Build a desk. Survive the open. Cards are market-making quotes, arbitrage legs, trend follows, vol sells. Bosses are Fed surprises, flash crashes, CPI prints, earnings vol. Win a run, ship a 20-hour campaign, learn actual quant-finance mental models by accident. Built by an ex-quant on Godot 4. Balanced by 100,000 overnight sim runs.
Each run is a 3-day simulated trading week. Day = act. Morning / mid-day / close = encounters. Close-of-day boss regime. Survive Friday close, you win the run. Most runs end Wednesday.
Limit-order quote, market-taking cross, delta hedge, volatility sell, factor rotation, carry trade, flight-to-quality. Rules are reduced to game mechanics but the intuition is the same a desk uses at an actual bank.
Drawdown past a threshold and you're forced to liquidate — run ends. But capital also gates deck size, position sizing, and access to exotic instruments. The classic deckbuilder tension, re-skinned.
Fed surprise. Flash crash. CPI print. Earnings vol. Liquidity drought. Each regime disables certain strategies and rewards others. Building a deck that handles three regimes is the meta-game.
Finish a run as Market Maker, unlock two-sided quote variants. Finish as Vol Seller, unlock straddles. Meta progression isn't stat-bloat — it's actual new strategies the real desk uses.
Ascension-1: add earnings vol to the boss pool. Ascension-5: bosses stack. Ascension-15: you're trading in 2008. No HP-inflation difficulty. Just more regimes that invalidate more of your deck.
You earn the bid-ask spread, every card, every turn. Until someone knows more than you do — then adverse selection bleeds you dry. Starter deck. Teaches inventory management, spread tuning, when to step back.
Two cards play simultaneously on paired venues — if the pricing diverges, you capture the spread. If it converges against you, both legs bleed. Teaches execution risk, venue liquidity, when to close vs. press.
You need two green bars in a row before you enter. Then you hold through chop — and hold through a stop-hunt — until the trend breaks. 70% of your trades are losers. The 30% pay for everything.
Sell premium, collect. The gamma clock ticks every turn — ignore it long enough and a single CPI boss turn wipes you. Steady income until it isn't. Teaches when the music stops.
Every indie deckbuilder has the same problem: one card breaks the meta, and you don't find out until playtesters are already furious. We built a Python simulation harness that plays the game against itself 100,000 times per night. Any card that plays in <2% or >25% of winning decks gets flagged before a human sees it.
Studios balance by feel, by Discord complaints, by the one streamer whose take gets amplified. We balance by 500,000 simulated runs a week, filtered through a statistical lens the designer can read in five minutes. A card change takes minutes to validate, not weeks. When we ship a patch, we know which archetypes got better, which got worse, and by how much, before anyone downloads it.
The designer is a CS senior who shipped ML systems in production. The game is the portfolio piece. The sim is the reason it balances better than its neighbors.
Proof: sim/runner.py v0 runs 500 games in <1s on a laptop. Flags cards. Ready to scale. Already in the repo.
One email per month. Build updates, sim-balance deltas, playtest findings, screenshots. When the Steam page goes up in Oct 2026, you'll be the first to wishlist. When the demo drops at Next Fest, you'll have early access. No spam, unsubscribe one-click, I read every reply.