Season, Expansion, or Update — Why We Classify Records, Not Games

Gather ARPG schedules in one place and you hit a problem immediately: not every game ships content the same way. Path of Exile opens a new league roughly every 100 days and resets characters; Grim Dawn releases an expansion every few years and resets nothing. Put both in the same table under the heading 'season' and the table fills up while the content becomes wrong.

So ARPG Seasons attaches a type to **each individual record**, not to the game it belongs to. As of 19 July 2026 we hold 237 rows across 13 tracked games, of which 234 are confirmed against a primary source: 223 seasons, 6 major updates, and 5 expansions. Just 11 records in total are not seasons. This article sets out the classification rules, why the judgement has to be made per record rather than per game, and which holes remain in our data.

One Question Does the Work — Does Anything Reset?

The dividing question is simple: **when this ends, does something reset?** If characters or a competitive track are periodically wiped, it is a season. If content is added once and stays, it is an expansion or an update. What separates those two is scale and commercial form: a paid or large-scale release that extends the game itself is an expansion, while a named free major patch is an update.

The part that matters is that we put this question to **the record, never to the game**, because games that both run seasons and ship expansions genuinely exist.

Where the 11 Non-Season Records Sit

Only 11 of the 234 records are not seasons — and that scarcity is exactly what makes the classification delicate. Eleven records are too few to notice in any total, yet different enough in character that letting a single one slip into the season sample would move the cycle averages on its own. So here is every one of the 11 and the game it belongs to.

What That Table Shows — Type Boundaries Cut Across Game Boundaries

Two quite different situations sit side by side in that table.

**First, games with no seasons at all.** Grim Dawn and No Rest for the Wicked. Grim Dawn's entire presence in our data is three expansions; No Rest for the Wicked's is four updates. Neither holds a single record classified as a season, so both are permanently absent from season-based aggregates — cycles, weekdays, volatility. That is why you should not expect a 'season cycle' figure on either game's page. Nothing is missing; the concept simply does not apply, and we do not manufacture values that are not there.

**Second, games where seasons and non-seasons coexist.** Diablo Immortal, Dwarven Realms, and Hero Siege. Diablo Immortal alone accounts for 51 confirmed seasons, more than any other game we track — and separately carries 2 expansions. Dwarven Realms and Hero Siege likewise each pair a season history with a single update record.

That second group is precisely why classification cannot be done at game level. Label Diablo Immortal wholesale as 'a seasonal game' and its 2 expansions leak into the season sample; file it instead under 'games that ship expansions' and 51 seasons of history vanish. Neither description is true, so the judgement is only ever made record by record.

Why Separate Them at All — Because Mixing Corrupts the Statistics

The classification is more than tidy labelling, because blending these categories distorts every aggregate downstream.

Grim Dawn's three expansions arrived in October 2017 (Ashes of Malmouth), March 2019 (Forgotten Gods), and July 2026 (Fangs of Asterkarn). Treat those intervals as 'cycles' and you get roughly 530 days and roughly 2,675 days. Fold those into a season-cycle aggregate and the genre-wide figure becomes meaningless. Expansions are not periodic in the first place, so the concept of a cycle simply does not apply.

The corruption runs the other way too. Diablo Immortal's seasons have a median cycle of 28 days — shorter and more regular than any other game we track. Drop its 2 expansions into that sample and the 28-day rhythm blurs; push its 51 seasons into the expansion list instead and you invent a game that ships expansions constantly. Only accurate typing lets both stories keep their own shape.

The classification does more than drive statistics — it also governs what the site surfaces first. Because expansions arrive only every few years, listing past ones with the same prominence as current seasons buries whatever actually matters right now. So expansions are held to **one entry per game in the main list, with already-released ones folded into a history section**. The entry that stays up is the most recently released expansion, or an upcoming one where a game has it.

There is one deliberate exception. **An expansion that has not yet released is never folded.** An imminent expansion is the single most important item on that game's page, so it always stays expanded. Only previously released expansions fold, and entries classified as seasons or updates are unaffected by the rule entirely. For Grim Dawn, that means the most recent of its three expansions holds the main list while the 2017 and 2019 entries move into history — none of the three becomes inaccessible.

What Our Data Is Missing — the Raw Numbers

A data site earns trust not by appearing complete but by stating what it does not know. Here is where things stood on 19 July 2026.

The Largest Hole — End Dates

The weakest column in our data is the end date, for two reasons.

**First, a large share of end dates are calculated rather than observed.** Studios rarely announce a firm season end date in advance, so when a next-season start is confirmed we treat that point as the prior season's close and fill accordingly. The approach lands close to reality but not exactly on it, since some games leave a gap of several days between one season closing and the next opening. This is why no aggregate on this site uses the phrase 'season length'. We use only 'season cycle', the interval between start dates, because start dates are confirmed rather than inferred.

**Second, the count shown in the listing differs from the count feeding the statistics.** Of the 234 confirmed records, 216 intervals actually enter the cycle maths. Two things account for the drop. Each game's most recent season has no successor and so produces no interval — eleven games, eleven intervals lost — and the entries for Grim Dawn and No Rest for the Wicked are excluded wholesale because neither has a season cycle to measure. Because we measure cadence as the gap between starts rather than as season length, rows whose end date was never confirmed can still contribute. Hero Siege benefits from that most visibly: its 16 classic-era records carry no verified end date but do carry a start, so this game's cycle statistics rest on 25 intervals.

That is why we deliberately show a different count for the list than for the statistics. Seeing 234 and 216 side by side on the same page is not an inconsistency — it is us refusing to hide which rows actually entered which calculation.

Practical Guidance for Readers

Three things will keep you from misreading any number on this site.

One: start dates can be trusted. They rest on official announcements and each record carries its source link. Two: a large share of end dates are estimates, so read them as 'around this time'. Three: derived metrics — cycles, medians, volatility — are all point-in-time snapshots that shift slightly as each new season is confirmed.

And on pages for games that hold no season records, the absence of cycle statistics is correct behaviour, not a data shortfall — the concept does not apply to those titles. How we operate the collection pipeline and at which stage human review enters is covered separately on our methodology page.

Finally, every figure in this article carries an 'as of 19 July 2026' qualifier. That is not boilerplate. We are actively adding tracked games, and each addition changes the total, the per-type counts, and the completeness ratios. Three games joined in this very snapshot, and both the per-type counts and their composition moved as a result. So we write numbers as 'this many, as of this date' rather than 'this many, currently'. If you return to this article and the figures differ from what the site shows, that is not an error — it means the dataset grew, and we update the in-text figures alongside the reference date.