This is not possible in Stratego, or any incomplete information game.
If it were , there would be a computer program who could beat platinum players , and this is far far away from the truth.
First, in 1999, computers (apart from Deep Blue, and even there Kasparov didn't play his best in the 1997 match) were also very far far away from beating chess grand masters. Only since 2005 and further (Fruit, Rybka and now Stockfish, Komodo and Houdini) did the top engines far outplay humans. Yet humans did play with such material guidelines for decades (maybe not as refined as in the Kaufman article).
Second, I think there are more reasons than incomplete information for Stratego computer programs to be not very strong (board size and repetitions extending the horizon effect are mentioned in Vincent de Boer's thesis). The main reason IMO is that the game is simply not as popular as others (chess, go, even draughts/checkers). As illustration, in other imperfect information games, such as poker/bridge/scrabble, top programs have equalized with or even outplayed human champions nowadays.
Third, a pure material only analysis will not be enough of course. But supplemented with material/information trade-offs, I don't see why simple guidelines couldn't be inferred from game databases such as from Gravon.
Simple question: is picking up a known opponent piece with a known piece of one rank higher (lieutenant with captain, captain with major, etc.) always good? Where does it stop? Picking known colonel with known general? Or when does picking up a known piece with an unknown piece become bad? Picking known captain with unknown major? There must be at least some guidelines.
Finally, in e.g.Texas Hold'em poker there are also pre-flop hand rankings. So imperfect information games are open to analysis. Of course, position, number of players and of course the later streets change everything, but denying that basic analysis is impossible or saying things like "it depends on the position" is not very productive. What is productive is trying to formulate how it depends on the position.
Edited by TemplateRex, 10 December 2017 - 02:33 PM.