Have a time variable and every time you do something (in either category or any other category) subtract from it. The things you don’t get to do don’t get done and if you can’t do enough things or certain necessary things, bad things happen.
Edit: And then just do the two variables as normal single variables.
You don’t need to make a stat for tracking that just track a temporary variable or a permanent one. That tracking practice is easy. If you do so you don’t loose teaching or scholarship knowledge you have you lost the practice on it or gain it. As in your example imaging you have a way higher teaching stat but you expend all time in scholarship… Game would track that and say.
You have a lot of knowledge of x subject however your lately lack of proper practice makes your exam redaction worse and less accurate. So your note is worse.
However in scholarship text would notice that you hard training has make you pass evn if you aren’t so skilled so you have more fatigue or you had to work harder than a skilled one. See no need a opposite stat and games gain deeper rp and replay value.
Here’s a link to a post onthe benefits of not selectable choices based on stats that you might find interesting : Knowing of stat checks
Here’s a post on degrading stats if you like that: Degrading Stats
Also, I’d like to point out that opportunity cost is not well modeled with opposed stats. In fact, when considering what kind of stat you use, you should ask yourself: if I had infinite time to do these things, could I do both? If yes, that means someone could be good at both things and you should not represent the stat with an opposed pair (your teaching/scholarship example). But even if I had infinite time, I couldn’t be both extremely cynical and extremely idealistic.
Edit: Actually, let’s think about opposed stats another way: couldn’t you be bad at both things? If you never studied or exercised, both stats would get worse. if you don’t study but do something else (say, hang out with friends), do you get better at sports?
So what if my opposed ability stats were not based on whether you had the skill, but how you used the skill? That’s then not a change in ability but a change in perspective. It’s like going from painting classically (say, Raphael’s Madonna + Child’s) to painting abstractly (early Matisse?), you could maybe do the other, but your brushwork is too different, you’ve been working the paint differently. But maybe - like grey / middle - you carry on being capable of both but not having experimented to the extreme with either school. you’re capable of both, you’re a master at neither (and in life sometimes being able to switch methods is good) If leave options for grey, could work? And if didn’t defined PC by opposites, and left it so the liar could tell truth etc?
It’s interesting. Before considering grey choices, I think my game had at least 80 clear different “methods” (e.g. balancing, max/min, efficiency with certain stat etc) considering the grey solutions and now it’s gone up to 270. 
without even changing stats! Is that an acceptable amount?
…then again if make more single stats, methods count goes up to 7290 
but then the opportunities to clearly indicate stats through text flavouring, becomes less clear to show and impossible to communicate each method well.
While I would most likely use single-stats for something along the lines of a skill (since, as I’d seen in many criticisms in Empyrean and I’m sure others as well, practicing one thing doesn’t automatically make you bad at another) I find opposing stats very interesting when it comes to a characters perspective and personality.
Say you have kind/cruel and optimistic/pessimistic. Those can be interesting tools for divvying up the narrative into different perspectives as people play different characters. A character who’s more kind than cruel could react one way to a person pleading for help, while a character that’s inherently more cruel than kind would react another. Same with optimistic/pessimistic, if you have someone who plays an overly optimistic character, they might view the world and any tension in the storyline a different way than a character who’s more pessimistic than optimistic.
These, I feel, help with immersion into the story, since they’re not so much challenges for the player more than they are tools for the author.
If you’re worried about it becoming too easy to maximize stats then you just need to make time the determining factor.
Say you’re about to go off on an expedition into an unknown land, you’ve got five days to prepare and each of those preparations take about half a day to do- so in short, you have ten actions. This causes the player to pick and chose what they want to improve on, and throughout this time they’re interacting with the NPCs and little intermediary events happen to keep it from feeling repetitive. Maybe what action you take that day will determine who you run into and talk to, what bits and pieces of other character’s personal stories are revealed to you and stuff like that. All the while, the player is still increasing their stats, and their responses to the NPCs will help you as the author write how this kind of character that responded X-way to Y-conversation would view the world outside.
It still made the player pick and chose what to raise but it wasn’t just a “raise this lower that” situation nor was it simply a stat-training-montage. It was woven into the story and had a long-lasting effect, so even the small choices like what stats to raise matter and added something to the overall narrative.
Personally, I think of stats more as tools for the author than as challenges for the player… but I can’t say much about the best way to use stats since in the game I’m writing I use stats as little as possible (favoring action-outcome-reaction-outcome kind of things, mainly using stats to- as I said above- change how the MC reacts to the world around them).
I don’t like opposite stats due to a… peculiar way of thinking.
For the sake of the argument let’s assume we’re dealing with ‘compassion’ and ‘ruthlessness’ stats here. The way most games work right now, the kinder you are the more points you get into the former but what if its all a ruse from your part to deceive the other party? What if you’re being charitable because you expect something in return, or are trying to get in their good graces so you can take advantage of them later? What’s more is that with opposite stats you can’t really do something like this because then your stats are too “neutral” to be able to pass the stat checks… resulting in faliures left and right and maybe even a game over or a bad ending.
I’m one of those people who believe people can sometimes be paradoxical in nature - be good and bad at the same, kind and ruthless, honest and deceptive. Or maybe I’m just the odd one out which is why I tend to find peculiar ways of thinking or perceiving a problem. That’s a possibility too. 
That’s less a problem with opposed stats as it is with the game assuming reasons for your choices. Instead of offering the choice:
- Save them!
it should be asking you:
- Save them because it’s the right thing to do.
I guess there could be compassionate reasons for not saving them, too (like maybe you’re hurrying to save a far larger group of people, or you know if you tried to help you’d only make it worse…).
But then what kind of changes would those choices incur in opposite stats? It feels like a tug-o-war between the two of them when I do something good with malicious intent or vice-versa…
No malice. You’re not leaving them to die because you hate them or because you enjoy their suffering, but rather because it’s more effort to save them than it’s worth to you. Consequently, this would be a selfish/selfless thing (and helping people for your own good is not selfless).
Of course, one thing you seem to be overlooking is that this would apply to any system of personality or morality stats, whether opposed or not. By the code, opposed stats don’t even really exist. There is no selfless stat; you’d be testing against 100-selfish instead. As long as the “opposed” stats are truly opposites, there should be no “tug-o-war” (as there would be, for example, if “guns” and “blades” were opposed stats).
I guess one thing you could do is have two sets of stats: public and private. The public stats would be based entirely on what you do (and would affect how NPCs treat you), while the private would be based on why you do it (and would affect flavour text, such as what you think of various things). That way you can play a villain who acts as a hero because it’s more useful to him without the game assuming you’re just a hero
I use both in my current WiP, or rather plan to as I haven’t started coding yet.
The opposite stats are personality related and add flavour text as well as a few certain choices that you don’t have to pick but that open up to you. Say, for example, your character is introverted - so they get the choice to become a stuttering mess if they so choose, but they don’t have to act that way. The extroverted opposite gets to let loose more than the introvert, but both can go for a middle ground. Likewise, a selfish or selfless character can act however they want, but the flavour text will be catered to their personality and one or two more extreme options open up that the opposite side doesn’t get. These stats are determined at the start of the game and don’t change a whole lot throughout the story, since they are core beliefs of your character.
Single percent variables are the things that you can influence and that determine your character growth, strength etc. They are the ones used for skill checks and let you fail or be the winner of a given situation. Character flaws that you pick at the start of the game also act as single percent variables because they add flavour text and clearly influence some stats.
I feel like this approach is a good way to make the player character feel unique while not sacrificing player input. Wanna play an introverted character who sucks at talking to others but loves to party? You can play that without any problem. Wanna be a brat who is loud mouthed but helps others grudgingly all the time anyway? There you go.
This is also great for relationships with other characters as they can react to your personality and start out with different friendship stats according to what your character is like. They can compliment your character based on their core traits or pick a fight due to them, making the NPC’s feel like they actually have a different opinion of your character based on some of your traits.
A opposite stats is not opposite really , in code is just one variable. if positive number is good if negative evil. But in truth there is only one . That’s why there is no flexibility or so many options available.
I don’t know why I never quite got it before this part of the conversation.
This all makes me realize that–since opposed stats are not really a pair of stats, just one that looks that way on the stat chart, the real difference between opposed pairs and non-opposed stats has to do with the way they are tested:
Principle 1:
If I have sweet/sassy as an opposed pair, it makes sense, storywise to have successes based on high sweet or high sassy. Or, to put that another way, there are times you will succeed because you have high sassy and times you have low sassy.
Opposed stats reward both high and low.
Principle 2:
A unipolar stat will be unlikely to award a success because it is low. If you have a low baking stat it is difficult to think of a situation where you’d rather have that, than a high baking stat.
Unipolar stats reward high stats, unless something funky is going on.
So dumb/smart is not a good opposed pair, because it is difficult to imagine stat tests that would reward dumb.
Methodical/quick is a better one, because you can imagine those tests on both ends.
Now pick away at my conjecture, so we can refine this.
It depends on your stat checks coded into the story - if you code it, a low uni-stat score “failure” can mean success.
I still think the consensus view on the opposite pair stats is missing the idea behind the mechanic.
I totally agree. But I would guess that that is by far the exception.
I have stats that reward being low or give a very different and funny experience. Like intelligence if you play low intelligence flavor text would be different and no worse or better. Stats could do whatever you want are so flexible. you could make that a variable set true when a stat reach a certain level and set false at other.Like if health drops 24 you caught a flu and up 55 you are good again. You could do so many stuff.
It really is a matter of design - for example: a stat check could reward dumb if you are putting on an act as a spy and want everyone to view you as the “dumb” character…
I think the opposite pairs and the uni-directional stats in a totally different framework then everything said here. Without going into details:
1: Unidirectional stats act as thermometers showing the temperature
2: Opposite pair stats act as a pair of scales indicating which has more weight
both are “pictures in time” and neither is necessarily in conflict… it all depends on your implimentation of game design.
If I were playing a game that had dumb/smart as opposed variables, and I wanted to play dumb, but I was too smart to do so, that would annoy me. 
Aside from that, I’m not arguing that you can’t do it. And definitely flavor text will be greatly influenced by low stats. But I’m musing about stat tests in particular, whether the big difference between the two types of stats is that one is more likely to reward a low stat, while the other isn’t.
Because, as @poison_mara and @ParrotWatcher noted, there no real difference between the two types of stats, aside from how they are presented on the stat screen.
Maybe something to keep in mind for your new game, eh?
There are a few things I can think of that “reward” dumb. The most interesting was in Fallout New Vegas, where dumb characters can get Arcade as a companion more easily then others (because he feels sorry for them, so it is out of pity). The other one would be people who are perceived to be dumb getting away with some (criminal) shit normal and smart people wouldn’t. Of course this “benefit” goes mostly to those who are perceived as dumb or dim-witted by society and the people around them and sometimes the greatest benefit is for the genuinely smart character to act dumb. One situation I can think of would be a smart enough royal bastard acting dumb enough to dupe his royal parent and sibling into thinking he would not be any sort of threat to them.
Of course stat wise this is best handled by either having a simple check for low intelligence or else for the appropriate skills at deception (charisma + cunning would seem like the natural default here).
Are you implying that you think I wouldn’t? 