I don't particularly like marking IA's either. Except good examples with interesting research questions, written by my own students of course. A couple of weeks ago I posted an example of an investigation for subscribers to mark. The idea was to enter marks in a google doc so we'd get some stats to look at. Only 4 people have filled out the form so far and to be honest I only got around to doing it myself today. With 4 bits of data (+1 for me) the stats are not going to be very significant but I can say the spread is not too huge (±2.5).

You can improve the reliability here -  Investigation marking exercise, but do it without looking at my marks which I have now posted below the exercise.