The Rung That Made Players Worse

20 Below’s Skill ladder began with Novice, a first step meant to show early progress. Playtesters complained about it, and the team raised a concern. The numbers confirmed it: at low Attributes, a Novice rolled worse than someone with no training at all. I kept it at first, then retired it three days later. Retiring it well meant making sure it didn’t punish players somewhere else, and carrying every saved character across.

Role: Producer: rulings.
Team: The playtesters and the 20 Below team, who raised it, and Claude (AI assistant).
Timeline: 25–28 September 2026.
Result: Training is never worse than not training now, and every saved character carried across with its points refunded.

The Problem

A Skill roll in 20 Below is 2d10, and you succeed at or under a target. An Untrained character ignores their Attribute: the target is the Difficulty alone. A Novice adds the Attribute, but rolls at Disadvantage, three dice with the highest two kept. The rules promised that each tier was strictly better than the last.

That was true for most characters. It wasn’t for the ones with the lowest Attributes. For them, the Disadvantage cost more than the Attribute added. So a new player who put their first point into a Skill got worse at it.

Analysis: Who Novice Failed

A new player with a low Attribute

Spent a point on training and rolled worse than if they hadn’t.

A player choosing where to spend

Was told each tier was strictly better, which the dice didn’t bear out.

A returning player

Had Novice Skills saved on a character, and needed them to follow the rules without losing anything.

A player testing builds

Wanted to try the new ladder without starting a character over.

The concern came from the table first: playtesters complained, and the team raised it. That is why the simulations were run. They used our five test characters, who have the same score in every Element, from Fragile Frankie at 1 to Maxed Maggie at 10. The exact odds settled it.

Chance to succeed on a Skill roll at Difficulty 7, by Attribute 1 to 10. Untrained is flat at 21 percent. The retired Novice tier is 11 percent at Attribute 1 and 15 percent at Attribute 2, below Untrained, and crosses it at Attribute 3 with 22 percent. Trained, now the first rung, is 28 percent at Attribute 1 and 94 percent at Attribute 10. — At Attributes 1 and 2, Novice sits below Untrained at every Difficulty. Mediocre Mike, at 3, is exactly on the crossover.

Design Decisions

It took five decisions, made over three days.

A timeline. Before 25 September, playtesters complain and the team raises a concern. On 25 September the simulation confirms it, and the first ruling is to leave it, for movement. On 28 September at 9:12, the ruling is to retire Novice: the disadvantage is too punishing for all of the players. At 9:19, keep every tier by name: reducing the damage allowed is punishing them in a different way. At 10:04, add Reset Skills because players are testing builds. At 10:10, the change is live, with the desktop releases the same morning. — From a complaint at the table to a retired rule, coloured by who drove each step.

First, leave it

The finding came up while the movement chapter was being written, and I ruled on it there: leave it. Movement has nothing to do with Skill use. The roll in that chapter is Athletics, an Everyman Skill, so every character starts it at Trained and Novice never comes into play. A player can also argue a different Element with a Descriptor.

Then, retire it

Taken up on its own, as a question about Skill use, the answer was different. My ruling: “the disadvantage is too punishing for all of the players.” Novice came out of the rules, the book, the Creator, the character sheet, the Battle Tracker and both Owlbear extensions.

The Skill ladder before and after, matched by name. Before: Untrained 0, Novice 1, Trained 2, Adept 3, Expert 4, Master 5. After: Untrained 0, Trained 1, Adept 2, Expert 3, Master 4. Creation points to reach Trained drop from 2 to 1 and to reach Master from 5 to 4. XP from Untrained to Master drops from 23 to 15. Social dice stay Trained 3, Adept 4, Expert 5, Master 6. Everyman Skills are still free at Trained. — Six rungs become five. Every rung gets cheaper, and every rung keeps its dice.

Don’t punish them somewhere else

There were two ways to renumber the ladder. Tier + 1 would shift every Skill’s dice down one step. Keep by name leaves a Trained Skill with Trained dice. Tier + 1 would have taken a die off everyone’s free Social attack, because Persuasion is an Everyman Skill and starts at Trained. My ruling: “removing the novice penalty was to prevent punishing the players, reducing the damage allowed is punishing them in a different way.” So dice stay with the name, and the weapons that sat at Novice move into Trained. Nobody loses reach.

Landed Social attacks needed to Fluster a target, keep by name against tier + 1, at every tier. Against an NPC with 5 Poise, a Trained attack takes 2.7 landed attacks under keep by name and would have taken 3.8 under tier + 1. Against a PC with 8 Poise, 4.1 against 6.0. Keep by name is better or equal at every tier. — Exact averages against Presence 3. Tier + 1 would have made everyone’s free Social attack weaker.

Carry every saved character across

Characters already existed, so the change had to reach them without anyone losing what they’d paid for. A character saved on the old ladder is converted when it opens, in the Creator, the character sheet, the Battle Tracker and the Owlbear sheet. Every tier keeps its name, and Novice becomes Trained. Anything paid for with Discretionary points or XP is recounted on the new rungs. A purchase that only paid for the old Novice rung gives its points back, and the Creator says once, in plain words, what changed.

A test character saved the day before the change, opened after it. Security and Investigation go from Novice to Trained. Insight stays Adept and gives back one Discretionary point. Every other Skill keeps its name, with the number one lower. The Creator's notice reads: Novice is gone from the Skill ladder. Investigation, Security are Trained now, at no extra cost. Trained now costs 1 Skill point instead of 2, so you may have Skill points to spend. The character has 4 Skill points to spend and 1 Discretionary point back. — A test character, saved before the change and opened after it, through the Creator’s own conversion.

Let players rebuild, and close the gap that broke a sheet

My players were building and testing characters, so the Skills step gained a Reset Skills button. It puts every Skill back where a new character starts and returns every point spent on Skills: the pool, Discretionary points and XP. The same day, a player’s character was reading −1 Skill points. It traced to a gap: a creation step let you lower something below what Discretionary points had paid for, and a later refund took it under its floor. Every creation step now stops at that floor: Attributes, Skills, Resources, Gift Levels and Gift Adders.

Four steps, each allowed on its own, rebuilt with a test character. One: Firearms at Expert, one tier bought with a Discretionary point, Skills pool at 0. Two: Firearms lowered to Untrained on the old Skills step, pool shows 3. Three: the 3 points respent, pool at 0. Four: the Discretionary tier refunded, Firearms below Untrained, Skills pool at minus 1. The fixes: Firearms now stops at Trained on the Skills step, and a Reset Skills button returns every point. — Rebuilt with a test character. No player’s file was used.

How the Work Ran

The table raised it; the numbers confirmed it. Playtesters complained, and the team raised a concern. The simulation turned that into exact odds I could rule on.
One change, everywhere at once. The rules, the book (formatted chapters, manuscript and web edition), the Creator, the character sheet, the Battle Tracker and both Owlbear extensions changed together. The Creator v0.12.34 and the Battle Tracker v0.3.4 went out the same morning.
Tested on copies. The conversion and the −1 were both checked with test characters, never a player’s file.
A release held for a ruling. Two AI working sessions were changing the same codebase at once, one on character portraits and one on Novice. The portrait commit staged a whole shared file that already held the other session’s unfinished conversion. For about five minutes, the live Creator converted saved characters while the rules still said Novice. The other session checked, and the conversions were correct under the new rules, so no saved file was damaged. Before building the installers, the portrait session found 28 files it hadn’t touched and stopped to ask. I ruled to release from main once the Novice change was committed, so the installers, the extensions and the live site all matched.

A timeline of 28 September from 10:04 to 10:37 in three lanes. The portraits session commits at 10:06, having staged all of state.js, and holds the release after finding 28 files it hadn't touched. The Novice session commits at 10:10 and at 10:12 confirms no saved file was damaged. The producer rules at 10:15 and at 10:30: release from main with both. Extensions are synced at 10:31 and both releases are published at 10:35. On main, 10:06 to about 10:11 is marked as characters converted while the rules still said Novice; from 10:11 the rules and conversion agree. — The lesson: check for anyone else’s edits before staging a shared file, and release from a known commit rather than the working folder.

Outcome

Training is never worse than not training now. The first rung is Trained, it costs half the creation points it did, and every rung above it is cheaper too. No one lost dice or reach in the change. Saved characters came across with their Skills named as before, points back where they’d paid for the old rung, and a notice saying what changed. Players testing builds can reset their Skills and try again.

What this doesn’t show yet is the table’s verdict. The playtesters who raised Novice are the ones who’ll say whether the new first rung feels right, and their next sessions are the real test.

Credits

Rulings	Me, as producer
The complaint that started it	The 20 Below playtesters
The concern that led to the simulations	The 20 Below team
Simulations, option tables, Package Deliverables and figures	Claude (AI assistant), checked by me