DivingDeal · The ranking
How do we rank?
A 5/5 rating from 2 reviews is not information: it is a coincidence. We cannot show it ahead of a 4.8/5 from 200 reviews. Here is how we decide, and with exactly which numbers.
The formula
score = (v / (v + m)) × rating + (m / (v + m)) × prior
v = number of reviews · m = weight of the prior, in equivalent reviews · prior = starting rating
It is a Bayesian average. Every business starts with a cautious rating, and its real reviews correct it all the faster as they are many. Few reviews, we believe you a little. Many reviews, we believe you fully.
Values in force
| Category | Review source | prior | m |
|---|---|---|---|
| Dive centres | 4.20 / 5 | 90 reviews | |
| Hotels | 4.00 / 5 | 210 reviews | |
| Hotels without a Google listing | Booking | 7.80 / 10 | 150 reviews |
| Liveaboards | liveaboard.com | 7.90 / 10 | 22 reviews |
In force since 29 September 2026.
Each prior is measured at the first percentile of its own population, the same percentile for all of them, never a rating converted from one scale to another. A Google score and a Booking score are never compared: centres are ranked among centres, boats among boats, and among hotels, those with a Google rating always come before the rare hotels that only have a Booking rating.
What it means in practice
| Business | Displayed rating | Reviews | Sort score |
|---|---|---|---|
| A | 5.0 / 5 | 2 | 4.217 |
| B | 4.8 / 5 | 200 | 4.614 |
B ranks ahead of A. That is the intended behaviour.
Why a pessimistic prior
Until 8 September 2026, our starting rating was the catalogue average, 4.82 for Google. That was a mistake: a poorly rated business with few reviews was pulled up towards that average, and so rewarded for its lack of reviews. A 4.4/5 on 19 reviews ranked ahead of a 4.7/5 on 232 reviews.
We replaced the average with a pessimistic prior. The correction now always pulls downwards: a lack of reviews can no longer benefit anyone, in any direction. Our internal rule forbids going back to a prior equal to the average.
What we never display
The sort score. It means nothing to you, and showing it would suggest that we rate businesses ourselves. On every page you see the real rating of the platform and its number of reviews, never ours. When the order of a list contradicts the displayed ratings, we explain it right there, with the numbers of the page.
What we cannot measure
The age of reviews. Google gives us a rating and a count, not dates. Two thousand reviews from 2016 therefore weigh as much as two hundred from 2025. We cannot correct that bias today, and we would rather say so than pretend it does not exist.
Fake reviews. We take the ratings of the platforms as they are. Their moderation is theirs, not ours.
This ranking is not for sale
No amount of money moves a business up this list.
History
| Date | Change |
|---|---|
| 29 September 2026 | Hotels: ranked on their Google rating, with a prior measured on hotels (4.00 out of 5, 210 reviews). The Booking rating now only serves hotels without a Google listing. Reason: one scale for almost every hotel. |
| 8 September 2026 | Prior moved from the catalogue average to a pessimistic prior at the first percentile. Reason: the old prior favoured businesses with few reviews. |
The parameters are re-measured every quarter on the real population. Every change is recorded here, with its date and its reason.