Randomization vs. blocking

[c10ck3r]

Okay, so (as some of you know), I currently am employed during my schooling doing research plots for seed and chemical companies (through a contractor). As I plant the actual plots, though, I am constantly reminded of the variation within a given field (which we attempt to avoid as much as possible of by staying off terraces, etc.). So, my question is, is our method the best way of eliminating variation, or is there another that would give us better results? For this exercise, assume that we are contracted to grow 400 varieties of soybeans at a given location, run in triplicate.

Currently, our plot layout program will place the three reps one behind the other, with randomization within each block, thus basically having three separate, 400 variety chunks set in front of each other.

The other option (in my mind) would be to have the 1200 plots all grouped together and randomized.

We are, of course, putting border around the edges either way to eliminate the edge effect, and are harvesting the middle of four rows for each variety unless there are extenuating circumstances (say, row 3 of 4 did not plant due to the tube coming out during planting, we would harvest rows 1 and 2 and make note).

Any tips, suggestions, feedback appreciated, my Elements of Statistics course was severely inadequate in teaching this kind of maths unfortunately...