Okay, I think I've got it.
When performing an action, there are four relevant modifiers. There's your talent or strength at the task, your opponent's difficulty or skill, the complexity of you and your opponent's goals, and any additional wagers you or your opponent make.
You and your opponent's power is determined by your stats and any additional modifiers; difficulty is determined by the general conditions of your goal or operation. For instance, performing athletics on a rain-slicked surface would have a higher power than if the surface were dry.
The complexity of a goal is the number of successes required to accomplish it. 1S are simple, 2S are moderate, and 3S are difficult goals. Simple goals are usually of the "quickly do this one thing" variety, such as assembling a simple item or tapping an enemy in combat. Moderate goals are usually reasonable actions, such as crafting a useful item or attacking an enemy in combat. Difficult actions are usually complex actions with a noteworthy goal, such as crafting a powerful item or hitting an enemy exactly where you mean to.
Wagers are additional success/failure sinks built into the action, and represent additional features of the action you intend to add. Positive wagers have a success cost, usually in line with normal action costs, and therefore increase the odds the action will accumulate additional failures in pursuing this secondary goal. Negative wagers have a failure sink, usually in line with normal action costs, and therefore absorb failures in a predictable manner, but trigger even if they do not receive their full allotment of failures.
As an example, let's run Googles the Octopus through two different actions. The first will be easy and sensible. The second will be designed by PCs.
For his first action, Googles attempts to pilot a ship to the surface of a lake below, scoop up some water, and fly off. The lake is calm, there are no severe winds, and so on, so the action is a base difficulty of 2; that is, it will be as though Googles is opposing a Strength 2 opponent. The action, "scoop up some water on the move" is also pretty trivial, so that'll be a simple action; that is, it will require 1 Success to complete. Finally, there are no wagers attached, so this will be a simple 4v2 roll until Googles succeeds once. Any failures accumulated prior to that point will go towards unfortunate complications, like the bucket coming loose or the ship hitting the surface of the water.
For his second action, Googles attempts to get a good sonar reading on the seafloor beneath an ocean currently in mid-tempest. Since the ocean is very clearly angry, this action has a base difficulty of 6; that is, it will be as though Googles is opposing a Strength 6 opponent. The action, "hold still long enough to get good sensor readings at the base of the tidal waves rolling over you while resisting hurricane-force winds" is utterly unreasonable, and will thus be difficult; that is, it will require 3, maybe even 4 successes.
Not knowing when to quit while he's ahead, Googles also has some additional wagers to add: First, he'd like to get chemical readouts while he's here. That's fairly reasonable, so maybe it just adds an extra success to the required total. Second, he'd like to avoid getting wet. Why an octopus is worried about his ship getting wet above an ocean in a storm is anyone's guess, but alright, he can try to avoid getting wet for another 4-5 successes. Perhaps dimly aware that he's reaching a bit, Googles throws in a negative wager as well: he's okay with straining the engines, perhaps requiring additional maintenance or repair later. That should throw in a good 2-Failure sink.
So we have a 4v6 roll, with Googles needing around 8 Successes at a minimum. Statistically that's going to give him 12 failures to go along with it, but he's got a 2-point sink, so that should only work out to 10 Failures total. That's still probably more than enough to make Part 2 of this mission take place at the bottom of the sea, so hopefully Googles' aversion to water only applies to his now-drowned craft.
Note that theoretically he could have ameliorated that by throwing in a bunch more negative wagers, but that would have channeled the failure more than prevented it; he's not going to get out of a 10-point deficit with "I'm fine with a bumpy ride" or "the coffee can suck for a bit." He might have been able to bash the everloving shit out of his ship to keep it airborne, though, which could have been worth it.