During the trial, it was undisputed that Trende’s computer simulation and resulting statistical analyses were only as reliable as the computer model he used, and the inputs and controls he incorporated therein. Yet in response to respondents’ criticism that he did not run a sufficient number of simulated district maps [5.000], Trende generated 10,000 additional maps, which confirmed his prior results. And in response to respondents’ criticism that he did not take into account certain criteria that the legislature was required to consider under article III, § 4 (c), he reran the simulations to use additional constraints, which also confirmed his prior results. In fact, the only criteria respondents now identify that Trende did not take into account was “communities of interest,” which he explained were notoriously difficult to account for because of their vague definition, a fact conceded by one of respondents’ experts. It is implausible that the failure to account for this one criterion in the simulated maps coincidentally resulted in showing that the enacted map had all four republican-leaning districts being more republican-leaning, all the next nine most competitive districts being more democrat-leaning, and the “safest” democrat-leaning districts falling within the range of the simulated maps.
Respondents also questioned whether Trende’s simulation for the congressional map had relied on redundant, i.e., repeated or partially repeated maps. Further, it was undisputed that Trende did not review, and did not submit for review, the simulated maps, and instead based his analysis on the aggregate data generated from those maps. We note, however, that respondents failed to object to the alleged redundancies or to the absence of the simulated maps, and that none of respondents’ experts presented their own competing simulation reflecting how the results might have changed had Trende conducted his model in a manner that they opined to be more appropriate. Moreover, none of respondents’ evidence contradicted the overwhelmingly consistent pattern Trende’s model produced. Respondents seem to argue that no computer simulation could ever account for all of the constitutional criteria, and thus petitioners would need to show proof of a “smoking gun” to prove partisan gerrymandering, such as an admission by the map drawers that they intended to favor a certain political party. Such direct evidence is rare and certainly not required for petitioners to meet their burden. Rather, petitioners may rely on circumstantial evidence (see League of Women Voters of Ohio, — NE3d at —), which, as established in this case, is just as probative to establish that the redistricting plan was drawn with partisan intent.