← Project Management · Instructional Design · Visual Information

Case Study

One Spider, Checked Three Ways

Real-time design and collaboration on 20 Below, a tabletop RPG I’ve produced since 2021: a gap in one monster’s venom, caught by three separate checks, fixed in the open, and tested with simulation before it ever reached a table.

Role

Producer: decisions and rulings.

Team

A GM playtester, George Szabo (game design) and Claude (AI assistant).

Timeline

One day, 24 September 2026.

Result

A rule left to the GM became one players can feel, simulated before play, and now the standard for every venom in the bestiary.

The Problem

The Brown Recluse’s stat card said its bite did no immediate damage, left the target Envenomed, and that “the wound worsens without treatment over the following days, severity the GM’s call.”

That last clause looks like a rule but doesn’t do anything. Worsens how, for how long, and what stops it? Every GM would answer differently. And the one creature in the book known entirely for its necrotic bite did nothing a player could feel.

How It Surfaced

Two people and an AI edit pass found it separately.

I didn’t catch it myself. What I did was check the AI’s flag against the team’s feedback and confirm that all three were pointing at the same gap. That’s the part of the process I’d defend: a machine’s flag counts once a person has checked it against what the team is saying.

Analysis

Naming the problem showed it was bigger than one spider.

• The book had two poison rules. The official condition, Envenomed, gave Disadvantage on Physical rolls. A separate bestiary trait had its own Water roll and cost an Exhausted level instead. I made Envenomed the only one, and the bestiary was updated to match.
• The recluse is necrotic, but its card did no damage. Whatever the fix was, the bite itself had to hurt.
• “GM’s call” was carrying a rule’s weight. A GM’s call is fine for judgment. It isn’t fine for a mechanic players will want to plan around.

Design Decisions

1. Mine sources, then propose. The assistant mined sources for concepts and found one pattern again and again: slow harm with a treatment window that closes. It proposed five mechanics built only from rules the game already had. My feedback was that it was too complicated. We wanted depth, but not excessive crunch.
2. Take the pieces that worked. My revision:
   • The bite does damage, with more dice for bigger spiders, and it counts as boosted.
   • From day 2, the victim rolls Water + 8 once a day. Succeed on day 2 and it’s over. Fail, and it takes three successful days to shake off. Each failed day costs an Exhausted level.
   • Antitoxin ends it.
   • Supernatural healing negates the venom, but the tissue is already dead, so it leaves a scar unless the healing can regrow a limb.
3. Throw dice before accepting anything. I asked for a simulation: 100,000 bites against four test characters, from a frail one (Fragile Frankie) to a maxed one (Maxed Maggie). The first question was how hard the bite should hit.

4. The numbers found a spiral nobody had seen. Exhausted 2 means Disadvantage on every roll, including the roll to resist the venom. Each failed day made the next one harder, and 56% of Frankies ended unconscious. My ruling was that the resist roll is never at Disadvantage. We also tested letting Envenomed’s own Disadvantage apply. That pushed Frankie to 80%, so it was rejected.
5. Make care worth getting. Medicine gives Advantage on the daily roll, but it never ends the venom on its own. With someone tending Frankie every day, the chance of ending up unconscious falls from 40% to 6%.

6. Let one fix become the rule. Writing the recluse’s rule exposed a gap everywhere else. Ordinary venom had no roll to resist it, so without Antitoxin a snakebite lasted forever. Every venom now gets a daily Water + 8, and one success ends it. The recluse is the tougher version, needing three. Along the way, Ki stopped being able to end any venom, Antitoxin’s equipment entry was rewritten to end it outright, and the Mutated recluse became the exception: it rolls every 6 hours, Antitoxin only gives Advantage, and Medicine does nothing.

The assistant’s work was checked the same way. Its first odds table for the Water roll assumed Attributes ran from 1 to 5; they run to 10. The assistant caught its own mistake while setting up the simulation, said so, and the correction went into the record alongside everything else.

Outcome

A clause that left everything to the GM became a rule players can feel and plan around. It was tested against the weakest and strongest characters before anyone rolled it for real, and it became the standard for every venom in the bestiary. The rules, the book, the status-effects page, the equipment chapter and the printable bestiary changed together, in one pass.

The process is the point. A playtester, a designer and an AI edit pass each caught the gap on their own, and a person confirmed they matched. The AI mined sources, proposed and simulated. Every decision stayed with the team, and each one was made with the numbers in front of us.

“I’ve spent a lot of my time on 20 Below trying to keep the rules balanced, from the damage scale and critical hits to the venom on a single spider. Feral makes that work possible: every design question gets tested with simulations and real numbers before it goes to the team to decide, and the result goes into every part of the game at once. I’ve never had to wonder whether a balance fix actually landed.”

Credits

Raised the problem	A GM playtester (uncredited by request); George Szabo, game design
Decisions and rulings	Me, as producer
Mining sources, proposals and simulation	Claude (AI assistant), checked by the team