Wild Crazy 8s started with 30 house rules. Nice, polite rules. Hearts give a little bonus when scored. The first card of each turn is worth a bit extra. Rules a grandmother could love, and a grandmother could balance.
There are 158 of them now. The store page says 150+ because I kept adding rules after I wrote the store page, and I would rather undersell by eight than update marketing copy every week.
The count is not the interesting part. The interesting part is the word the store page uses: multiply. Rules in this game do not sit politely next to each other adding small numbers. They compound. Heart Throb makes every scored heart worth extra Boost. Encore pays out on every third card you play in a single turn. Put them in the same run with something that extends your turns and you have built an engine where each piece feeds the next. That is the whole appeal of the genre, and it is also the problem.
The problem with multiplication
Additive bonuses fail gracefully. If one rule is a little too strong, your score is a little too high. Multiplicative bonuses fail catastrophically. If two rules multiply when I expected them to add, the score is not off by a margin. It is off by an order of magnitude, and the player has stopped making decisions because every decision wins.
So each new rule does not get playtested against the game. It gets playtested against the other 157 rules. The regret in this post’s title is a spreadsheet where I track which rules can see each other’s output. Every time I add a rule I walk the sheet and ask the same question: what happens when this touches the ugliest thing already in the pool.
Rarity is the first line of defense. Rules come in four tiers, and the shop rolls Legendary rules two percent of the time. The engine-defining pieces are rationed. You will usually see one true engine per run, and the run becomes the story of what you found and what you bolted onto it.
The bot that plays badly on purpose
The second line of defense is a bot, and the bot is deliberately mediocre.
Here is the reasoning. If I test balance with a bot that plays well, I learn how strong a build is in expert hands, which mostly tells me about the expert. But if a weak player can win with a build, the build is carrying them. Strength of build and strength of player get separated by handicapping the player.
So the probe works like this: force a specific build into existence, hand it to a greedy bot that just plays the highest-value legal card with no planning, let it shop badly, and run that many times over. My curated control build, the one I consider fair, wins somewhere between 30 and 40 percent of those runs. Any build the greedy bot wins with more than about 60 percent of the time gets flagged. Not because a good player winning is bad, but because a bad player winning that often means the deck plays itself.
Nine archetype builds go through that gauntlet: pure chip scaling, wide engines, combo chains, economy loops, the works. Combo builds actually read low on the probe because the bot keeps resetting its own combo by drawing, so I treat their numbers as a floor and playtest those by hand.
On top of the probes there are fresh-seed sweeps that track how far runs get before they die, floor by floor, across the game’s nine floors. The shape of that curve tells me more than the win rate does. A healthy curve bleeds runs gradually. A cliff at one floor means a boss is doing something rude.
What the spreadsheet bought
The payoff for all this bookkeeping is that I get to be irresponsible in the design and responsible in the tuning. A rule like Encore, which rewards long turns in a game where turns are usually short, only exists because I could check what it does next to every turn-extender in the pool before it shipped.
There is no grand lesson here except the one every deckbuilder designer learns eventually: the fun lives exactly one step away from the broken. The craft is building fences one step past where the players will stand. 158 rules, 36 bosses, nine floors of fences.
The game hits Steam on November 10. If you want to find the combination I missed, wishlisting it is how you make sure Steam reminds you.