How to read a leaderboard

A ranked table invites a quick conclusion. This page is about the places where the quick conclusion is wrong.

The first half holds for every board. The second half takes each live board in turn — Transfers, Team Value and Player Value — and what its own columns mean. The terms used here are defined in the glossary.

Zero is the interesting number

Not one, and not the top of the table.

Every skill score is measured against a baseline: the answer anyone could give without a model of their own. Zero means a participant did no better than that — which is a real thing to know, freely available, and requires no model at all. Above zero is the whole claim.

• On Transfers, zero is the league base rate for the player's role. Above it, a participant is extracting something about individual players that the league average does not contain.
• On Team Value, zero is the scoreline. Above it, a participant is seeing something in matches that goal difference does not.
• On Player Value, zero is naive plus-minus: each player valued by the goal difference he has accumulated while standing on the pitch. Above it, a rating is seeing something in a player that his own on-pitch goal difference does not.

That also makes the scale unintuitive. Skill scores are not percentages and not accuracy, and the boards' scales are not comparable with each other: +0.21 is a strong Transfers result, while a season of Team Value resolves differences of about five hundredths and Player Value about fifteen thousandths.

Negative is shown as negative. It means worse than the baseline, which is a real outcome for a confident model pointed at the wrong questions, and clamping it to zero would hide the difference between a bad model and a cautious one.

The five misreadings

1. Reading the ranking before the interval

Early in a season the interval is the story and the point estimate is barely a story at all. A handful of questions with little football behind them cannot separate a good model from a lucky one, and the interval says so.

If two intervals overlap heavily, the board is telling you it cannot yet order those two rows — even though it has to print them in some order.

Both Game Value boards say this in words. Every row reads ahead of, level with or behind its baseline, depending on whether its interval clears zero, and two rows that both read level have not been separated from that baseline, whatever their rank numbers say. Read the words before the number.

2. Reading overlap as "no difference"

The reverse trap, and the subtler one. Two overlapping intervals do not mean two participants are indistinguishable.

Participants answer identical questions, so their errors are correlated, and the difference between two scores is far better determined than either score. On a real twelve-case Transfers cohort, two participants with heavily overlapping intervals — 0.498 [0.387, 0.594] and 0.454 [0.334, 0.559] — had a paired difference of 0.044 [0.035, 0.053]: a clear separation, twelve times tighter than either interval.

Both Game Value skill scores are built this way already: each row's interval is the paired difference between the participant and its baseline, resampled on the same fixtures. What no board publishes is the paired difference between two participants.

So: overlapping intervals mean this board cannot tell you from the columns alone. It does not mean the two are equal.

3. Reading a high score without its coverage

A participant answering 40 of 110 questions very well is not a better model than one answering all 110 nearly as well — it is a model that was allowed to choose.

Declining is legitimate and is not scored as error. It costs coverage instead, which is exactly why coverage sits next to the score on every board. Read the two together, always. A high score at 35% coverage describes a narrow model, and that may be precisely what you want — but it is a different claim from a high score at 95%.

What reduces coverage differs by board: on Transfers, refused and unanswered cases; on Team Value, competitions not entered, matches not priced and values delivered late; on Player Value, appearances left unrated — and the board splits that figure, because part of the gap is the Arena's own and is marked as such.

4. Reading an unranked row as ranked

Some rows are shown but never ranked. Their rank reads —, and they sort below every ranked row whatever sortable column you choose.

• On Transfers they are provisional: the evidence behind the row is too thin to place it, usually because the participant entered recently and its cases have accumulated few appearances. Early in a season this is true of everyone.
• On Team Value they read no score: there was nothing to publish that week, such as a signal that ordered nothing. A missing score is not a score of zero, which would claim the participant had matched the scoreline.
• On Player Value they read not scored yet, and early in a season that is the normal state rather than the exceptional one. A segment needs prior ratings for at least nine of the eleven on the pitch, so the opening rounds of a season are scorable for nobody: no player has a prior appearance to be rated on yet.

None of these is a judgement about the participant. All of them say the row cannot be placed yet.

5. Reading two rankings as one

This one belongs to Player Value, which publishes two rankings, one under the other, and never blends them.

The first asks whether the goals followed the players a rating liked. The second asks whether the coach did — whether a rating anticipates who is about to play more. They are different claims, and a product can be clear of the baseline on one and level on the other without any contradiction.

There is no combined figure, and its absence is not an oversight. Weighting agreement with goals against agreement with selection would be the Arena inventing the most important number on the page, and it has no defensible weighting to offer. A reader who wants one ranking should read the first and treat the second as a separate finding — not average them by eye.

Two further things follow:

• The second ranking resolves several times coarser than the first, on orders of magnitude fewer pairs. Level there carries even less of a verdict than level on the score.
• The second is agreement, not truth. A rating that disagrees with coaches may be wrong, or may be early. The column is named for what it measures and claims nothing beyond it.

Entry point matters

Entry is open all season. A participant that entered in November answered with more of the season already observable than one that entered in August, and is scored only on what happened after it entered.

They did not answer the same question, and every board is built to show that rather than hide it:

• Coverage carries it. A late entrant has fewer questions it can still be scored on, so its coverage is lower by construction.
• The season column carries it. Its trajectory starts partway through, with a visible gap where it was not yet scored.
• On Transfers, Info at lock says how much football its players had already played when its predictions locked.

A late entrant with high coverage would be the surprising thing, not a late entrant with low coverage — unless the row says scored under an exception, which is the one case where the Arena has waived the rule and which every such row carries beside its name.

The Transfers board

On the home page. Each row is a participant's predictions for the same transfer cases, scored against what the players have done since.

A row is provisional below the floor of appearances per scored case printed under the board.

Confidence is a separate ranking

It answers a different question: does this participant know when it does not know?

It is never folded into the skill score. A well-calibrated but inaccurate model and an accurate but overconfident one are different products, and collapsing them into one number would serve neither.

"No data" is not zero. A participant that sends no confidence has made no claim about its own certainty, which is different from having made a badly calibrated one. It sorts last in both directions, because floating an absent claim to the top of an ascending sort would read as the worst-calibrated participant on the board.

A constant confidence also scores no data. Asserting certainty on everything carries no ordering information and gains nothing.

The cohort strip counts the benchmark, not the participants

Those four numbers are identical for everyone, which is why they are published once above the table rather than repeated on every row.

• Cases in registry — every eligible transfer into the covered leagues.
• Void — the player was transferred again inside the window, so he can no longer be observed at the club he was predicted for. Voided for everyone.
• Unobserved — signed, resolved, and has not played. A football outcome, not a pipeline failure.
• Scorable — has at least one appearance. This is the coverage denominator.

The gap between the first and the last is real attrition, and it is shown rather than quietly absorbed. A board reporting only the scorable count would be hiding that far more players were signed and that many of them have not yet kicked a ball.

Filtering by league

The filter selects destination competition — the league a player moved to, which is the league his percentiles are computed against.

Two things change when you filter:

• The skill score is recomputed for that scope, weighted by cases. It is not an average of per-league scores: a ratio recombined by averaging would weight a 60-case cohort like a 120-case one.
• The interval appears. Intervals come from resampling cases within one cohort and cannot be recombined across leagues without the case-level scores, so the board shows one only when a single league is in scope, rather than inventing a range.

The Team Value board

At the team board. Each row is a participant's values for both sides of every fixture in the competitions it entered, each value counting only toward fixtures kicking off after its deadline, scored against the scoreline.

How a row got there

Not every row belongs to a company that entered this benchmark, and the board says which is which.

As of today there is no vendor-submitted row on the Team Value board: three are the Arena's own models and one is a published figure the Arena collected. That is a transitional state and it is stated here rather than left to be inferred from the labels.

The two descriptive columns

Right winner and Draw agreement are not sortable, and the ranking ignores them. They exist because a skill score of +0.0135 is hard to repeat back to anyone, and "called one more match correctly" is not.

They can disagree with the ranking, and have: over 2025/26 the shots entrant had the best right-winner rate on the board while its skill sat below four xG rows. Right winner reads only the sign of a signal, where Somers' D ranks every pair of fixtures by margin. Neither is wrong; they answer different questions.

Draw agreement is the one to read carefully. It is a ratio of a participant's own numbers — the mean gap between its two values on drawn matches, over the same mean on decisive ones — so a model in xG and a model in shot counts are comparable, and scaling every value changes nothing.

• 1.00 — drawn matches looked exactly like the rest. It saw nothing coming.
• Below 1.00 — it told drawn matches from decisive ones.
• 0.00 on the scoreline row — by construction, since a drawn match has equal goals. It is a floor, not a target, and no participant should be read against it.

The scoreline has its own row on those pages, with its own pre-match call, and every participant is read against it: a call goal difference also made was free, so it counts for nothing, and what a team page totals is the disagreements. Zero there means a row has told you nothing the league table had not.

Both columns can now be audited match by match. Each one is a count over fixtures, and every fixture it counts is published: browse a team and open any of its matches to see each participant's standing signal before kick-off, which side it favoured, and what happened — for a drawn match, the gap between its two values against the gap it usually leaves. Matches inside the burn-in are shown and say they count towards nothing, because a fixture a column did not count should not look like one it did.

Every column header carries a ? that explains it in place.

Read the standing, not the rank

Each skill score carries one of four words, decided by its interval rather than its point estimate:

• ahead of the scoreline — the whole interval is above zero;
• level with the scoreline — it contains zero;
• behind the scoreline — the whole interval is below zero;
• no score — nothing to publish that week.

A season of five leagues resolves roughly five hundredths of skill for a participant whose signal is its own, so most first-season rows will read level. That is the honest result, not a failure. The count printed under the board — how many rows are separated from the scoreline — is the number to read before the order.

Why there is no league filter

One skill score across all five competitions, never five. The spread between competitions reaches 0.109 and is not stable between seasons, so a board per league would invite a reading none of its numbers supports. Where a Transfers reader filters by league, a Team Value reader reads one number.

What the strip above the table counts

Four figures, identical for every participant and published once: the fixtures scorable, which is the coverage denominator and reads — until the first scoring run; the five leagues the one score covers; the three horizons averaged into it; and the weeks published.

Horizon decay

Each fixture is forecast from every match known before it, then with each side's most recent match withheld, then the two most recent. The column draws the participant's Somers' D at those three horizons over the scoreline's.

The drop is a diagnostic. A metric of current form falls away as recent matches are withheld; one of durable strength barely moves. Both are legitimate, and the column shows which a participant has. Read the drop against the scoreline's own: in a week where results were hard to order, both fall together and nothing has been learned about the model.

Deliveries

Lateness is published because it is not private. A value delivered after its deadline is kept in the record and counts toward no fixture, so a row with late deliveries has fewer fixtures behind its score than it could have had — which is a fact about the score a reader is being asked to trust.

The Player Value board

At the player board. Each row is a participant's ratings for every appearance in the competitions it certified on, each rating counting only toward matches kicking off after it was due, scored against the goals scored while those players were on the pitch.

One board, two views. A switch above the table chooses which ranking is shown — the score or the selection column — and each is a link, so a view can be sent to someone. Everything above about not blending them applies; the columns are listed separately because they answer separate questions.

The score: did the goals follow your players?

The selection column: did the coach pick who you rated?

Before the column has read for anyone — it needs every club to have played about ten matches — the view says so in place of a table, and no count is printed: a zero would be a measurement.

Every column header carries a ? that explains it in place, as on the team board.

Right winner, and why it sits so close to plus-minus

It is there so a result can be repeated back — called 29 more segments than plus-minus — where a skill score of +0.015 cannot. Expect it to sit within a point or so of plus-minus: over the rehearsed seasons both products and the baseline called the winner of 58 to 63% of decisive segments, in step with the score. It reads only which side a signal favoured, where Somers' D ranks every pair by margin, so the two can disagree, and a few segments either way is noise.

The team board also publishes Draw agreement. This board does not, on evidence: on segments it reads 0.92 to 0.98 for every product and for plus-minus alike, because a drawn segment is nearly always a goalless stretch of twenty to forty-five minutes, and no rating of who is better foresees that nothing will happen.

The floor is the column to read first

It has no counterpart on any other board, and it exists because of one specific failure mode: a rating that encodes little more than which team is better will still order goals reasonably well, because better teams score more.

The floor is what that rating is worth. It is a signal built from the league table alone — each side's mean points per match before the match, home minus away — scored on the identical segments. No lineups, no ratings, no knowledge of who played.

In both complete rehearsal seasons it landed about 0.004 below plus-minus — on Somers' D, +0.2142 against +0.2185 in 2024/25 and +0.1834 against +0.1879 in 2025/26. Knowing who is on the pitch at all was worth roughly half a hundredth over knowing only which two teams were playing.

That is the honest shape of this board: the bar is nearly saturated before a participant says anything. Beating plus-minus is the claim, clearing the floor is the precondition, and a row sitting between the two has told you about clubs rather than about players. Nothing is ranked by the floor; it is published so this can be checked rather than asserted.

Why two rows show two different plus-minus figures

On the team board the baseline is one number, the same for everyone. Here it is not, and the first time a reader notices it looks like a bug.

The baseline is recomputed over each participant's own scored segments. A participant scored on four leagues and one scored on five are not being measured on the same football, so a single shared baseline would fold the difference in coverage into the difference in skill — and the gap between two rows would stop being a statement about ratings.

Read each row's skill score against the plus-minus figure on its own row. Comparing one participant's Somers' D against another's baseline compares nothing.

Scored under an exception

A row carrying this label was scored on segments that were played before the participant certified. Hovering the label gives the reason and the period it covers.

Ordinarily a rating counts only toward matches kicking off after it was due, so a participant that certifies in September is scored from September and nothing earlier. An exception waives that for a named participant over a named period, and it is written onto the row and into the published snapshot rather than applied quietly.

Every row on the board carries one today, and both say the same thing: the ratings are figures the source had already published against each appearance as the season went along, unchanged and fixed before any result they are now scored against, which is why the Arena judged that nothing had leaked. They were entered when the board opened, and the exception lets the season before that date count.

The label stays on those rows permanently. A score computed under a waived rule is a different claim from one computed under it, and a reader cannot tell the two apart unless the board says which it is.

How a row got there

The same three provenances as the team board, shown beside the name:

As of today both rows on the Player Value board are entered by the Arena: neither is vendor-submitted. That is a transitional state and it is stated here rather than left to be inferred.

Not scored yet is the normal early state

A segment is priced on the mean prior rating of the eleven on the pitch, and needs prior ratings for at least nine of them on both sides. In the opening rounds of a season nobody has a prior appearance, so nothing is scorable for anybody — not for a participant, and not for the baseline.

A first reading of the selection column comes later still: it needs five club matches before a pivot and five after, so it appears only once every club has played about ten.

The pages behind the board

Under Players, beside the leaderboard: every team, its scored matches, and each match cut into its three segments with both elevens. It is where a row's one figure is opened back up into the football it was computed from.

Three things to hold on to when reading one:

• The side share leans; it does not fill. Ratings are bunched — nobody is rated 2 — so a clearly stronger eleven still reads about 51 to 49. The bar is centred on 50 with its edges at 45 and 55, and says so. A share of 52 is a large lean, not a small one.
• One segment proves nothing, and the page draws no verdict from one. Many end goalless. Nothing is marked as called right or wrong; the board pools thousands of segments, each compared only with others that began in the same score state, before it says anything.
• What is published is relative and coarse, on purpose. A player is shown against his side to a whole percent and a side to one decimal. No rating is published, for any participant: the rows entered by the Arena are a vendor's own published ratings, and the pages are not a way to rebuild them match by match.

Past matches only. A lineup exists once a match has been played, and a page once the weekly run has scored it. A rating before the match is the average of what a player was given in earlier matches whose ratings were already due — nothing about the match on the page is in it.

What produced a number

Every board names what its numbers were computed under, because a score is only meaningful alongside the rules that produced it.

• Under the Transfers board, the pins: contract, case registry, scoring metric and reference pools, each linked to the version it names where that version has a page.
• Under the Team Value board, the contract version — and on each participant's page, the five pair counts per horizon for it and for the scoreline on the same fixtures, which are enough to recompute every figure on the row.
• Under the Player Value board, the contract version, the bootstrap seed and the number of replicates — and on each participant's page, the five pair counts for the participant, for plus-minus and for the floor, all on the identical segments. Somers' D = (C − D) ÷ (n₀ − n_Y) and the difference of two of them is the skill score, so any published figure can be recomputed without the Arena's code. Where a row was scored under an exception, the snapshot says so.

A published snapshot is never rewritten, so a number from week 3 still reads the same in week 30 — and still names the rules that were in force when it was produced, not the ones in force when you read it.