr
8&1430&43263&2988&2,3,3&&&127&127& 517&.247&2988&241&.834921&390&.444048\cr
9&4862&246675&11856&3,3,3&&&&266&&&&574&.914286&983&.978857\cr
10&16796&1430715&45580&3,3,4&&&&&&&&1377&.587302&2492&.993468\cr
11& 58786& 8414640&180960&3,4,4&&&&&&&&3322&.336508&6343&.812317\cr
12&208012&50067108&743160&4,4,4&&&&&&&&8055&.810467&16201&.746633\cr
\noalign{\hrule\smallskip}
\omit\rlap{limit}\hfil
&4.0&6.75&4.5&&&&&&&&&\approx 2&.48\ldots&\ge\approx 2&.65\ldots\cr
}}

\parindent0pt \advance\parskip by 3pt

$n$ inner points in a triangular convex hull

$C$ = Catalan numbers

$A$ = abstract stacked triangulations = ternary trees with $n$ inner
nodes and $2n-1$ leaves = $\binom{3n+1}{n}/(3n+1)\approx 6.75^n$
% stmin.py

$D$ = 3 chains, of lengths $_{i,j,k}$. The limit exponent of 4.5 has
been established by Marc.
% 3chains.py

$T_{\min},T_{\max}$ all triangulations of a point set (Oswin, from the
database)

$ST_{\min},ST_{\max}$ stacked triangulations of a point set (Oswin, from the
database)

$ST_{\mathrm{avg}}$
Average over all realizable order types of the given
cardinality (Oswin)

$E_{\min}$ %,E_{\max}$ 
lower %/upper
 bounds on stacked triangulations (G\"unter)
% bestperm.py
It seems that the lower-bound examples (for $n$ a multiple of 3) look
like a series of nested concentric
triangles, where successive levels are rotated by 180 degrees.
The points lying close to a line through the center (they lie on both
sides of the center) are probably not uniformly ``curved'' but they
lie in such a way that a line through two such points cuts the points
between them evenly. It remains to define such a family precisely and
count
the stacked triangulations for this family.

$U=$ something random
defined by the recursion
$U_n = \sum_{i=1}^n U_{n-i}\cdot
   \frac{\sum_{j=1}^{i-1} U_{j-1}U_{i-1-j}}{i-1}  
$.
(For $i=1$, the value of the fraction is taken as 1.)
%     ( sum(f[j-1]*f[i-1-j] for j in range(1,i))/float(i-1) if i>1 else 1)
%     * f[n-i] for i in range(1,n+1))
% rnadperm.py
Maybe this is something related to the degree-3-vertices?


$R$ = average number of stacked triangulations on a random set, according to
Emo's recursion
$R_n = \sum_{i+j+k=n-1}R_iR_jR_k\cdot \frac 2{n+1}$.
% rnadperm.py


Prob[the balanced stacked triangulation with $n$ inner vertices can
be embedded on a random point set] $\approx 0.61886974^n$ (when $n$ is
of the form $(3^k-1)/2$).
\\
This is probably $>$
Prob[any other fixed stacked triangulation with $n$ inner vertices can
be embedded on a random point set].

\end{document}