An Odd Paradox
Imagine you're playing as Lizzies and face this situation:
The Saurus in the yellow box only has Block, and your re-roll is already used. Would you still be tempted to make him dodge, say, to mark the Dorf blitzer?
I play rather conservatively, yet I would. Am I out of my mind? Not at all. Allow me to explain what I like to call an odd paradox.
Firstly, nothing can prevent Dorfs from scoring. Pressuring them signals that you want them to score on their next turn, after which you could attempt a 2-turn TD before the half ends. This strategy might not be a probable move, but if you don't try to make something happen on the BB field, nothing will happen.
Secondly, letting the Saurus stand there also contains a risk; it is in fact riskier than dodging, even if the Saurus has Block and neither Blocker has Mighty Blow. According to elyoukey's calculator (http://www.elyoukey.com/sac/), the chance of getting knocked down after two 2D blocks is a little over 80 %, while failing a Dodge on a 5+ is near 66 %. In fractions and reversing the odds, this means the Saurus has 1 / 5 chance to stay up after two hits, while he has 1 / 2 chance to succeed in dodging. If we consider cases with Mighty Blow and a Saurus without Block, the difference is more dramatic.
Now, be honest - did you really think that dodging with a Saurus gave you better odds than letting it take two hits?
Suppose your AG1 Saurus has Dodge. You might want him to dodge before he gets Break Tackle. How do you feel about the odds?
Suprisingly, they're quite good: 56 % or 5 / 9, i.e. 1 / 3 + (2 / 3 x 1 / 3 ). Even more suprisingly, they're the same as getting a PUSH on a 2D block without a RR. Which means that a One-Turn Touch Down (OTTD) that requires pushbacks is often less probable than dodging with a Saurus with Dodge!
Blood Bowl offers many similar situations where we make intuitive, but erroneous, estimates. Only cold calculation can help us see through these paradoxes. The only other viable alternative is to embrace a heroic conception of probability.