PFP Lab Notes

PFP Lab Notes

Archives
Log in
Subscribe
November 22, 2026

What I learned after grading other people's picks

Hi, Casey here.

Two weeks ago one of our replies got 26K views under a much bigger post, and I wanted to know if it moved the betting market. We watched 54 NHL games against rules we wrote down before we looked. By those rules, nothing moved. One of the three measures came close, but not close enough to count. So we couldn't detect anything.

That's the kind of thing PFP does now: write the rule first, then check.

Here is what grading other people's picks has shown so far.

We've graded more than 600 picks from more than 15 builders and pick accounts. You can see all of it on the Board.

About 43 out of every 100 picks closed at exactly the price they were posted at.

About 100 picks couldn't be graded. The most common reason: the line they posted wasn't on offer at DraftKings when they posted it.

We save each pick the first time our bot sees it, usually within a few minutes of the post. If someone edits or deletes a post after that, the pick we saved still counts.

I hold myself to the same bar. My first model went 50-50-2 and lost 9.7% on the bets that won or lost. I published the record anyway, and froze it the same day.

This is early. More than fifteen accounts and a few weeks is a start, not a verdict. I'll keep grading and tell you what holds up and what doesn't.

Casey Pixel Forge Plays

Don't miss what's next. Subscribe to PFP Lab Notes:
Continue the conversation:

Older → PFP Lab Notes #6: units fell, the close didn't
Twitter
pixelforgeplays.com
Powered by Buttondown, the easiest way to start and grow your newsletter.