An open beta on a private server has nothing to do with a beta in the marketing sense. It is not a product showcase and not a reward for registering early. It is the last chance to find out how the system behaves under real load — before the people arrive for whom there will be no second attempt. At Valor this stretch took a month: from mid-March to the tenth of April 2026.
A calendar of one month
- 11 March — the open beta test is announced; in the same post the project pushed back the release date.
- 15 March — launcher update 1.0.2, which required a full client reinstall.
- 16 March — maintenance, preparation for the open beta.
- 27 March — launcher update 1.0.5.
- 28 March — launcher update 1.1.5.
- 2 April — Valor Chat update: voice chat and screen sharing.
- 10 April — the open beta closes and release preparation begins.
- 18 April — release.
The density of that list says something about the character of the period all by itself: three launcher versions in two weeks is not a planned development schedule, it is the pace of reacting to what live players are showing you.
Load cannot be written by hand
A synthetic test can open a thousand connections and fire off a thousand identical requests. What it cannot do is distribute them the way people distribute them. Real load on a 3.3.5a server is almost always uneven: players gather where it is profitable or convenient. Starting zones in the first hours, the bank and the auction house in the city, the narrow corridor in front of a dungeon entrance, the same vendor for everyone at once.
It is exactly this clustering that surfaces things you never see in quiet conditions: the database query queue, updating the list of visible objects, spawn handling, chat delays. A core can hold the total number of connections and still sag noticeably when half of those connections are standing on one spot on the map. There is exactly one way to check that — invite people.
Live players instead of test scenarios
The second thing an open beta gives you is unpredictability. A tester works through a list: picked up the quest, turned in the quest, ticked the box. A player does it differently. They approach a task from the wrong side, break the order of phases, abandon a chain halfway, find a shortcut past what was intended, and almost always look for an edge — the fastest way to get gold, reputation or an item.
For a server with x3 rates and its own content this matters fundamentally. Custom races, mount crafting, hardcore mode, the Dota mode — there is no reference version to compare their behaviour against: no original exists that says how it is supposed to work. The only way to find out where a mechanic like that misbehaves is to hand it to a large number of people and see what they do with it.
The economy is a subject of its own. The rate at which gold flows in, auction house prices, whether professions pay for themselves: all of that only adds up out of mass behaviour. On ten test characters you can see that a recipe works. That the recipe is more profitable than anything else in the game becomes visible on a thousand.
The launcher on other people's computers
The third job of a beta is technical and the most thankless. The client gets installed not in a lab but on machines the team knows nothing about: different Windows versions, antivirus software, permissions, paths with Cyrillic characters, old 3.3.5a builds people have accumulated over the years. Update 1.0.2 required a full client reinstall — probably the worst thing you can ask of a player, and far better asked in March during a beta than in May after release.
The Valor Chat update of 2 April, with voice chat and screen sharing, belongs to the same category. Voice means microphones, drivers and network restrictions at providers; screen sharing raises performance questions. Features like these either get road-tested on a live audience or they do not get road-tested at all.
Why the test closed eight days before release
The closing date of the open beta is not a formality, it is a calculation. Too short a gap before launch means the fixes accumulated over the month go live untested: they get applied in the final twenty-four hours and everyone hopes for the best. Too long a gap devalues the test itself — the data goes stale, the final build diverges from the one that was tested, and the interest the beta warmed up has time to cool off.
Eight days is enough for a specific set of work: apply the fixes, build and run through the final version, prepare the starting state of the world and the database, ship a launcher update and give people time to download it. At the same time, eight days is too short to forget what all of it was for.
For the server itself and its rates, see the 3.3.5a server page; the chronology of announcements from that period is preserved in the news.
