Posted by chainsaw4 on 9/13/2017 2:59:00 PM (view original):
Looks like your algorithm needs tweaking...
Why? Because you beat UGA in OT (which is basically a coin-flip) creating my first incorrect prediction for your season? Nah.
Over the course of an entire season of games for all 120 D-IA teams, I'm historically around 83% correct projections. That's 5 out of every 6 games that my algorithm gets right for all 820 or so games played. (The success rate is higher for just human-human or human-sim games; those damn sim-sim games are so wildly erratic) I'll take that every time and I doubt even the most astute observer could best that mark.
Even if you exceed my 3-10 projection by a wide margin, that will not form the basis for necessary change. I'm looking on the macro scale at my success (or not), so micro inaccuracies are of very little concern to me. But let's wait until the end of the season before we even decide just how wrong my algorithm was about Colorado State.
And if you wildly deviate from my projection AND there's something that I can learn from that, you can believe I will incorporate that into my projections. I'm constantly looking at how I can improve this thing. What's funny is that sometimes my projections are self-defeating on small scales. I've heard direct feedback from multiple coaches who've gotten a chip on their shoulders when I didn't rate or project them high enough, so they went on a coaching rampage and "broke" all the assumptions that underlie my algorithm for them/their team. If so, good for you!
9/13/2017 3:29 PM (edited)