Man + Machine: Pitcher Projections

This article is part of our Man + Machine series.

Picking up with the optimizer analysis, the focus this week was to evaluate how accurate the optimizer ranked starting pitcher by projected points.

Three days worth of data was collected. I was on a mini -- and very refreshing -- no-computer vacation for a few days, so these are only initial observations and results. As always, though, there were still plenty of telling outcomes to hit on.

Each day, I pulled the 10 pitchers with the most projected fantasy points from the two sites. Wednesday, though, I used the projections for all 15 games for FanDuel, but I only used the projections from the 10 night games for the Draft Kings list. I wanted to switch it up slightly, and see how the optimizer did with fewer aces.

I did not track the projected points to compare them to the actual points, and the actual rankings in the tables only include the 10 hurlers tracked. Though, tracking all starters and comparing projected to actual fantasy points would also be excellent avenues to explore in the future, albeit extremely time consuming ones.

So, we'll see there.

Results

Reflections

First, it is worth noting the starter with the highest projected point total never finished with the most fantasy points out of the top-10 pitchers pulled. In fact, they never finished higher than fifth.

With that said, it is a small sample size, and in five of the six sample groups there were a number of high-end starters taking the hill. Additionally, Cole Hamels threw a no-hitter, and Zack Greinke's start came after a cross-continent flight following the birth of his first child.

Plus, with a closer look, you'll see the projected top-five arms fared well and finished among the top-five ranked 19 of 30 times, which aren't bad results for fantasy sports. Also, looking at just the five full-day results, the optimizer went 17-for-25 returning top-five performances.

Two specific examples stuck out for me as I was pulling the names: Carlos Martinez and Dallas Keuchel. Saturday, Martinez was the eighth-ranked starter, and while there were plenty of aces, he had an absolutely dreamy matchup and was the second-best arm. Keuchel was a surprise absence from FanDuel rankings against the Royals on Sunday. He wouldn't have been in play for me, but it was still surprising. As it turned out, the optimizer was bang on.

I'm going to circle back on the Keuchel example later.

The average point total of the 30 pitchers pulled from FanDuel was 13.16. At DraftKings, the average fantasy point total was 20.3. Removing the five highest and five lowest scores returns fantasy point averages of 13.18 at FanDuel and 20.48 at DraftKings.

So while there were a few extremes, the averages didn't sway too far, and also, the averages are solid marks to return value in the majority of contests. When you're spending up for an ace arm you're looking for approximately 18 and 30 points on FanDuel and DraftKings, respectively.

As seen from the above results, aces have off days, too.

Tips

It seems more appropriate to look at the pitcher rankings in terms of tiers and then narrow down your selections accordingly. Earth shattering analysis, I know. Still, every day tracked, a pitcher ranked in the five-to-10 range returned a top-three showing. So you shouldn't always be blindly looking at the top options, which is even more important on two-pitcher sites.

Chris Liss has talked frequently about riding the chalk arms on FanDuel throughout his DFS Amateur Hour series, and there are no qualms in the approach here, especially after the optimizer performed well with the recommendations.

The key is honing in on the right top starter, and the No. 1 ranked or most expensive hurler isn't always the right choice. In his current form, Clayton Kershaw might be an exception, and at times this season, Max Scherzer and all the other high-end aces have shown to be next to unhittable.

However, as noted, aces have off days, too, which is why it was interesting to notice Keuchel so far down the rankings. He is a top-tier pitcher 95 percent of the time, but against the Royals on Sunday it was both telling and confidence instilling to see him excluded from top billing by the optimizer. Look out for those instances.

Additionally, look out for instances where a hurler -- Mike Friers on Wednesday -- are included. You should be skeptical and gauge whether the juice is worth the squeeze.

What's Next

The vacation put a hold on playing contests competitively, so I'll look forward to jumping back into the fire this week.

Additionally, I've continued to track the hitter values, so it'll be nice to re-examine those with a larger sample size. Plus, there is room to continue tinkering with how to best track pitchers.

As always, if there is anything else you're looking to see covered, please don't hesitate to fire away in the comments.