Submission contract
This contract is in draft. Field names and the endpoint are not frozen, and no submission is being scored yet. What is settled is the shape of the ask, which is deliberately small.
What a participant sends
One value per player per month. That is the whole submission.
No model, no methodology disclosure, no endpoint for the Arena to call, nothing to keep running. A participant chooses which players it covers and sends a value for each, once a month.
| Field | Meaning |
|---|---|
player_id | Resolved through the Arena's published entity registry |
value_eur | The participant's valuation, in euros |
as_of | The date the value is asserted for |
Why monthly, and why it cannot be sent later
The fee board scores a value at four lead times before a transfer — 12, 6, 3 and 1 months. Nobody knows in advance which players will move, so the Arena selects the right snapshot afterwards, from the history a participant has already sent.
That has one consequence worth stating plainly:
A value that did not exist before a transfer cannot score that transfer. History cannot be backfilled, and a month not submitted is a month that can never be scored.
A participant that already publishes its values openly can be collected by the Arena on the same monthly grid instead of submitting. Either way the observation grid is identical, so no participant is scored on a denser history than another.
How coverage is counted
The Arena scores wherever it holds a transfer fee, in any competition. A participant is scored only on the players and transfers it covers, and its coverage is published as its own column:
coverage = what the participant priced ÷ what the Arena holds
A participant covering 40 competitions of 80 is scored on those 40 and marked at 50%.
Covering less is not a way to score higher
Fees in a wealthy league may be far more predictable than in a thin one, so a participant covering only the richest competitions could otherwise post a better number by choosing an easier population.
Two figures are published to close that:
| Number | Set | Answers |
|---|---|---|
| Own coverage | the participant's own scoreable players and transfers | How good is this where it operates? |
| Common core | what every ranked participant covers | The only number two participants can be ranked on |
Ranking is on the common core. A participant's own-coverage score is published beside it and is not comparable between participants.
Listed participants
Some valuations are published openly. The Arena scores those without waiting to be asked, under conditions published in advance:
- The method, cohort, baseline and statistic are documented before any score exists, and do not change to suit a result.
- The same rule applies to every participant, including the Arena's own baseline, which sits on the same board.
- A row states how it got there — a participant that submitted and one the Arena collected from public data are marked differently. Being measured without asking is defensible; implying someone entered a competition they never entered is not.
- A participant may dispute the data, ask about the method, or publish a response beside its row.
Refusals and gaps
A player a participant does not price is not an error and is not scored as a wrong answer. It reduces coverage, which is published, and nothing else. A participant that prices only the players it is confident about will show a high score against a low coverage, and a reader can see both.