The hot tech in RPG skill systems these days seems to be a cluster of concepts related to “failing forward”, ensuring that the outcome of a failed skill check is interesting so that play can continue in one form or another despite flubbing a roll.
This goes along with other concepts like only asking for a roll if both outcomes are interesting, giving mixed outcomes on skill checks so that a failure is just success at a cost (and the player has some authorial control over what their failure looks like), explicit stake setting, and so forth.
But let’s focus on that failing forward thing for now. In D&D, some skill checks have always been fail-forward: if you fail at moving silently, you get detected, and play that out. If there’s nothing to detect you, then the failure didn’t do anything but ratchet up tension, and that was in the service of the story too: that seems fine. Failure doesn’t stop the story, it changes it. Into a story about the inside of some guardian Beastie’s gullet.
Some skill checks are not (exactly) fail forward, though: if you don’t pick the lock or climb the wall or bust down the door or whatever, you’re still in status quo ante and the die roll didn’t mean anything but a wasted 6 seconds. Out of combat, even the threat of a fall doesn’t mean much, since while it threatens a fall from great height, playing out each of the segmented climbing steps is a giant pain in the butt and most DMs I’ve seen let a single check carry for the whole height.
Yes, you can be in a situation where you need to get through the door right now, but if you aren’t, what are we to do?
3/4e kind of solved this with take 10/20 rules, statistical arguments about capabilities over time: if you can make the check 20 times, assume you roll a 20 on the die and go with that.
I reject this concept in place of the simple “how long do you want to spend?” system. It has die rolls that natter without degenerating to the DM deciding on a binary “you can/not do this depending on whether the DC is 20+your modifier or not”.
Here’s how it works.
In round-by-round combat, you can only spend 6 seconds and accomplish things you can, y’know, do in six seconds. These rules don’t really apply. If you fail on a thing, you can try again next round, and the round thereafter, and so on.
Out of combat, the player says what they’re doing (“I’m searching the room!”, “I’m picking the lock!”, ” I’m rigging up climbing gear and climbing the wall!” and so forth). The DM sets an expected period and a difficulty class for the task. The DM should inform the player of the period, and also hint at the difficulty (though that may be influenced by hidden factors).
Expected period: example tasks.
1 round: busting down a door, manipulating a device, listening at a door, climbing 30 feet
1 minute: searching a 10×10 section of corridor, picking a lock, climbing 300 feet
10 minutes: searching a 30×30 room, skimming a book for clues, building a barricade, climbing 3000 feet
An hour and a half: searching a 100×100 room, reading through a book thoroughly, hunting for game, climbing 30,000 feet (seriously?!)
Ask how long the players are willing to spend on this pursuit.
Adjust the narration of how long the task actually took by the amount by which the player’s check beat the difficulty class (or failed to do so!). If the players are unwilling to spend the amount if time this would indicate in order to succeed, they spend as much time as they were willing to spend, get whatever proportionally success seems reasonable, but fail overall. They may adjust their answer upwards now if they want to just keep trying.
Success can be faster!
If the roll beats the DC, the time the task took was pretty much the expected period.
If it beats the DC by 5+, the task is accomplished in 2/3 of the expected period (6* the next shortest period).
By 10+, 1/3 of the expected period (3* the next shortest period).
By 20+, 1/10 of the expected period (simply use the next shortest period).
Consider a bonus action 1/3 of an action, and “not an action” 1/10th of one.
On the other hand, failure can be slower.
If the roll misses the DC, the player may spend longer to increase their chances of success.
They may spend half the next largest period for +2 on their roll, the entire value of the next largest period for +5, or (at the DMs option) two periods up for +10.
However, if what they are attempting has hazards for failure (such as climbing under hazardous circumstances), those are applied before the character has the potential to spend longer to succeed: they fail.
Example of play:
Cwinn the dwarf, armed with a crowbar, wants to break through a stout wooden door in the Dungeon of Hathory. Cwinn’s player, Chris, declares his intention to break down the door and that he’s not leaving ’till it’s broken. The DM sets the DC at 10, and Chris rolls for it — with advantage from the crowbar. The total check value is 17, handily more than 10. Since this isn’t more than 10 points higher, but is more than 5 points higher, the DM determines that it takes 2/3 of the next shortest period. Since that base period was 1 round, it’s not enough to matter here: Cwinn busts open the door.
However, if we consider Lellin, Cwinn’s wizardly companion, separated from the rest of the group. Louis, Lellin’s player, makes a similar attempt. With Lellin’s total of 4 (tired, hungry, lacking crowbar and not up to the task at the best of times!), Lellin can only bust down the door at the cost of 10 minutes effort. Since that’s two orders of magnitude more expensive, the DM may also rule that Lellin is unable to bypass the door until the situation changes.
Why would you do these things?
Because the amount of time characters spend on a task reveals how much the player wants to risk failure and time is always secretly a resource.
If the players characters have to climb a cliff, it’s not particularly difficult, but the DM still wants to roll for random encounters or whatever, the interesting roll isn’t the climb check: it’s the random encounter roll. These rules give the DM a quick way to approximate how long a task will take and then move on to whatever’s interesting, without saying “you fail on your sixth attempt, but may retry for a seventh in 6 seconds!”
That kind of narration works well when the environment is chaotic, changing from round to round, but most skill check scenarios are static, so the only interesting “partial success” outcomes relate to resources expended.
TL;DR: it strikes the balance between the DM making a snap judgment without any guidance and boring the whole table with a thermodynamics simulator.