Season’s greetings to you all!  It’s almost 2017, which means the Sub Terra Kickstarter is juuuuust around the dark and shadowy corner.  As a late present, we’re releasing the latest draft of the rulebook for your perusal.  Give it a quick read!  I’ll wait.

These rules are almost certainly going to be what we ship with (though feedback is always welcome!).  They’ve existed in roughly this form since July, and have been nearly fully locked down since September.  This allowed the previous posts in this series to accurately reflect the game as you’ll encounter it in the wild, but the absence of a presentable set of rules has made it a bit awkward as I’ve had to reference concepts indirectly.

Therefore, in this post, I thought it’d be interesting to give some direct commentary – explaining the process and reasoning behind why things ended up the way they did.  This is going to be more of a gigantic DVD extra of behind-the-scenes release notes rather than neatly structured prose.  It’s probably something you’d want to dip in and out of rather than devour in one sitting.

(I was inspired by this post by Donald X. Vaccarino giving card by card commentary of the Dominion base set.  It’s clearly ridiculous to compare Sub Terra to Dominion, but hopefully this may be of interest to future game archaeologists in a similar way)

If this isn’t your sort of thing, I give you permission to stop reading now and go enjoy yourself.

Still curious?  Kick back with some mulled wine, keep the rulebook open in a separate window, and follow me…


  • Sub Terra wanted to be a purely cooperative game.  The main tension in the game is that between sticking together for safety and spreading out for faster exploration – the game doesn’t really work if there are too few cavers (as the isolation is too hard to overcome) or too many (as you’re always near friends who can rescue you when you fall over).  Four to six cavers felt like the correct number, so rather than limit the number of players to this range I allowed players to control multiple cavers at low player counts.
  • Cavers start with three health points.  This has remained constant since the very first prototype – it’s an aesthetically appealing number and provides enough strategic room without being overwhelming (i.e. being on one/two/three health all carry distinct levels of risk depending on your situation, in a way six distinct levels wouldn’t without a heavy increase in the complexity of the game system).
  • The number of tiles used directly controls game length.  My aim was always to create an experience that lasts “around an hour”, with enough texture to keep that hour sufficiently interesting without becoming overwhelming.  Earlier versions of the game had more tiles and fewer tile types, flaws which were gradually corrected.  I’ve also tried out a “quickplay” half-hour version for conventions which uses a simpler mix of half the tiles, which works quite well – this might find its way into the rulebook before launch if there’s sufficient demand.


    First paper tile prototypes (orange squares are cave-in tiles)

  • The size of the hazard deck is the “clock” that controls the level of challenge – the larger the deck, the more turns you have to escape the cave.  The numbers in the table were arrived at through healthy trial-and-error.


  • The round structure has always broadly been “do actions, then do hazards”.  I settled on the one-hazard-card-per-round thing fairly early on, after experimenting with dice-based alternatives that scaled with the number of players (these were all super inelegant).  Balance is kept because the hazards themselves mostly scale to the number of players.

  • Horrors move before resolving a hazard card because:
    • (a) This was the cleanest way to stop horrors from moving off their spawn tiles in the same turn.  Keeping unseen threats contained to their own tiles (and not neighbouring tiles) was critical to keeping the push-your-luck gameplay fair.
    • (b) Because there aren’t always horrors in the cave, it’s easy to forget to move them once they appear.  The solution we went with was to add a horror reminder marker that sits on top of the hazard deck, so you physically can’t draw a card without interacting with this horror-related thing.  I’ll concede that it’s not ideal.

  • For a lot of development, there was no starting player marker – every round started with the same caver.  This made the turn structure very easy to follow (keep going round in a circle without breaks), but it had some unfortunate side effects:

    • Cavers moving later in the turn order always had more information about what tiles everyone was ending the round on.
    • As hazards always occurred at the same point in the turn order, late-moving cavers had more time to be revived and missed fewer turns while unconscious.  A “recovery phase” was introduced so that healed cavers only revived at the end of the round, but this fought against how players instinctively wanted to play the game.
    • Certain characters were better at certain points in the turn order, allowing experienced players to game the system.  For example, the Leader is much more powerful if she always goes last.

A rotating starting player marker fixed a lot of these issues while only making the turn structure a little bit more complex.

  • The end-game scoring system changed a few times throughout development.  Initially, it was highly individual: “If your caver gets out, you win.  If your caver doesn’t get out, you lose”.  The idea here was to add more emotional heft to leaving someone behind or risking yourself to rescue a friend.  This worked great for groups of friendly altruists, but not so great if you were playing with unfriendly win-at-all-costs griefers – the way the game works, running off and finding the exit by yourself and stranding everyone else in the cave is entirely possible.  This spent some time as a pure co-op ‘shared victory’, before I realised I could safely reintroduce the asymmetry by scaling the ‘win’ based on the number of other cavers that also ‘win’.

  • Unconsciousness has had a lot of thought put into it (ironically).  Yes, it’s almost player elimination, and no, it’s not particularly fun to experience.  But this negative experience is necessary as a threat to stop players taking stupid risks, and it incentivises other players to come rescue them.  While most sessions will have one or two cavers fall unconscious, you should be able to rectify this in a few turns if you’ve been playing sensibly, and as this is a cooperative game there’s nothing that stops the unconscious player from continuing to give advice and participate in group discussions.  I did experiment with some low-probability auto-revive mechanics (e.g. each turn, if you roll a 6, regain one health), but this stopped other players from coming to pick you up as you’d get better automatically in time, and this meant you spent more time out of the game.

  • Skill checks were introduced as a way to standardise a bunch of different dice checks.  They’re set at 50-50 because this felt “fair”, even for a wide spread of outcomes.  When the success chance was greater than the fail chance, consequential failure felt much worse.


  • Individual turns in Sub Terra are deliberately short to keep control moving quickly between players.  Taking up to three actions per turn still gives you enough meaningful choices without bogging things down, and it allows hazards to interrupt multi-turn plans-in-progress.

  • The action point system flipped a few times between what you see today and a more restrictive “short action / long action” system (“take either 2 short actions or 1 long action per turn”).  The latter system was cleaner but more restrictive, and in the end I decided that “action points” were such a familiar concept to most gamers that the actual complexity cost to using them was fairly low.

  • Exerting was an early addition to the game.  Using Explore as the last action you take in a turn is very risky as you don’t know how dangerous your destination tile will be, yet repeated Exploration is crucial to getting through the full tile stack in time.  Exerting added an extra safety net to the double-Explore turns (you can always exert-move back to your previous tile), as well as allowing you to convert health into actions at points of crisis (e.g. rescuing friends, dashing to the exit).  It was a very simple way to add significant depth.

  • The basic actions didn’t change much throughout development.  The triad of Reveal/Move/Explore was in from the start, and Heal was added once it was clear that a single dedicated healer character wasn’t enough.  Run was added a bit later to make backtracking easier/faster, which in turn helped exploration become more fun as you could safely push further before having to turn around.

  • The hazard actions, however, changed a number of times.  The very early drafts of the game had vastly more specialised roles – for example, only one character could place ropes, and only one character could clear cave-ins.  This ensured everyone had a purpose, but because that purpose was usually critical it made decisions obvious (“keep doing that thing you do”).  To keep the game interesting, it made sense to allow every caver to tackle every obstacle inefficiently, and allow certain characters to do certain actions faster (along with other, less critical, bespoke perks).  There’s now more tension between working inefficiently by yourself and calling a specialist over to do that thing instead.


  • From the start, I always wanted horrors to be nigh-unstoppable deadly threats.  This wasn’t a zombie-fighting game – I wanted the focus to be more on teamwork, risk and evasion rather than linear combat.

  • Horrors started out by moving semi-randomly, but the dice overhead was very irritating.  A consistent “slowly chase the closest caver” behaviour kept the pressure on while allowing multiple ways to deal with them regardless of which character you were.

  • The despawn for distant horrors was added to stop situations where horrors got too far away to matter, but you still had to move them every round.  The seven tile limit was just long enough to make escape a challenge, but short enough that the horrors could effectively catch you or cut off your escape.


    • The five hazard types (tremors, floods, gas, cave-ins, horrors) have all been around in some form since the first prototype – there’s only so many grokkable things that can go wrong in a cave environment.  Some had drastically different effects, though.


  • Tremors were kept functionally the same throughout development, but have gone through several name changes (including “Rockfall” – which was too close to “Cave-In” – and “Injury”).  Because they can hit you regardless of where you’re standing, they add to the feeling that you’re not safe anywhere, and make going down to your final health point feel that much more dangerous.
  • Floods always placed flood markers onto the board, but they initially worked by growing the flooded area every time a flood occurred (spilling onto other unrelated tiles).  While flavourful, this sometimes led to massively overflooded passageways that were impassable in a reasonable amount of time, which was unsatisfying as it forced a loss outside of player control.  The final version keeps the desirable gameplay of sometimes making backtracking more difficult while removing the excessive variance.
  • The initial concept was for Gas to be of the explosive variety.  Like floods, gas clouds used to emerge from a point then spread outwards.  Cavers moving into the gas cloud had a chance to trigger an explosion, which would remove the cloud but knock-out everyone caught inside it.  This ended up being too complex and dice-heavy to be fun, and served a similar function to floods in that it hurt when backtracking.  The final version is themed around choking, toxic gas, and keeps the significant level of danger while also changing the value of exploration in “gas turns”.  Only dealing two damage allows “being on three health” to be a more aggressive way to tackle the hazard.
  • Cave-Ins always placed rubble markers on tiles, but the mechanism to determine which tiles caved in has shifted a lot.  The purpose of these was always to block off passageways rather than deal a lot of damage (though the low probability of high damage is important deliberate psychology).  To keep them distinct from floods, the intention was that these were worse if you were by yourself because they were costly to clear, whereas floods were worse if you were in a group because they taxed everyone equally.  The “one-in-three” chance of collapse gave a good spread of cave-ins throughout a game session, and this eventually moved from being all {12}{34}{56} to unique number pairs to make it harder to sit on disjoint sets, upping the risk of a blowout.
  • Early drafts had no limit on Horrors – some games had five or more horrors chasing you at once and were clearly unwinnable.  Intermediate versions capped horrors at three, but had a complex “build-up” mechanic that allowed “dead” horrors to more aggressively respawn later.  The final version ditched that complexity for a straightforward one-card, one-spawn system.  The additional move granted to horrors was added to stop them being so predictable – the fact that they can move two steps in a turn helps keep you feeling uneasy when they’re nearby.
  • The severe hazard cards were a fun way to make harder difficulty settings more surprising.  The Flood+ and Gas+ are especially good examples of this, as they increase the amount of health lost by those events by one.  I’ve lost track of the number of times I’ve been knocked out from full health by standing on a gas tile “because the worst that can happen is that I lose two health”.  This keeps you on edge, and stops the game becoming too predictable.

  • For a large part of development there were also miracle cards that had a positive effect rather than a negative effect (e.g. “remove all horrors”, or “everyone gains one health point”).  This did result in some fun moments where you’re saved after thinking all had been lost, but they ultimately added far too much variance in terms of player challenge (a game with three miracle turns has three less hazards to deal with AND three bonus effects).  Cutting these was a good move.


    Miracle playtest cards


  • Initially, turn order remained constant, and this was a “recovery phase” where unconscious cavers who’d been healed that turn regained consciousness again – the implication being that you could be unconscious AND on positive health during the preceding round.  Understandably, this confused a lot of people.  It was simpler to introduce a rotating starting player and keep “unconscious” and “on zero health” tightly linked states.

  • It would arguably have been “better” to have the starting player marker rotating anticlockwise, in the opposite direction to the action phase’s clockwise turn order.  This is because it minimises the longest gap between player turns:
Turns before your next turn (5 player game) Longest gap
No rotation 4 – 4 – 4 – 4 – 4 4
Clockwise rotation 8 – 3 – 3 – 3 – 3 8
Anticlockwise rotation 5 – 5 – 5 – 5 – 0 5

If we assume that prolonged downtime is a bad thing, then it’s clearly better for everyone to wait five turns (and occasionally have a “double turn”) than for one player to occasionally wait eight turns.

As it turned out, having two directions of motion was really confusing for most testers, and the cooperative nature of the game meant that downtime wasn’t that bad anyway as you could always communicate and discuss strategy.

If this bothers you, feel free to switch to an anticlockwise rotation – the game will function almost identically.


    • The “In the very unlikely event…” rule has never actually been invoked at any point during testing.  It’s there because it’s mathematically possible to draw four dead ends before you draw a T-junction or crossroads, and at that point you’re stuck unless you brought the Engineer along to blast a hole out of your claustrophobic death cave.  While I’m sure this will happen at some point out in the wild, it’s not the real reason the rule exists: it was written to stop people worrying that it could happen and always picking the Engineer “just in case”.

      Without the rule, these occurrences will mainly happen in the first few rounds of the game, and at that point you can easily restart without being too angry at me or RNGesus.

    • The set connectivity of the tiles (number of dead-ends, number of crossroads etc.) has increased over time.  Initial tilesets were very claustrophobic – I perversely liked how annoying running into dead-ends felt, as this does a great job of feeding your despair.  However, this wasn’t ideal from a game playing perspective, as this vastly reduced the number of exploration options available, which meant players had less to think about.  The more open cave we have now results in many more “press on”/”regroup” decision points, while retaining some of the infrequent moments of obstruction.

    • (The water, gas, cave-in, and horror tiles have been discussed in the Hazard Phase section, above)


  • Squeeze tiles have gone through a ton of changes since the first prototype.  I didn’t feel I could make a game without caves without including these – they’re close to the first thing people think of when discussing caving.  The trick was finding a version of them that felt intuitively correct while still remaining fun and easy to comprehend.  I experimented with two-space tiles (a nightmare with terminology and explosives tokens), single caver limits, and skill checks.


    Prototype squeeze tile (two-space)

    Eventually, I settled on the final version which acts like a pre-flooded water tile – it’s not something you really want to push through unless you have no choice, as you’ll either have to backtrack through it (costly) or the rest of the team will have to follow you through it (even more costly).  Allowing an exploring caver to enter the tile for a single action point does feel a little unintuitive to me, but this was far more interesting to play with than the version which just turned the Explore into a Reveal.

  • Originally, Ledge tiles were the only tiles to use ropes – they functioned much as they do now (i.e. ‘soft’ dead ends), except they briefly had a “two-space tile” thing going on where the rope bridged the gap between the two spaces (“chasms”).  When their rules were merged with Slides about halfway through development I was able to throw out the confusing two-space thing entirely.


    Prototype chasm tile (two-space)

  • Slide tiles used to function like giant pushy arrows that forced cavers that moved onto them out the other end into a completely new tile.  I wanted a threat for reckless explorers that wasn’t just potential damage or unconsciousness, but separating the group against its will.


    Prototype slide tile

    It quickly became clear that you needed some way to get back over Slides, as if you were pushed into a dead-end then only the Engineer’s explosives could rescue you.  After experimenting with a bunch of different alternatives (skill checks, long actions, having a team-mate nearby) this merged with Ledges to make ropes more of a thing.

  • Rough Terrain tiles were the last tile type to be added, and went in thanks to some complexity budget being freed up by the Ledge/Slide merger.  They act as yet another threat to reckless exploration, and add some interesting Squeeze-like decisions on whether you should press on through them or go back the way you came.


  • Originally, there weren’t any characters.  I knew the game needed asymmetry between players, but this started out in the form of equipment instead of defined roles.  The Shovel let you dig out caved-in tiles, the First-Aid Kit let you heal cavers back to full, the Explosives let you blast holes in walls, the Rope let you place rope tokens (this wasn’t one of the better items), and the Axe made you immune to horrors and let you chase them around.  Once things got a little more settled, I realised that I could separate each player’s powers further, and they turned into characters with pairs of abilities.

  • The design objective behind the characters was to create distinct roles that gave noticeably different play experiences, while keeping the shared actions useful enough that you didn’t need to include any specific character in your team.  Any set of characters is playable, though I think you’ll stand a slightly better chance if you have a balance between “explorers” (rapidly reveal tiles, fast backtracking), “shapers” (connect tunnels and keep them clear for others) and “protectors” (keep everyone alive).  I had these archetypes in mind throughout the design process, and most characters overlap at least two of them to varying degrees (e.g. Medic is an Explorer/Protector, Geologist is an Explorer/Shaper).

  • The Medic was probably the first character to be designed, and was definitely the first to reach his final form, where he’s happily sat for the last eight months of development.  Initially, the Bandage action had a maximum number of uses, but this was dropped for being annoying to track and for not really impacting gameplay (you heal someone when you need to, and there’s no real need to hold back on this).  His second ability was initially the Leader’s Direct action, but it quickly became clear that it was conflicting with his primary purpose (healing requires you to move to patients, directing results in you staying fairly still – I think I was envisaging you remaining fairly central while directing and then running out to people who needed it?).  After changing this to Sprint, I never looked back – the abilities synergise perfectly, and make the Medic a capable explorer even when there’s no healing to be done.

  • The Bodyguard was one of the initial batch of characters.  He’s always had five health, and he’s always had a way to interact with horrors (though this started out as the Axe’s immunity and push abilities).  During a simplifying step around halfway through development, the push ability changed to a simple “kill an adjacent horror”.  Later on, once the character concept was finalised, his second ability was tweaked so he could act like a tank and protect cavers on his tile from damage.  This made him a lot more interesting – it’s not as powerful an ability as it might first appear, because if you’re on the same tile as someone else you’re probably not exploring efficiently, but it gives him a secondary role as a ‘healer’ of sorts and is fantastic when rushing the exit or as a ‘safe space’ to regroup to.

  • The Engineer was created around the Explosives item, as early testing showed this to be one of the most fun and useful roles.  The concept remained the same, though it slowly got more and more nerfed (cost, cave-in) as its power became clear – being able to explore knowing you can connect two passageways and shortcut half the map is very strong.  Her second ability went through a number of different (weak) options, before settling on something that gives her some protection and synergises well with her own first ability.  The reason it’s not full cave-in damage immunity is because the more ‘safe’ tiles there are for a caver, the less tense the game becomes as you can nearly always end your turn in safety.  (At one point this also protected all others on your tile from cave-ins as well, but it turned out this was completely busted)

  • The Geologist was the first dedicated explorer character to be designed.  Initially, he looked at the top two tiles of the stack, placed one, and put the other back on top; this gave him similar control over shaping the environment, but had the weird effect of letting the caver on his left gain some info about the next tile they were going to place.  After going through far too many iterations, I settled on the current “one tile in reserve” mechanic we have now.  His second ability is deliberately weak to compensate for his powerful first ability, though saving an action point here and there does have its uses.


    Early Geologist prototype

  • The Leader appeared once it was clear that the Medic wanted to run around rather than yell orders at people.  Spreading action points around is fun and powerful, but it was important to limit the Direct ability to one use per turn as I didn’t want the Leader to just sit there doing nothing while doubling someone else.  Her second ability synergises with her first ability by essentially giving her more action points over the course of the game (via exerting), so she doesn’t lag behind as much even when giving orders.

  • The Climber took the Rope item, with the original intent being that having a Climber on your team would make exploring the cave easier as you could now cross ledges.  When the rules changed so that anyone could place ropes (badly), the Climber’s second ability had to be buffed to compensate.  This went back and forth between being a protective ability and an exploration ability, before ending up on the current easier movement buff.  She’s now a great explorer (due to fast ropes and easier backtracking), and a competent rescuer (as she can quickly get to unconscious cavers).

  • The Scout appeared because I wanted another dedicated ‘explorer’ class to complement the Geologist.  A lot of early versions were completely overpowered (a free explore action every turn, anyone?), but she eventually settled on an ability where she could ‘burn’ dead-end tiles so she’d never get blocked.  Later on, I realised I could simplify the ability while reducing the power variance and making it more interesting by allowing her to burn any tile, but at most three times per game (this was a proud moment for me).  Her Stealth ability was designed to make her more resilient at running off on her own (though this is still not a great idea…).

  • The Diver was the last character to be designed.  For most of development, he was actually the Chaplain, who was created as a ranged way to deal with horrors (first by moving them around, then by destroying them from range).  He also had a cool one-shot ability that let you ignore a hazard card.  As the backstory for the game was finalised, the Chaplain concept felt more and more out of place, so he was switched for the less supernatural Diver.  The Diver fulfils a similar role to the Climber – he’s good at exploration and rescue, is a little more resilient, but a little less good at helping those near him.  His long range ‘teleport’ ability can be really powerful if played correctly, as it gives you more time exploring before you need to regroup, and is a great way to jump to unconscious cavers halfway across the board.

    Prototype Chaplain


Hopefully you found this post interesting!  Or maybe you will find it interesting but only after you’ve played the game.  Or maybe you hated it intensely but read it anyway out of spite.  Live your life however you want, I guess.

In my next and final post, I’ll be back to a more article like structure, discussing the three Sub Terra expansions that we’re going to be producing for the Kickstarter!  Hopefully I can persuade you that they’ve had just as much care and attention spent on them as the core game, and are just as worthy of your time and shelf-space.

See you then, and have a great new year!

Tim Pinder

Tim Pinder

Game Designer, Sub Terra

Indie gamedev robot. All opinions, fatalities and apocalypses are the responsibility of my creators, who should have worked harder on the control problem