Jahmyr Gibbs
3+ catches
Behind the pick +
Expected average: 5 catches. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
I turn curiosity into things you can use.
Interactive products. Experiments with data. A little design, a lot of figuring it out.
Explore the workA grillz configurator that makes a custom piece feel real before it exists. Pick individual teeth, explore materials, and build a design that is yours.
Open the configurator
28 individual teeth. Mixed materials. 96 opal colors.
Immediate previews, fit guidance, and shareable designs.
Product design · Interaction · Visual systems
Portfolio demonstration · Ordering disabled
Football research, distilled into a focused set of game and player picks. The strongest estimates, the reasoning behind them, and the limits of what we know.
Explore the picksOctober 11, 2026
A preserved pregame forecast from the research archive. These are not today’s picks.
3+ catches
Expected average: 5 catches. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
25+ receiving yards
Expected average: 60 receiving yards. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
175+ passing yards
Expected average: 245 passing yards. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
5+ catches
Expected average: 7.5 catches. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
5+ catches
Expected average: 8 catches. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
1+ rushing / receiving TD
Expected average: 0.9 rushing / receiving TD. Based on 4 prior games.
Archived analyst scenario. Conditional on an ordinary workload; not calibrated.
Estimates apply to the exact threshold shown. They are experimental and uncalibrated, assume the player is active with an ordinary workload, and are not guarantees.
The October 11 archive preserves the original analyst scenarios. New scheduled editions use a separate four-game baseline derived from THE.PREDICTOR’s research. The baseline does not reproduce the manual scouting adjustments.
Game leans rank by projected score separation, with a three-point minimum. Player picks rank by estimated probability at the displayed threshold, with one selection per player.
Only earlier completed games are used. Incomplete player histories are excluded. Chosen distribution assumptions are not calibrated confidence. Injuries, current lineups, and changing roles remain limitations.
The historical geometry experiment did not pass its improvement tests and is not promoted here. The web baseline has no completed out-of-sample validation period or measured hit rate to report.
Player data: nflverse ↗All times Pacific. Picks are frozen between editions. No live revisions.
I’m interested in the space between an idea and a working thing.
My work brings together design, code, data, and AI-assisted development. The goal is to make something useful—and make the thinking behind it visible.
Take another look