I just came across this thread and thought I'd add a few things.2010
(just to add to what Sprung
wrote): The tech tree was already flat in CA. The switch to ZK was less a redesign and more a way to remove the last of the OTA models inherited from BA. The bot factories had already been diversified and were reasonably well covered by models in CA, but the vehicle and aircraft factories had a lot of double up and old models. The 1faction fork started off as a crazy idea to skip all this work but then gained traction. ZK started with:
Four bot factories, shuffled around from CA to lean into the cloak/shield/spider/jump differentiation that was already being developed. There were many proposals and deciding on the best was tricky.
Three vehicle factories, which were relatively straightforward to create. The handful of interesting Arm units were made into hovercraft (kitbashes by Saktoth of Scifi's hovers), while the Mr. D models carried on as Tank and Rover (previously just called Light Vehicles).
Two air factories, which were just a rotation and merge of the unique units from the mixed CA air factories.
A ship factory, made up of the models that existed and left to be dealt with later.
Most of the Striders we have today. There were more, but they tended to be duplicates and lacked models. I recall there being a laser+emg strider very similar to Dante, and long ranged tacnuke nuke artillery for the corresponding role of Merlin. These units can probably be found in BAR.
Amphbots were added around 2011-2012.
I think the initial motivation was inclusion in linux distros (there were more linux people around back then). This was also about the time we switched from caspring hosted svn to google code svn. The switch to ZK and google code lost us a fair number of contributors (some hated google), but we had previously lost people who couldn't see CA removing the OTA content any time soon, so there was a tradeoff.2012
(ish): Zero-K splits off to its own infrastructure after disagreement with infrastructure developers. From my perspective they were very stubborn regarding extending the protocol to allow for new stuff (Eg matchmacking, more advanced planetwars) and would make sudden changes that broke our autohosts.2013-2014
: Evolution RTS is greenlit on Steam and released in 2014. I don't think Steam was on my radar at the time, but now it looked like a possibility. Looking back at the dates we actually put up a greenlight campaign five days after the Evolution RTS release.
Zero-K was greenlit, although opinions varied on when or what sort of release we should go for. I thought there was a lot of work to do before ZK looked like a real game, given how poorly the IRC+BattleRoom paradigm of old Spring fared for the initial Evolution RTS release. We were on a bit of a time limit though, since active developers were dwindling to low single-digit numbers, and the playerbase was dropping off with little influx from Spring or advertising.2015-2017
: A lot of polish work that isn't necessarily interesting for people who started out playing then contributing to a multiplayer RTS. Campaign, matchmaking, tutorials, AI, UI, optimisation etc... There was still a bit of time to do easier and more fun stuff. gajop
saved us with the lua framework for the lobby, as that let me pick it up and write most of the lobby myself, freeing up Licho
: We seemed about ready, so we released on Steam about four years after greenlight. Steam was a lot more crowded after four years, but I think we did ok out of it. The impetus to make ZK look like a game, with polish and singleplayer, was also an important result of Steam. It also provides an easy install process. 2020
(ish): I decide (or at least decide to act on the theory) that longevity comes from modding, restarting the process that kicked it off (sort of) 23 years ago. Attempts are made to efficiently use our limited manpower to better support modding. Luckily Spring is built on this sort of stuff, so it just required a bit of infrastructure and lobby work.2021
: Here is what we look after a decade.
"daily unique players" is the number of unique accounts that were a player in a game that ended up generating a replay (ie no spectators, singleplayer, or exited games). Each point is the average of 28 days.
The bump in March 2020 is an aberration that appeared to affect many games. No need to dwell on it.