About this game Ember Vault
Development notes
One engine for play, preview and proof
Play, resume, card previews and the server's daily check all call one deterministic step function. A rejected action changes nothing, not even the random generator; previews run on a copy and show public numbers only, never your draw order. Runs begun under rules 2 to 4 keep their old numbers and still verify, but the daily board takes only the current rules.
Measuring each card with a bot
I added one copy of a card to a fixed decision bot's starting deck and ran 300 seeds at heat 1. Firebrand raised the bot's win rate by 18.7 points and Conflagration by 17.7, while the rares Fallen star and Flame forge lowered it by 12.7 and 4.0. Since stacked burn grows as a triangle, I cut their burn from 6 to 4 and 7 to 5, and made Fallen star 2 energy for 24 and Flame forge 1 energy for 3 strength; all four then landed between +9 and +14. These are tuning samples, not player win rates.
Rewards worth choosing
Score once counted only gold still held, so buying anything lowered it, and the bot finished runs holding 177 gold on average. Each rare was as likely as any common, and elites gave one random relic. Rules 5 counts spent gold, adds a 90-gold shop relic, makes each reward slot 15% rare and gives elites a sure rare plus a pick of two relics; the bot's heat-2 win rate rose from 19% to 31%.
Updated: