Tardy for this party…So sorry 🙂
We are 2 hours post Sunday kick off; Thursday night came and went without my pre-blogging. This is new for me; I’ll get better. With that, here is who I played on Thursday along with my full team. This will give you a sense of play/bench choices I had.
As I mentioned, I bombed my draft. Out of the 3 leagues I drafted (1 PPR and 2 Standard), the league against my family / the one that started it all, received a grade C with a 61% likelihood of making it to the playoffs, ending the season ranked 5th (out of 10). To give perspective, last year, I was graded a B- going into Week 1, but had a likelihood of 81% of making the playoffs, ranked #2 in terms of wins and losses (11 wins, 3 losses –> and this was exactly how I ended the 2016 season) . My other two leagues graded better; both ranked as draft grade B with league 2 showing a 77% likelihood of making the playoffs, ranked #1 spot BUT with 7 wins and 7 losses (ouch). League 3 shows a playoff likelihood of 67%, ranking 3rd (8 wins, 6 losses).
All in all, I troll the waiver wire heavily and think my model has an uncanny accuracy at finding this year’s rookie-of-the-year as I mentioned in my prior post.
Going into Thursday, League 1 lineup and points scored as of 3:27pm on Sunday against my competitor. Keep in mind, as of Thursday morning, Week 1 was projected by NFL.com to go to the other person (not me), and she had gained 10 projected points between Monday and Thursday. But NFL.com projections are often incorrect. Also, my Dolphins and TB players were unexpectedly benched due to the Hurricane so that was another event previously unmodeled. Also, with Ezequiel Elliott playing in Week 1, my temporary backup (McFadden) was another forced bench. But here is who I went with:
So far, in the RB position, we nailed it (we being me and my model). Both games were played on Thursday but both players exceeded publicly available projections by a landslide. Plus of course my competitor had L. Bell, who (as of Q4) performed way under public projections. But my model flagged him as underperforming bumping up my likelihood to win this week. I also did a good job working with my subpar drafted team in terms of bench: most significantly, J. Crowder. When I updated the model, his name was flagged for underperformance and moved to bench. I had Duke Johnson in but swapped for Gillislee despite all the mixed feedback and noise online about him.
J. Reed (TE) came in 2 points under what I expected but there is still time on the clock. And my competitor has a lot of stout players on her roster to player. Hopefully D. Baldwin and A. Thielen bring home something good – My poor Seahawks defense was an emotional draft (sort of a hometown for me). Just like T Brady was the worst QB for me because I have a visceral response to his game play style evident by the points above. Hopefully he will perform at par with what everyone expects from him.
Musings on ML, Deep Learning & AI 