My programming partner and I spent most of the free day yesterday in the campus cafeteria studying, and particularly practicing, logic for the exam today. Even though I had thought there was nothing left to do except wait, in fact these five hours or so were extremely helpful in doublechecking a few concepts and getting practised, and hence faster, in the standard routines that were sure to come up in the exam, like the resolution algorithm.
Unbelievably I slept really well before this third exam, even though I expected it to be rather hard. First-order logic can quickly get nasty, what with the five or six transformations a formula needs before you can even start resolving, and resolution itself is full of minor detail and thus quite easy to get wrong.
But in fact the exam was entirely doable. All problems were small and manageable, so I never really broke a sweat. Mind you, there were plenty of them, like around two dozen or so, and I really needed almost the entire 90 minutes to solve all but two. On a minor thing I had absolutely no clue, so I just ignored it. The other problem was rather complex (a reduction proof using the halting problem for Turing machines), and I had only a rough idea how to solve it, but since there was some time left, I quickly scribbled nearly a full page of prose and sketched a vague machine hoping for a few grace points. This tactics had worked for me in the Math 101 exam.
All in all I am fairly confident of a good grade. In most exams the maximum number of points well exceeds the number needed for an A+, called the “grading ceiling” (Bewertungsgrenze). Usually this goes with the suggestion that you can exclude a few problems right away, but the more usual practice is to try and solve most or all problems so to compensate for a few casual mistakes in each one. I call it the shotgun tactics–a rough shot at all problems usually is more cost-effective than making double sure of getting fewer problems right, because no professor will give you no points at all if you at least tried. In any case, in math traditionally the points buffer provided by the grading ceiling is extra large. In Math 101, an A+ was valued at 100 points out of 120, and our logic professor announced the buffer would be in the rough dimension of the VAT rate. In Germany that rate is currently at 19%, so we might need only 80 to 85 out of 100 total points in today’s exam for an A+. Though his favorite quip is that he wouldn’t give any guarantees as to which VAT rate applied: There is also a reduced 7% rate for some goods. Depending on which one it is, I might get an A. In any case I am fairly certain of a B.
Unlike programming on Monday, thus, this was a very fair exam, more or less predictable and quite doable. Programming however still rankles and is the favorite topic of debate even right before or after other exams. Hardly any of us doesn’t feel that whatever the final outcome it is extremely unlikely to be a fair reflection of his or her programming abilities. Many of us would be quite curious as to how long the professor himself would need to solve his own exam. Reportedly the technical assistant had a look at it and approved. It can hardly have been a very thorough look, and it’s quite unlikely he tried to actually solve the problems. One of my co-students said her high-school teacher had the maxim that if the students had 90 minutes to do an exam, he must be able to do it in 30. I wish our programming professor had heard of that one.
Mind you, a few of us sent him an email right after the exam saying it was way impossible to solve even half of the problems in the allotted time. He said we shouldn’t worry. If he had in fact screwed up (something he is always ready to admit) it was usually possible to “iron it out” to some extent in the grading process. But “some extent” would have to be a really drastic adjustment in this case. In his exams, the grading ceiling is usually at 95% of the points, which is the textbook definition of an A+, so in fact there is no buffer at all. Considering that the best of us barely managed to complete the solutions to the 3 to 4 minor ones of the 6 problems in the exam, he would have to lower the ceiling to 50 to 60% of the points, and I don’t think that’s likely to happen. But we will see. Just in case, we will contact administration to find out if there is any recourse should he stick to his original grading scheme and fail 90% of the course. I do doubt there is, but there is no harm in finding out. Unfortunately, the most likely outcome is that he will make some adjustments so that a majority of us pass with a poor to average grade. I wonder if I wouldn’t rather fail and take the exam again with a different professor.
Now there are three free days before the final two exams, databases on Tuesday afternoon, and automata theory on Wednesday afternoon. And then we will have survived the infamous second term. And in fact our first year of studies will come to a close.