How we rate Android games

Where the shortlist comes from

A shortlist starts as a much longer list of candidates. Some arrive because a reader wrote in, some because a title has been sitting on our own devices for months, and some because a store category page put them in front of us. Nothing arrives by developer submission — we do not accept them, and a message asking to be considered is filed and not acted on.

From that longer list we remove anything we cannot install and run ourselves, anything whose listing does not match what installs, and anything already covered in a recent edition without a meaningful change since. What survives gets played. Only then does the ranking get argued out.

The list is short on purpose. A page of forty entries is a directory, and a directory cannot hold an opinion. Four entries can be defended line by line, which is the standard we hold ourselves to.

The first criterion in practice

The first hour is where most Android games either earn a place on a home screen or lose it. We note the point at which the game stops instructing and starts trusting the player. A tutorial that highlights a button, waits for the tap, then highlights the next button is pointing rather than teaching, and we say so.

We also note what happens on failure early on. A game that explains why an attempt went wrong is doing more for a new player than one that simply resets. On a touchscreen this matters more than it does with a controller, because the player cannot always tell whether the input or the intention was at fault.

The second criterion in practice

The hundredth hour is shorthand, not a measurement. We do not claim to have played any title for a hundred hours, and any site telling you otherwise about a four-game shortlist is inviting you to check. What we do instead is look for the structural signals that a game will still be interesting long after the novelty has gone: whether new mechanics arrive or only new numbers, whether difficulty rises with the player's understanding or with the clock, and whether late content reads as authored or as recombined.

Arcade titles are judged on this criterion too, and often score lower for it. A run-based game is not obliged to have depth, but if it does not, it cannot claim the top of a shortlist that includes something that does.

The third criterion in practice

Every game asks for something. Storage and battery are the visible costs; attention and money are the ones a listing does not state plainly. We read the price field as it stands on the store — a one-off figure, or a free entry with purchases, or a free entry with purchases and advertising — and we treat each of those as a different proposition rather than as degrees of the same thing.

Where purchases sit inside the core loop rather than beside it, we mark the game down. Where a wait can be removed with a payment, we mark it down further, because that is a design that has decided your time is the lever. None of this is a moral position; it is a statement about what the game returns for what it takes.

How the X.X/10 figure is reached

The score is a single number on a scale that runs from 5.0 to 9.8, published to one decimal place. Read in words: a figure in the low fives means we can describe an audience for the game but would not put it in front of a newcomer. The sixes mean the game works but asks something awkward of the player — setup, tolerance for repetition, or a narrow taste. The sevens are solid recommendations with a stated reservation. The eights are titles we would hand over without a caveat, and the nines are reserved for a game that does something well enough that we would keep it installed after the edition closes.

We do not publish anything below 5.0 because a game we would score under five does not reach the shortlist at all. We do not publish 10.0 because the scale has to leave room above whatever we happen to have seen.

There are no percentage weights behind the number. We considered publishing some and decided against it: any split we wrote down would be a decoration over a judgement, and readers deserve to know which of the two they are looking at. Three criteria are discussed, the desk agrees a figure, and the figure is written next to the reasoning that produced it.

Why our score differs from the store's stars

A star average on a Google Play listing answers a different question from ours. It aggregates ratings left by players, frequently at the moment of greatest pleasure or greatest annoyance, across years and across devices we have never touched. It is a useful signal and we reproduce it on every card for exactly that reason.

Our number is one desk's assessment, weighted toward craft and long-term value. The two figures should not match, and when they do it is coincidence. This is also why a title with a lower star average can sit above one with a higher average on our page — on the current shortlist, an entry showing 4.8 stars against a single review sits last, because one review is not a sample and the setup work it asks for is real.

What this desk does not do

We do not rank by download count alone. A game installed half a billion times has demonstrated distribution, not quality, and the two are routinely confused.

We do not take developer submissions, and no position on any page has ever been sold. There is no arrangement under which a title could pay to move up, down, or onto the list.

We do not score a game we have not run ourselves, and we do not carry over anyone else's verdict. We do not publish a score we cannot explain in a single sentence — if the reasoning will not fit, the entry is not ready.

Corrections and re-checks

The shortlist is standing rather than seasonal. We re-check it on a regular schedule and sooner whenever a price or a permission set changes, because both alter the third criterion directly. Entries move down as readily as up.

Every change is dated. Where a store figure moved after publication, we say that it moved rather than quietly restating the new value. If you believe a figure on our pages is wrong, write to the desk and tell us which listing you read and when.