Know thyself. A fantastic example.
I agree that this IS real work. At least for me, I believe that to do this well, there has to be at least something about this that is fascination to fun about it. If this were work, in the sense of something that I was making myself do, then I wouldn't be doing it well.
NOT-ADVICE; just way that I see things. Worth working out your own examples to see if you agree.
The way I see rolling in general, and spreads more particularly... when a position is close to the money or in the money, and the purpose of the roll being considered is to buy time for the share price to turn around and the position to move back OTM (vs, a convenience roll - confirm the profit for this week, while opening the position for next week), then:
There is a balance between rolling too soon, and rolling too late that we're looking for. Of course if the crystal ball was always clear then it'd be easy, but the crystal ball is cloudy.
The factors that affect the roll - everything else being equal we'd like to roll with the time value as close to 0 as possible. The rationale is that on a roll we are buying out the time value remaining in the old position so we can sell the time value in the new position. The less time value in the old position, the less there is to buy out. Or a different way of thinking about the same thing - our earnings are in the time value; the less of there is remaining, the more that we've earned. That mostly means that the closer to expiration, the better.
The other factor is just how deeply ITM one goes. Generally speaking the quality of the available roll decreases as the share price approaches the short strike and continues getting worse as you go further and further ITM. It seems like the very best rolls are ATM or slightly OTM, but are still good enough that waiting a few extra days while going slightly ITM is a better tradeoff (time value melting away).
Too soon and you're buying out more time value than you want to (lower quality position to roll into), and too late may see you get too deeply ITM, thereby lowering the quality of the position you can roll to.
My own bias in this tradeoff is to roll early and for maximum strike improvement (subject to a net credit). There are two reasons for this:
1) The maximum strike improvement is going to get me the best protection if THIS move is a biggie. Most of them aren't, but if this is the one, then I'd like to have given myself the very best chance of staying ahead of it, so the shares can reverse.
2) I view the roll as buying time and that means I'll have a week (or more) of effectively no earnings. Earnings going to 0 for a week or more is way better than realizing losses. And since I'm in this for income, and I've got enough buffer from the weeks that win, one of the uses for those winning weeks is to allow for weeks with no income. Oh - and realized losses are bad, really bad, for income
Rolling early if the shares aren't gapping up and down will bias towards still OTM. If I'm close to expiration and OTM, then I might split the difference on the roll - find out my max strike improvement and then come back 1 strike so I get a bit of income that week while still providing lots of coverage if this is a BIG move against me.
From what I've seen high IV provides significantly better rolling choices, while also being an environment in which more rolls are needed (can't get something for nothing).
I learned at the beginning of the year when IV was high that I could get an effective roll while $80 ITM. Shares around 720 and strike around 800 on some puts, so 10% ITM at that point, where effective = credit + strike improvement available, whether I take the strike improvement or not. A month later with IV down, I no longer had an effective roll when $40 or $50 ITM. I don't have a formula here - I just regularly setup candidate rolls when its possible that I might be needing them soonish so I can see just what's available.
I learned that there are situations where a 1 week roll is barely or not effective, while the 2 week roll IS effective. Or 1 week vs 3 or 4 weeks. I personally choose to not roll out to more than 4 weeks, at least without a lot of thought about it. I've rolled out to 1 and 2 years before and really didn't like that dynamic. Once I'm up to 3-5 weeks to expiration then it might be time to take assignment (or the loss) and move on.
I've seen situations where the 2-4 week roll is much more than 2-4 1 week rolls. And heck - if you're deep enough in the doodoo, then I've rolled for as much as a month just to avoid the weekly work (and get a better credit).
When really deeply ITM on naked puts and calls, the time value will approach 0 much sooner than the day before expiration. I keep an eye on that and have been in situations where I was rolling 1 week ahead of expiration to avoid early assignment (a good time to be considering monthly rolls). That doesn't guarantee no early assignment - I just don't try to push time value really, really low. Probably around .30 to .50 is as close as I've gone (guessing at the number - it is at least directionally accurate).
And last that I can think of right now - spreads have different roll characteristics from plain short options. These are best seen in a really tight spread, but the dynamic applies to some degree to any spread. To the degree that a spread behaves like a short put|call, then the rolls that maintain the spread size should be expected to behave similarly or the same. From what I've seen this relationship holds pretty well through 1/4th of the way into the spread. So $50 on a $200 wide spread.
The spread stops rolling for a credit or strike improvement at the mid point.
Rolls past the midpoint of the spread are a net worsening of the position, either by paying a debit to maintain the strikes, worsening the strikes, and/or bringing more capital into the trade.
And somewhere in there - probably the midpoint and lower - adding more time to the roll doesn't change the quality of the roll that is available. I think that what's going on here is that the % change in time value for the long and short legs are now changing roughly the same, so whether you add 1 week or 4, the time value relationship between the two changes about the same, thus little or no extra value in more time. Maybe even a worse roll with more time - I haven't tested this though.