Start a project
All work

Case study · Founder venture

National Gridball League.

I’m designing Gridball, a new team sport inspired by boxball, and the systems a league needs to run it. A sport with no history has no evidence to lean on, so I built ways to test decisions before anyone plays. This case study follows three of them.

NGL broadcast open: the chrome NGL logo and a red ball in a lit arena
Venture
National Gridball League, LLC
Years
2024 – now. First rulebook in 2024, league formed in 2025
My role
Founder and Commissioner: sport design, product, design direction and every ruling on this page
Built with
AI coding agents (Claude Code and Codex), directed and reviewed by me
Status
Concept, on Rulebook V3.9. Playtesting comes next
Visit
nglgridball.com
The problem
Every rule, tool and screen had to be decided before a single real match. There was no season to learn from, no players to ask and no scoreboard that understands the game.
My approach
Study how established sports and their vendors solve the same problem, test rules in a match simulator, and write the answers into software and the rulebook together.
Where it stands
Rules tested on thousands of simulated matches, a scoring console rebuilt after an audit, a broadcast system that runs in a gym, and a public site. Nothing has met real players yet.

Three decisions in this case study

  1. Make the big scores feel earned, like a touchdown.
  2. Make the score impossible to get wrong at the table.
  3. Make a sport nobody knows readable, in a gym.

The sport in one minute

Gridball grew out of boxball. At my Miami arts high school we played regular boxball in a seventh-floor dance studio every single day. It was our school’s sport. Gridball is my answer to a question that game left me with: what could it become if it were designed for competition, with teams, officials and a league? It’s still a concept, and playtesting comes next.

  • The Grid. A 24-foot square of four equal quadrants: Prime, Major, Minor and Entry. Two teams, two players each on the Grid, from co-ed 14-player team sheets.
  • The rally. Prime serves. Every return goes to another player’s quadrant after one bounce. Miss, and you’re eliminated.
  • The Z. Players below the gap move up, Entry to Minor to Major to Prime, and a teammate comes in at Entry.
  • Scoring. An elimination is 1. Holding the top of the Grid for four rounds earns the big scores, 3 and 7. Those are where the first decision begins.
How to play on nglgridball.com: Follow the Z, the route from Entry to Minor to Major to Prime Interactive rotation explainer: choose the eliminated quadrant and step through the exit, rotation and the Next In player arriving
The public How to Play page teaches rotation by letting you run it: pick a quadrant to eliminate and step through what happens. See it on nglgridball.com.

Decision 1 · Rules

Making the big scores feel earned

A Gridlock is worth 7. I wanted it to feel like a touchdown or a three-pointer: rare, earned by a team, and something a crowd can see coming. Under the original rule, one player holding Prime for four straight rounds scored it.

A rule like that can’t be judged by reading it, so I built a match simulator to play it out. It runs full matches with elite rosters, coaches calling plays, Next In choices and timeouts, and scores every round with the console’s own code. No two matches are the same, and none are scripted. The first studies said the rule was broken:

41Gridlocks a quarter in the first model
56%of all points from the two big awards
15.6%of rounds where the Prime player was eliminated
Gridlocks and point share from 500 simulated matches under the original rules; the Prime rate from an audit of the same model. A score that happens 41 times a quarter isn’t a touchdown.

Finding the cause

The tempting fix was to cap the award. I looked for the cause first. An audit of the simulated serve found a hard, fast drive with no bounce in Prime and almost no risk of a fault. It ended about three in ten rounds on its own, so Prime almost never lost. That isn’t a scoring problem. It’s a gameplay problem, and I set a rule for myself: fix how often the big scores happen through how the game is played, never through a technical patch on the score.

Options I weighed

OptionWhat the simulator showed, and the outcome
Cap the award at once per player per stayGridlocks fell by half, to 21 a quarter. Turned down: the 7 and the 3 are team scores, and a cap per player makes fans follow one player instead of the team.
Hand possession to the other team after a GridlockThe closest games of any option. Turned down: the other team answered 82% of Gridlocks right away, so games became ping-pong.
Fix the serveAdopted. The serve became underhand and must bounce in Prime first. An illegal serve gets a warning, then a point to the other team. The Prime player is now eliminated in about one round in five.
Make the Gridlock a team scoreAdopted in V3.9. A Gridlock now takes one team holding Prime and Major with the same two players for four rounds. One player holding Prime alone is the 3-point Grid Hold.
The first two were measured on 500 simulated matches in the earlier model, before the serve fix. I kept the point values and the four-round count as designed.

What changed

I tested the new definitions on 300 matches in the updated model before they went into the rulebook.

300 simulated matchesBefore V3.9After V3.9
Time between GridlocksAbout every 2 minutesAbout every 6 minutes
Quarters with no GridlockNoneAbout 1 in 5
Longest wait for a Gridlock7 minutes23 minutes
Points from Grid Holds and Gridlocks47%34%
Median winning margin57 points38 points
Stronger team wins59%53%
A Gridlock became rare enough to feel earned, and games got closer. These are model numbers, and the first playtest will recalibrate them.

A rule change in a league touches everything. This one reached about 300 files across the website, the scoring console, the mobile app, the printed books and the simulator. I wrote one specification, recorded thirteen follow-up rulings it raised, and tracked every file against it. It shipped to the site, the console and the app on November 2, 2026. The graphics changed too: the Gridlock takeover now shows the pair that earned it.

The original Gridlock takeover: one player with plus 7 and four round markers
BeforeOne player, four rounds in Prime
The V3.9 Gridlock takeover: Gridlock plus 7, two players named, four straight rounds in Prime and Major
AfterTwo players, four rounds in Prime and Major
The Gridlock takeover before and after V3.9.
A trailer cut from one simulated match. The simulator drives the real console and display pages, so every film is also a test of the scoring and broadcast software.

Decision 2 · Console

A console that can’t get the score wrong

No scoreboard on the market understands Z rotation or the Grid’s awards, so I built one. The person using it sits at a folding table in a school gym, scoring live on a laptop or iPad while a crowd watches the display it drives. One wrong tap changes a result.

The first version, from March 2026, worked in demos. Before trusting it with a real match, I had it audited against how professional scoreboard consoles work. The Daktronics All Sport 5000, a standard in gyms, sets clear conventions: separate start and stop controls, a visible running indicator, confirmation before editing a stopped clock, and recovering the previous game after a power cut. The audit, run by an AI agent against the code, reproduced five serious defects:

  • Undo could leave the clock showing as running with nothing timing it.
  • Points scored in practice mode carried into the live match.
  • Eliminations could be recorded before the match started.
  • Pausing and resuming counted a second round.
  • The same player could stand in two quadrants at once.

Rulings the design couldn’t make

Some of the bugs were really unanswered rules. When does a round end? What happens if the clock runs out mid-rally? Those can’t be settled by a screen layout, so I ruled on them as Commissioner and updated the rulebook to match:

  • A round ends only at an elimination.
  • If the clock runs out mid-rally, the rally finishes.
  • Timeouts happen only between rounds.
  • A penalty elimination uses the normal rotation, and an ejected player leaves the Grid and can’t return.

What I changed

  • A layout built on the conventions. A match strip that always shows the score, the clock and whether it’s running. The Grid in the center, matching the official diagram. Next In beside each team. A control rail where pausing can never start a new rally. Practice runs as its own session.
  • Every action is an event. An append-only log, so undo, corrections and the final record all come from the same history.
  • It survives a crash. A live match recovers on reload.
  • The display has to answer. The board acknowledges each update, and the console warns the operator when the display falls behind or drops off.
  • Tests before every change. The scoring engine has 122 automated tests that run before any change ships.
The March 2026 console: a dense grid with Eliminate buttons, rosters, an event log and a column of small manual-score and challenge buttons
BeforeMarch 2026: Start New Round and Pause side by side
The current console: score strip with the clock, the Grid with a Grid Hold count in Prime, rosters with Next In boxes, match controls and the activity log
AfterNow: one Start Round control, a match strip, and the activity log under the Grid
The operator’s console before and after the audit. The current capture is a demo match between two demo clubs.

Decision 3 · Screens

Readable in a gym, for people who’ve never seen it

Nobody watching a first match will know the rules. And no Gridball arena exists. Early matches will be in school gyms and rec centers, on a taped court, with a TV or a projector as the display. So the screens have to teach the game as it happens, on hardware the venue already has.

I looked at how a newer pro sport handles this. Major League Pickleball’s venue screens put big scores, team crests and player names on a dark field. Scoreboard vendors keep the scoring software separate from the display software. My first arena display, from March, read like a web page that happened to be on a big screen.

The arena display

  • A fixed 16:9 canvas. It scales evenly to any TV or projector, with nothing that scrolls or rearranges.
  • Score and time first, the Grid second, everything else third. The square Grid stays in the center, with the league logo above the clock.
  • Show what’s at stake. The hold count sits in the Prime cell, so the crowd sees a big score coming before it lands.
The March 2026 arena display: concept crests, scores, a small Grid of four cells and a text line for Gridlock alerts
BeforeMarch 2026: the Grid small, the stakes in a line of text
The current arena display: the logo over the clock, both scores with timeout dots, the Grid with who holds each quadrant and the Grid Hold count, fouls and Next In
AfterNow: a 16:9 canvas, the Grid in the center, the count in Prime
The arena display before and after. The current capture is a demo match between two demo clubs.

The broadcast

  • Build the moment before it lands. The score bug tracks the pair’s Gridlock count, calls a Gridlock watch at three of four, then scores it. The arena gets its own “one round from Gridlock” moment.
  • Graphics from the event log. Graphics fire from what actually happened at the table. The arena display shows every moment; the stream shows only the big plays, one at a time.
  • A producer can’t change the score. Graphics run on their own feed, separate from scoring. In testing, forged score updates from the graphics side were refused everywhere.
  • It works without internet. Gyms don’t promise Wi-Fi that works, so the console, the producer app and the stream overlay can link over the venue’s local network with a pairing code.
  • Sponsors get proof. The console logs every second a sponsor is on air, counting only while a screen confirms it’s receiving, and exports a report per sponsor.
  • One identity. One league typeface from the website to the on-air graphics, and color from the NGL brand and each club’s own colors, not a different color for each kind of score.
Score bug: Home 2, Away 0, with the Gridlock count 2 of 4 across the Prime and Major cells
2 of 4The pair’s count across Prime and Major
Score bug with a Gridlock Watch banner: number 1 A. Rivera plus number 3 J. Chen, survive this round for plus 7
Gridlock watchOne round away, both players named
Score bug after the Gridlock: Home 11, Away 0
ScoredPlus 7, once, to the team
Arena moment: One round from Gridlock, number 1 A. Rivera plus number 3 J. Chen, 3 of 4 in Prime and Major
A Gridlock as a viewer meets it: the count, the watch, the score, and the arena’s own moment at three of four. Demo match.
Match open, team intro, starting lineup and the in-match score bug. Arenas, players, clubs and sponsors shown are concept renderings of the long-term vision, not real people or brands.

The rest of the league

A league is mostly logistics. I designed the back office around each role’s actual day: a competition wizard for seasons and tournaments, a schedule board the Commissioner publishes from, roster checks, and a match day flow with check-in, a go/no-go and a final sign-off at the table. Results flow from the console into public pages for fixtures and standings.

Officials can’t work from a website, so every rule change also rebuilds the printed books together, and they never disagree.

Official Rulebook cover, 2026 edition, version 3.9
Official RulebookVersion 3.9 · 62 pages
Game Officials Handbook cover, 2026 edition
Game Officials HandbookVersion 1.0 · 47 pages
Scorer’s Table Handbook cover, 2026 edition
Scorer’s Table HandbookVersion 1.0 · 63 pages
Quick Play Guide: the court, teams, game structure, scoring and the serve on one page
Quick Play GuideOne page, everything for a first game

What isn’t proven yet

  • No real match has been played. The simulator’s players are a model. How often Prime survives a round is still an assumption until real players show otherwise.
  • The clock is still open. In the model, a clock that runs only during rallies stretches a match to between three and five hours. A running clock brings it to about 80 minutes. The playtest will decide.
  • No real operator yet. The console has never run a live match with a crowd watching.
  • Screens in a real room. Projector readability at viewing distance, and the cloud broadcast path, still need testing in a venue.

The first playtest will measure the rally, the rotation and the scoring with real players. Then I’ll recalibrate the simulator, rerun the studies, and see whether V3.9 holds.

What I learned

Designing a sport is designing a system, and every part leans on another. One scoring rule changed the console, the graphics, the books and the stats at once. The simulator let me argue with my own rules using evidence instead of instinct, while they were still cheap to change. Looking for the cause before the fix mattered most: the serve fix did more than any cap would have. And some design problems were really rules nobody had written yet. Once I answered those as Commissioner, the screens got simpler.

Numbers come from my match simulator. Clubs, players, arenas and sponsors shown are concepts or fictional. Console and display captures are demo matches.

Credits

Sport design, product and rulings
Jaylon Ballard
Built with
Claude Code and Codex, directed and reviewed by me
League
National Gridball League, LLC, with its founding directorate
Note
National Gridball League, Own the Grid and the NGL logo are trademarks of National Gridball League, LLC.