Table 4 How well does model selection work? Description of distributions when selecting models according to the FRST potentials and the SVM. $\overline{ni}$ is the relative frequency of targets for which a selection procedure suggests improved models. $\overline{fim}$_<, $\overline{fim}$₌, and $\overline{fim}$_>, are the relative frequencies of selected models with decreased, equal, or increased TM-score quality, respectively. min qim and max qim are the minimal and maximal quality improvements achieved per target. $\overline{qim}$ is the average quality improvement over all targets. qim is the average quality improvement for the targets that the selection procedure suggests improved models for.

$\overline{ni}$

_<

$\overline{fim}$ ₌

$\overline{fim}$ _>

min qim

max qim

$\overline{qim}$

qim

0.51

0.23

0.50

0.27

-0.51

0.35

0.0016

0.0031

0.40

0.14

0.61

0.25

-0.21

0.29

0.0064

0.0160

$\overline{ni}$

$\overline{fim}$ _<

$\overline{fim}$ ₌

$\overline{fim}$ _>

min qim

max qim

$\overline{qim}$

qim

0.70

0.35

0.31

0.34

-0.43

0.29

0.00047

0.00068

0.58

0.22

0.43

0.35

-0.27

0.35

0.00774

0.01339