CarbonylR&D

Platform · Scheduling, analytics & video

League operations platform

A scheduling engine, a complete club website and a sports analytics stack, from one service for recreational sports clubs. Seasons that are fair by construction, game nights that run themselves, and numbers a club has never had about its own teams and games.

DisciplineOptimization & analytics
IssuePublic extract
AccessBy invitation

The problem

Clubs run on spreadsheets and plugins, and players quietly leave.

Schedules by hand

Some teams always get the late game or meet the same opponent twice, and every rain-out means rebuilding by hand.

Lopsided games

Mismatched teams produce blowouts, and blowouts are why players stop coming back.

Tools that take a cut

Generic sign-up tools and website plugins take a share of every payment and still leave the club stitching systems together.

No view of the club

Nobody can see which teams are leaving, why, or how this season's registrations compare with the last.

Approach

Six rules the service is built on.

R-1

Hard rules are never broken

The engine returns a complete schedule, or exactly what could not be placed and why. It never quietly bends a constraint.

R-2

Change with the least disruption

Rain-outs, late joiners and withdrawals are re-solved so that only what has to move moves. Games already played are never touched.

R-3

Fairness is measured

Team ratings learned from results suggest levels and, when the club asks, pair the rest of a season to avoid mismatches.

R-4

The club keeps its money

Payments go through the club's own payment account. The service takes no share, and every amount is kept to the cent in a ledger with its refunds.

R-5

One service, many clubs

Each club has its own site, domain, look and data, fully separated from every other club.

R-6

Private by default

Public pages load no third-party scripts or trackers, and players can download or delete their own data.

Scheduling engine

A complete season, or the exact reason it cannot be.

clubconstraints build improve complete schedule what could not be placed,and why rain-out, late team, withdrawal:re-solve, moving only what must
FIG. 1The club's fields, nights, teams, levels and can't-play times go in. The engine builds a season and then improves it. Out comes a complete schedule, or a precise account of what could not fit. Changes during the season loop back and move as little as possible.
E-1

Season structures

Round robins with balanced home and away, pools, and playoff brackets that fill themselves from the standings and the league's own tie-breakers.

E-2

Shared resources

Fields and nights shared across leagues, can't-play times, players on more than one team, and levels, all solved together rather than league by league.

E-3

What would fit

When a season cannot be completed, the engine works out what would make it fit, such as doubleheaders, later fields or extra weeks, without saving anything.

E-4

Referee assignment

Officials assigned fairly across all of a club's leagues from their availability, with their own schedules and what they are owed.

Analytics & machine learning

Numbers a club has never had, about its teams, its season and its games.

The analytics run on data the club already produces: results, registrations and, increasingly, video of its games. Every model is measured on data it has not seen before it is allowed to inform a decision.

gamevideo detect &track players pixels tofield meters movement:speed, sprints play: possession,passing, control checked againstground truth,then shown
FIG. 2Ordinary game video becomes tracked players, is calibrated from pixels into field meters, and yields movement and play metrics. Each metric is checked against measured ground truth before it can appear on a game page.
FIG. 3A game page from match analytics: the summary, both teams' passing networks, and pitch control at one moment, with each player's direction of run. Drawn from the platform with every word and number removed; positions are illustrative.
A-1

Team strength

Ratings learned from game results, weighted by margin of victory and carried between seasons, tested out of sample against a no-skill baseline before anyone relies on them.

A-2

Fair pairing

Suggested levels for returning teams and, when the owner asks, the rest of a season paired by strength so that predictable blowouts are avoided before they are scheduled.

A-3

Retention & forecasting

Who comes back and who leaves, what the leavers were worth to the club, and each open registration forecast against last season's pace.

A-4

Player tracking

Players detected and tracked through ordinary footage, assigned to their team by appearance, and placed on the field in meters through camera-to-field calibration.

A-5

Movement physics

Distance, speed zones, sprints, accelerations and heat maps for every tracked player, from the calibrated tracks.

A-6

Tactical analytics

Possession, passes and passing networks, team shape, and pitch control: which team controls which part of the field at each moment.

A-7

Field geometry

A layout planner that measures every cone and marking for small-sided fields, in walking order, kept per field and printable.

A-8

Open research

Keeping each player's identity stable through crowding and occlusion, so that individual numbers become as reliable as team ones.

FIG. 4The owner's numbers: teams by year, who came back and what the leavers were worth, and which games were foreseeably one-sided. Drawn from the platform with every word and figure removed.

Running a club

Everything a club runs, in one place.

C-1

Seasons & standings

Round robins with balanced home and away, pools and playoff brackets, shared fields and nights, tie-breakers and brackets that fill themselves.

C-2

Game night

Self check-in with waivers, staff and referee screens, scores from phones, forfeits counted automatically, weather watch and rain-out handling.

C-3

Registration & payments

Forms with waitlists that move up on their own, deposits and balances, teammates paying their share, memberships, discounts and late fees.

C-4

Pickup, classes & events

Drop-in games with passes, classes, camps, clinics and events, each with rosters, waitlists, payments and attendance.

C-5

The club's own website

Pages, look and domain carried over from the club's current site, player pages in more than one language, calendars that stay current and alerts where players already are.

C-6

The owner's view

Money in and owed, registration pace against last season, who came back and who left, and team ratings, with staff permissions and a full audit log.

C-7

Programmatic access

A signed API for every scheduling operation, and exports to spreadsheets and calendars.

C-8

Operator console

Every club on the service at a glance: new clubs set up from scratch or from an existing site, billing, data exports and delivery checks.

Engineering

Built to the same standard as everything else we ship.

Q-1

Decisions on record

Every design decision is written down as it is made, and every piece of deliberately deferred work is kept in a register rather than lost.

Q-2

Tests that can fail

A large automated suite with property-based tests, and rules tested by mutation: the code is broken on purpose and a test must catch it.

Q-3

Checked in a real browser

Every page is checked in a real browser for accessibility and phone layout, in any club's colors.

Q-4

Security built in

Two-step sign-in, forgery and origin checks on every form, a separate signing key for every club, and personal data that players can download or have removed.

Connections

The same ideas, elsewhere in our work.

X-1

Robotics & learning lab

The video layer turns footage of play into positions, speeds and the physics of the ball. The lab runs the other direction, from physics to learned movement. Two directions across the same bridge.

X-2

Market simulation research

Ratings that update from results are an estimate of hidden strength read through noise, the same kind of problem as reading the state of a market from its order flow.

X-3

Geospatial intelligence platform

Placing players on a pitch from video and placing a camera's view into a city are one problem: pixels mapped into a world frame.

X-4

Ultrafast electron dynamics

Building a season that satisfies every constraint and finding a pulse that drives a molecule to a target are both searches through enormous spaces, made tractable by their structure.

Engage

By invitation.

Every engagement begins with a conversation and is scoped before work begins.

Clubs

Run your club on it, with your site and your data brought over.

Subject: Access

Custom scheduling

The engine applied to your own allocation problem: venues, staff, rooms or fleets.

Subject: Build

Partnerships

Associations, facilities and federations that serve many clubs.

Subject: Partnership

Investment

Briefings and a demonstration, under confidentiality.

Subject: Funding
contact@carbonyl.org