Valahigh · Portfolio · Clearifi

A recruiting-intelligence layer for men's basketball.

Clearifi pairs a cross-level talent-evaluation score with Webster, a computer-vision film engine. Built transparency-first and abstain-first: every score and report is explainable, versioned, and confidence-tagged.

Overview

What it is

Clearifi is a recruiting-intelligence platform for men's basketball that normalizes fragmented player data - high school, club, college, and transfer - from public and provided sources, and brings game film into one consistent, coach-facing view. It is an intelligence layer, not a directory or a highlight service.

At its core is a transparent talent-evaluation score that aims to place players from different levels on a single comparability scale, with cross-level translation still developing, and a computer-vision engine, Webster, being built to turn raw game film into structured, play-level evidence attached to a player's card.

The design ethos throughout is transparency-first and augmentation-not-replacement: every score and report is explainable, versioned, and confidence-tagged, built to support a coach's judgment rather than override it.

ii.

The walkthrough

See it run

A short walkthrough of the coach-facing view: the cross-level comparability score, and how Webster is being built to turn film into confidence-tagged, human-reviewable play-level evidence. Also live at clearifi.us

iii.

Inside the deck

The through-line

Clearifi · Deep dive

The box score loses the player. We give it back, honestly.

A deeper look at the market, the build, the engine, and where it goes.

The founder

It started as a spreadsheet.

Alex built a spreadsheet to help his younger brother find his real fit with colleges, and the opportunities he could not see. The spreadsheet became a framework; the framework became the platform. Today he is founder and sole owner of Valahigh Co. LLC, the company behind Clearifi and Webster, and built both the platform and the computer-vision pipeline himself.

01 · The market

The system celebrates the ~7% who make it and forgets the 93% who don't.

~540KHS boys play basketball

~3.6%reach the NCAA

~1%reach Division 1

Overlooked is not unqualified: late developers and small-school athletes are screened out by exposure and timing, not by ceiling. College is no safety net either - the transfer portal leaves thousands of rostered athletes without a program every year. That forgotten cohort is the market.

What Clearifi is

One platform for every side of recruiting, built on honest evaluation.

Webster, the computer-vision engine, is being built to turn raw game film into review-gated, play-level evidence. Carry places every player on one cross-level scale. A human editor reviews the read, and every correction trains the engine. It reaches each side of recruiting as a player advisor, a right-fit program finder, and tailored hubs for every role, over a large multi-source player database, with Webster proven on real game film.

The landscape

The field is moving from exposure to evaluation.

The open corner is where evidence from the film meets a full, cross-level platform. Film tools capture and host the game but do not read it. Rankings sell human opinion and star ratings. Exposure platforms sell access, not evidence. Drill-based computer vision reads one move at a time; pro and team tracking reads full games, but only for the NBA and top programs. The white space is honesty: transparent, cross-level tier guidance, plus the access intelligence the market withholds.

02 · The build

A working product, not a prototype.

The platform, at scale Advisor, Find Program, persona hubs, and profiles, live today.

Webster on real film Proven on a full game: scorebug-anchored events and context moments.

The evidence pipeline Review-gated: surfaced as a candidate, never auto-credited.

Value across personas

The job each persona hires us for.

Player Clarity: transparency into where they really stand, plus right-fit visibility.

Parent ROI: a clear return on recruiting spend, on a clear timeline.

Coach Leverage: centralized operations and data to advocate for players.

AAU director Control: efficient operations and transfer-portal transparency.

Evaluation and intelligence value. Never a placement promise.

The value web

One engine. Value on every side, and every side makes it smarter.

Player, parent, coach, college program, and event director each draw value from the same engine, and each one feeds it film, usage, and corrections in return. Every game filmed, every correction, every search makes the next read sharper and more valuable. That compounding is the moat.

03 · The engine

Webster: from pixels to a review-gated skill read.

1 · Detect & track Working: players and ball, stable on a full game.

2 · Identity The gate: owned-label, full-resolution reads, the bottleneck.

3 · Scorebug event Working: anchors every scoring event.

4 · Make / miss Abstains on misses for now.

5 · Attribution Abstains, never guesses; clears once identity volume lands.

6 · Context to skill Working: why it mattered, into a skill profile.

Proven on one game. We abstain rather than guess at every step that is not there yet.

Carry

The read the market never had.

A transparent, cross-level comparability measure built on production and paired with TES: a benchmark, not a subjective ranking, built on what a player does rather than reputation. It is conservative and auditable, and it never promises a player will move up - it describes today, not destiny. One scale across levels means the small-school scorer the rankings skipped and the showcase prospect they ranked sit on the same production scale. The overlooked player finally has a number, review-gated, never a promise.

04 · The vision

A transparent benchmark for a market that runs on hype.

The wedge: prove it down-market on small schools and overlooked players, let the benchmark hold up against real production, then bring the transparent value hook to the top. Transparent comparables already anchor negotiation in pro sports; as Carry is validated against realized value, it can become that reference point for college. Benchmark-anchored allocation across every sport is the long-horizon direction - roadmap, not shipped.

Why now

A monetized, high-churn market that still can't get the truth.

Newly monetized NIL made a player's value explicit, and schools now share revenue with athletes directly.

Constant movement 30,000+ athletes enter the transfer portal every year, 2,000+ in D1 men's basketball alone, and roughly 1 in 3 never land at a new school.

Demand already paid Families already spend heavily on youth sports and recruiting tools; the money buys hype, not the truth.

Why basketball first

Prove it on the hardest sport. The same engine extends outward.

Worst funnel, most churn, most overlooked, and the hardest computer-vision test - contained 5-on-5 broadcast film. If it works here, it works. The same engine extends to football, volleyball, soccer, and beyond: same framework, more sports, the dial turns but the method does not. Multi-sport is roadmap, not shipped.

01 / 14

iv.

What I built

Two things, working together

The score

A transparent talent-evaluation score that puts players from different levels - a high-school recruit, a club standout, a college transfer - onto one common comparability scale so they can be viewed side by side, with cross-level translation treated as developing and unvalidated.

It is explicitly a comparability measure, not a prediction of future outcomes. That was a deliberate design call: I scoped it as a cross-level comparability measure rather than an outcome forecast, and shipped an auditable, versioned score instead of a black-box "will succeed at the next level" number.

Webster, the film engine

Webster is a computer-vision engine designed to turn game film into structured, play-level evidence - proposing who did what and when where the footage supports it, and abstaining otherwise - plus a per-skill profile on each player card. It orchestrates well-understood open-source building blocks - object detection, multi-object tracking, scoreboard reading, jersey-number reading, pose, and appearance embeddings - into a single pipeline rather than leaning on one proprietary model.

It is abstain-first and human-in-the-loop by design: it proposes rather than asserts, reports an explicit confidence with every output, and abstains when uncertain instead of guessing. Results route through a human correction step; those corrections are captured to inform future model training once rights-cleared, labeled data is available.

The output format is designed so that every analysis shares the same structured, versioned shape, making two reports directly comparable, and it is built to reduce the analyst-to-analyst subjectivity, fatigue, and drift of ad hoc scouting. Identity is approached in a privacy-respecting, BIPA-conscious way, using jersey number, team color, on-court position, and roster context as evidence rather than asserted identity, and abstaining when the footage is not clear enough - no facial recognition, no biometric identifiers anywhere.

Under the hood it is a modern web stack (React / TypeScript front end, a Postgres-based backend) plus a GPU-accelerated cloud vision pipeline, with a clean architectural boundary that separates the scoring engine from any single data source, so the same engine can run on different underlying feeds. The rights posture is deliberate: any model training will use only openly licensed footage (Creative Commons / public domain), shipping models are never trained on copyrighted broadcast footage, and only permissively licensed open-source components (MIT / BSD / Apache) ship in the product.

Data and scoring surfaces are built and operational; the computer-vision engine is in active development. No CV models have been trained or shipped yet, and results - including modest ones - are reported honestly. Accuracy on real film is work in progress.

See it live at clearifi.us Back to Valahigh