Week 5 NFL Daily Fantasy Football Results - Model Still Improving, Yet My Worst Week Ever (-$158.00)

• Week 1 Results
• Week 2 Results
• Week 3 Results
• Week 4 Results

$0.00

As in, nothing. Nada. Zilch. Big fat zero. I have been sitting here for the past hour or so just marveling at my own inefficiencies.

I know this whole thing is a process, and I realize there still is much to be learned in a loss. But, it is just hard not to feel deflated and embarrassed after seeing all my best intellectual efforts go to waste this week.

My week 5 ended up with me losing all $158 of my lineups split 70% in H2H and 50/50 competitions, 30% in GPP's. I got to about an average lineup of 138 points this week which isn't terrible, but obviously is not going to do it if I ever hope to consistently cash in the long run.

This puts my season at a $190 loss overall, or -35% profit.

However, my overall algorithm r^2 has increased to well over 28% and climbing. So what gives? Well, just because I am getting better at picking players doesn't mean I am getting better at picking lineups. I had several inefficient choices in players that were okay, but just not the home runs I needed. Guys like Emmanuel Sanders that performed up to my expectations, but not over them. These guys have been killing me.

But, the largest problems are my misses, of which I have had many. Guys like Zach Ertz, Dontrelle Inman, and Steve Smith that I have been falling in love with for salary and matchup purposes and under perform greatly.

My Best Lineup - 165 Points

This lineup was in a GPP Thursday-Sunday ticket that just barely fell out of the money. This is a theme that haunted me this week. I had some excellent selections here, but chose to fill in with guys like Sharpe and John Brown for salary purposes that just way underperformed. Sucks to be out of the money when a lineup hits 165, but those are the breaks.

My Worst Lineup - 107.3 Points

My Ravens plan that obviously backfired greatly on me. I felt they had a great match this week, Flacco was rated high on the algorithm, and I felt pretty confident in my low salary fill ins. I placed this in $30 worth of H2H and 50/50 tickets. Every single one lost.

My Heartbreaking HUGE 50/50 Losses - Only 3-4 More Points Needed For a $120 Swing

I entered both of these lineups exclusively in 50/50 tournaments spread across $60. As of the Sunday Night game, both were cashing. As soon as OBJ caught his garbage time touchdown, the cash line moved to 149 in both fields and I was left out on the doorstep.

I really loved these lineups for 50/50, specifically because of what I felt at the start of the day were pretty safe plays. Was not expecting Inman to blank, although in retrospect he was probably a terrible play with Gates coming back to make 5 solid options for Rivers. I am still super disappointed in Ertz, he had what I still feel was a recipe for success, just had a bad game. And, the Ravens just underperformed by all metrics. Hard to say if I was just unlucky in these spots, but the way the week went overall, it probably was just my overall poor lineup construction that sealed the deal. I should have opted for guys with better injury situations like Cameron Meredith than guys like Steve Smith and Inman in already crowded situations. This is something I will take into great consideration in weeks to come.

Inside the Data

Below is my box plot of the differences between predicted and actual results for Week 5, 2016 with examples of players that fell in each range. The majority of my results were between a 1.2 point median overestimate and a 4.5 point third quartile overestimate.

My "Normal" Distribution

Assuming all of my differences between predicted and actual are normally distributed (think bell curve), about 68% of my predictions lie on the interval [-8.34, 7.29]. This is about where I have been over the past two weeks, but the range is trending inwards. This means I am getting much better at narrowing down potential game performance ranges on players. Again, negative numbers here mean underestimates, and positive numbers mean overestimates.

Results by Player -- Week 5, 2016

I have to be disappointed with my major losses this week, but I also have to make the choice to learn from them and just move forwards. I have already picked up on a few extra signals in multiple linear regression testing that is promising for my Week 6 projections. I am also trending upwards in my overall r^2. And, I am getting very close to Week 8, the week where my backtesting has proven profitability. If I can just hold on until then, I am very optimistic about the weeks to come! I hope you continue to ride along with me!

---

GIF Source

1. GIF - Willy Wonka

---