Two charts, one week, no agreement

The hundred, as Steam ranks them by play

Click a column to sort by it. # always stays Steam’s played rank, so wherever you sort, the disagreement stays on screen.

Nothing on the chart matches that.

How far apart are the two charts?

Every row above carries two ranks: where the game sits when Steam orders the hundred by how much they are played, and where it sits when Steam orders them by how many people are inside at this moment. Those are two different questions, so they come back with two different answers. Here is how far apart, counted out of the hundred rows on this page. The right-hand column carries the running total down to a hundred.

Of the 100 rows, 18 are played enough for this chart but not crowded enough for the live one, and that brackets the live chart’s 100th place without our ever having seen it. The smallest crowd still on it is Warhammer 40,000: Space Marine 2 at 19,457; the largest one missing from it is Rocket League at 18,631. The line falls somewhere in those 826 players.

Does one column tell you which way a game will disagree?

Of peak is the share of a game’s own busiest moment this week that was playing at the instant this page was built. Take the games that appear on both charts, sort them by that share, cut them into three equal groups, and compare the two ranks inside each group. A positive average means the group sits lower on the live chart than on the played one; a negative average means it sits higher.

The three groups do not come out in order this week, so the column is not predicting anything at the moment and the six games below are this week’s extremes rather than an illustration of a rule. That is worth knowing on its own: two things push the column up and they do not always pull the same way. A game whose players mostly live in one part of the world looks full while that part is awake; a game people sit inside for hours also looks full, at every hour. Both read as a high share of peak, and nothing else on this page separates them. None of it accounts for the 18 missing from the live chart altogether: across the 17 of them that report a live count, share of peak runs at a median of 98% against 102% for all 99 rows that do, so they are not simply caught in their quiet hours. They are smaller at any one instant than the live chart’s 100th place, and that is the whole of it.

A hundred rows. Where are the people?

A ranked list gives every row the same height, which quietly suggests the rows are comparable to one another. They are not. Add up the Playing now column and ask how much of that total each slice of the chart is holding.

Half of everyone counted here is inside the top 9 games. The middle row has 43,159 playing and the busiest has 29 times that; the 99th has 7,145. One number on this page is worth not trusting: the week-peak column adds up to 8,381,123 across all 100 rows, and that crowd never existed — it is a hundred different busiest moments summed as though they were the same moment. Only the 8,166,118 above was counted all at once. Every figure in this section leaves out 1 row: Steam publishes no live count for it, and a missing count is not a count of zero.

Where every number on this page comes from

1. The two lists. Two official Steam endpoints, one request each, a hundred rows each. One ranks by how much a game is played, the other by how many are in it right now. We do not pick the games and we do not reorder them before printing.
2. Column #. Exactly the rank the played chart returned. Steam does not publish the formula behind it. What we can check is what it is not: it is not the peak column and it is not the live column, both of which sit right next to it and both of which put the list in a different order.
3. Column Live #. The same game’s rank on the live chart. off chart means it is played enough to make the first hundred but is not in the live hundred at all — usually a game people play in long sessions rather than all at once.
4. Week peak. The most players in the game at one time during the chart’s week, from the same response.
5. Playing now. Read once when this page was built — the time is in the top right. It comes from the live chart, and for the games missing from it, from one extra official call each. It is a snapshot, not a ticker, and it does not update while you read. not given means Steam publishes no live count for that game at all, which happens with playtests and demos. Those rows are left out of every total and every share on this page rather than counted as zero, and each section that drops one says so.
6. Of peak. Playing now divided by week peak. It says how full a game is against its own busiest moment. Two things push it up and this page cannot tell them apart: it being daytime in the region most of the game’s players live in, and the game being one people sit inside for hours at a stretch. It is not a quality score, and a small game can sit at the top of it.
7. Δ wk. Last week’s played rank minus this week’s. Games with no rank last week are marked new.
8. Titles. From Steam’s store, one request per game, cached here so we are not hammering it every refresh.
9. The three sections above. Counted from the same hundred rows, nothing fetched for them. Every figure in them is worked out when the page is built and printed into the file, so they change only when the chart does — and so they are still here with scripting switched off. Where a section draws a conclusion, the sentence is generated from the numbers it is about, not typed in once and left to rot.
10. What we cannot show. These endpoints give one instant and one weekly peak — no history. So there are no trend lines on this page. Drawing a curve through two points would look like data and would not be any.