I looked at buying on a surprise after the prior report being negative, and it was not as good as when both were positive. Here is a new monthly screen using the last two eps surprise periods...

The screen uses qseps.s which is one of three Quarterly Surprise (a.k.a. announcement day) fields that could be used:

epslr.s [SI Quarterly Surprise-EPS]
qseps.s [SI Quarterly Surprise-Est]
qsdiff.s [SI Quarterly Surprise-Difference]

qseps.s is the quarterly earnings estimate for the latest reported fiscal quarter (i.e. the EPS estimate on announcement day). The screen ignores actual EPS, and instead uses past estimates; looking for qseps.s 20% higher than the previous quarter.

A 20% increase in epslr.s is 20% earnings growth.
A 20% increase in qseps.s is 20% estimates growth.
A 20% increase in qsdiff.s is 20% growth in earnings surprise.

==== SIP field descriptions ====
Quarterly Surprise-EPS
Data Table Name: EPS_LR
The latest reported quarterly earnings. The quarterly period matches the Quarterly Surprise-Est period.

Quarterly Surprise-Est
Data Table Name: QS_EPS
The quarterly earnings estimate for the latest reported fiscal quarter.

Quarterly Surprise-Difference
Data Table Name: QS_DIFF
The difference between the actual reported earnings and the consensus earnings estimate for the most recently completed fiscal quarter. It is calculated in the following manner: Actual (announced) EPS – EPS Estimate