What the rating counts: an average and the ratings behind it

A rating on a store page is an average and a count together. On 6 November 2026 this desk read both on 4 listings in Nigeria and the split by star underneath. The table below is the reading; the text explains how to take it apart.

The rating figures and the stars

Average, number of ratings and the count and share for each number of stars. Pages read here: MetaTrader 4 and 5 in Nigeria.

An average with a count under it

The star figure at the top of a listing is a summary. Under it sits the number of ratings it summarises, and under that the split into five bars. The table prints all three. On 6 November 2026 this desk read both on 4 listings. MetaTrader 4 Forex Trading (Google Play): 4.5 from 1,024,688 ratings. MetaTrader 5 — Forex, Stocks (Google Play): 4.6 from 1,234,015 ratings. MetaTrader 4 (App Store): 4.7 from 7,296 ratings. MetaTrader 5 (App Store): 4.7 from 24,386 ratings.

Two figures on one Google Play page

On the Google Play page of MetaTrader 4 Forex Trading the header shows 4.5 stars (4.529 before rounding) and, in the “Ratings and reviews” block lower on the page, 4.5 (4.532) with 987,576 ratings; the header's count is 1,024,688, a difference of 37,112. On the Google Play page of MetaTrader 5 — Forex, Stocks the header shows 4.6 stars (4.619 before rounding) and, in the “Ratings and reviews” block lower on the page, 4.6 (4.621) with 1,199,907 ratings; the header's count is 1,234,015, a difference of 34,108.

Both figures are printed by the page, which does not explain the gap; the table therefore shows both rows for each Play listing without choosing. Rounding is part of the picture: two honest roundings of numbers a few thousandths of a star apart can read differently.

The same count, split by star

The split in the table is worked out from the counts the page gives. MetaTrader 4 Forex Trading (Google Play): 712,380 of 987,530 are five-star (72%), 167,683 four-star (17%), 58,015 three-star (6%), 19,259 two-star (2%) and 30,193 one-star (3%); MetaTrader 5 — Forex, Stocks (Google Play): 913,555 of 1,199,876 are five-star (76%), 189,760 four-star (16%), 51,391 three-star (4%), 18,209 two-star (2%) and 26,961 one-star (2%); MetaTrader 4 (App Store): 5,640 of 7,296 are five-star (77%), 1,135 four-star (16%), 325 three-star (4%), 88 two-star (1%) and 108 one-star (1%); MetaTrader 5 (App Store): 19,549 of 24,386 are five-star (80%), 3,079 four-star (13%), 931 three-star (4%), 342 two-star (1%) and 485 one-star (2%).

Shares are printed to the nearest whole percent, so a column of shares can add to 99 or 101. The counts beside them are exact as the page gave them.

Why the count matters more than it seems

4.7 from 7,296 ratings (MetaTrader 4, the App Store) and 4.6 from 1,234,015 (MetaTrader 5 — Forex, Stocks, Google Play) are written the same way and carry very different weight. With 7,296 ratings a single new rating moves the average by at most 0.00069 of a star (five divided by the count); with 1,234,015 it moves by at most 0.00000405.

This is arithmetic and not a view on the apps. A small count can swing between two readings; a count in the hundreds of thousands is steady by comparison. Hence the date beside every rating: a rating moves, and a figure without its date is half a figure.

A worked example of how firm an average is

Take the average a page shows, multiply by the count, add new ratings and divide again; that is all an average is. Worked through for each page that shows a rating: MetaTrader 4 Forex Trading (Google Play), average 4.529 over 1,024,688 ratings: about 29,883 new one-star ratings in a row would be needed to pull the average down by a tenth of a star; MetaTrader 5 — Forex, Stocks (Google Play), average 4.619 over 1,234,015 ratings: about 35,067 new one-star ratings in a row would be needed to pull the average down by a tenth of a star; MetaTrader 4 (App Store), average 4.660 over 7,296 ratings: about 205 new one-star ratings in a row would be needed to pull the average down by a tenth of a star; MetaTrader 5 (App Store), average 4.676 over 24,386 ratings: about 682 new one-star ratings in a row would be needed to pull the average down by a tenth of a star.

These are calculations on the page's numbers and not forecasts. They show why an average built on a large count moves slowly, and why one one-star rating that vanishes on a listing with many ratings matters on a listing with few.

The words on a rating line

Four terms recur on a rating line. A rating is one star score from one person; a review is a rating with text; the average is the sum of ratings divided by their number; the count is how many ratings the average covers. Play shows the count in short form such as “104K reviews”, which this site reads as the exact figure in the page's data, and the App Store prints “Ratings” in full.

Different stores, different populations

Play and App Store ratings are two separate collections, by two companies, from two groups of phone owners, for apps shown under their own developer names. Placing the averages side by side is fair; subtracting one from the other to find the better app is not, since the pages do not say they rate the same product. Play labels its ratings “Ratings and reviews are verified”; the label is repeated here untested.

What this page leaves out

Reviews are left out: they are other people's words, they change hourly, and a few picked by a store's ordering are no sample. The pictures on this site have the review cards hidden for that reason.

Only the recomputable part remains: an average, a count and five bars.