Why it is built this way

Developability is how much value would be created by closing the one limiting factor holding a player's price down.

This page is for the reader who wants to know where the model would break. Every weakness on it is one we would raise ourselves in a meeting.

Three things it refuses to do

Most instruments with a formula in them are built to produce an output. This one is built to withhold one in three specific situations, and that is the clearest thing we can tell you about how it works.

It refuses to produce a score. Four screening questions run first. Any single no and no number is produced, only a reason and a routing answer.

It refuses to project a value without evidence. Value figures come from comparable players who closed a similar weakness at a similar age and level. Where there is no comparable set, no figure is projected and we say so.

It refuses to reward more hours. The ranking denominator is capped. Beyond the cap, additional hours do not count toward projected value.

Why the score is broken out rather than given whole

A single number invites a reader to treat it as a verdict. The breakdown makes it arguable.

Two players can sit at the same index for opposite reasons. One is cheap and straightforward. The other is a project whose main risk is that the work will not show up in a match, which is where it gets paid for. Those are different decisions and a single figure hides the difference.

The breakdown is also what tells a coach which factor is the binding constraint, which is the thing he actually needs on Monday.

The firewall

The index never invents a currency figure.

The Developability Index measures the size and feasibility of a prize. The value uplift comes from comparable players who closed a similar weakness, and from nothing else. The index tells you which projected uplifts are good bets. It does not generate them.

Without that separation, a reader is entitled to assume the money figure was worked backwards from the score, and there would be no way to prove otherwise afterwards.

Every value band we report carries the size of the comparable set it came from. A band with a hidden denominator is a sales figure.

What we do not claim

We do not claim sole credit for a change in value. Age, minutes, form and market movement all contribute. What the record evidences is that a named skill moved in the trained direction, on our timeline, against a prediction made in advance. That supports causation. It does not prove it.

We do not have a track record yet. What exists is a specified instrument, a written delivery model, and a defined unit of evidence. Cases are being built. When we have them, the ones that failed will be published at the same rate as the ones that worked, because a case set with no failures in it has been curated.

The practice design layer has a method and a prediction, not an evidence base. It is specified and it is in pilot. We would rather say that than let you find out later.

The scores are priors. They are set from reasoning and evidence, then corrected against outcomes. Where a case is the first of its kind, confidence is low by design and the reported range is wider.

Where the numbers come from

The factor scores are set against written anchors, so that two assessors looking at the same player have a reason to land in the same place. Reliability between assessors is something we intend to measure and publish rather than assert.

At the level we work at, the instrument runs as an expert assessment against video the club already holds, not as an algorithm over a purchased data feed. That costs scale. It buys a defensible assessment in leagues where the feeds do not reach, and a written justification for every judgement, which a model output does not give you.