Parabolas were fitted to the first 8 or so n values for confidences of 95%, 99%, 99.9%, 99.99%, 99.999%. It was further decided to average the second order term and apply it to all curves after k=16. The predicted values were loaded into the scamper program. Further work on the estimation of these (k,n) pairs is required but this is a useful starting point. The important point is that the nested loops involved in their calculation prevent results for k>9 because of processing time required.
The updated scamper program was used to analyse the UDP paths where there was no clump and the corresponding TCP analysis found load balancing. The vice versa of this should probably also be analysed. The result was 13 out of 42, though this reanalysis run was some months after the original analysis. This means that 13 paths found load balancing where none was found before.
The published k,n values for k up to 16 were graphed against the data generated by the recursive program, k up to 9 fitted and extrapolated. Some longer extrapolations of all the curves were produced as well though the usefulness of these is unknown other than that the values are probably above the real values.
The paper on stopping values puts a bound of 30 nodes on the values used.