How to

How to read this ledger

The five stamps, the coverage column, and why an empty cell is an answer.

The five stamps

Every figure on this site carries one of five marks. They are not a rating of the bookmaker. They are a statement about the sentence the number is sitting in: who said it, and how much work went into checking that they were in a position to say it.

OP Recorded. Read off the operator's own Ghanaian pages: a cashier screen, a terms clause, a licence footer or its own API.

CP Computed. Derived from a price feed the research pass pulled itself, with the sample size and the date on the record.

RP Reported. Published by somebody other than the operator, or agreed across several pages with no primary document behind any of them.

ES Estimated. Filled in by the research pass rather than found. It is not evidence, and this ledger prints it as a figure under protest rather than as a fact.

-- No entry. Nothing is on file. In this market that is usually a checked answer rather than a gap, and the page says which.

The one that changes how you should read a page is the fourth. An estimated figure was put into the fact base by the research pass rather than found anywhere, in one case explicitly set to the middle of the market so a record would not read as blank. It is still used, because removing it would drop a real bookmaker out of the table altogether, and it is still marked, because a figure nobody sourced doing work in a ranking is the thing a reader most needs to be told.

Today that applies to Bangbet and to nothing else in the index.

The coverage column

A score computed over four criteria and a score computed over five are not the same number, and printing both to the same precision without saying which is which is how a table invents confidence it has not earned. Every row in the index therefore carries the share of the model that record could answer.

A criterion with nothing behind it is not scored at zero. It drops out of the average on both sides, so the score is an average of what could be read. That is the only honest way to score a thin record and it has a cost: a bookmaker that answers three questions well and five not at all would float upwards. The floor is what stops that. Below 60 per cent of the model answered, a record is not ranked at all.

Betway answers 90% of the model, missing deposit floor.
Premier Bet answers 90% of the model, missing deposit floor.

What the columns are

Cash out: The median on the record, in minutes, assembled from what the market publishes about the operator. Never a timing of ours.
Floor: The smallest amount that opens an account: the lower of the cashier minimum and any qualifying deposit a welcome offer demands.
Rails: How many cashier rows the operator publishes. It is a measure of how much the operator says, not of how good the payments are.
Model read: Coverage. The share of the 100 points that this record could answer.
Index: The weighted result on a 0 to 100 scale. Getting paid is 46 of it.

Questions about reading it

What does an empty cell mean?

That nothing is on file, and in this market that is usually a checked answer rather than a gap. The alternative, filling it with a market average, would make a bookmaker that publishes nothing indistinguishable from one that publishes exactly the average, with no way to tell afterwards which was which.

Why is a score computed over less of the model still printed?

Because refusing to print it would be a quieter kind of dishonesty: the reader would see a shorter table and assume the missing books were rejected. The coverage column says how much of the model each record answered, and anything under 60 per cent is not scored at all.

The model itself, criterion by criterion, is on the index page, and the entries behind every figure are in the ledger.