Kavan

You can say that you can't generate optimum result if you don't know the fight duration. Another way to look at it is if you know what the optimum is under different durations you are better prepared to handle the reality. One thing one will notice is that optimum sequence does not change drastically when making small changes to fight duration. Most differences will occur as result of additional cooldowns becoming available.

I'll show an example analysis for my specific case using sequence reconstruction in Rawr. It is good to know at least an approximate fight duration. If you don't know whether it is going to be 3 minutes or 10 minutes then all bets are off. Let's say we're expecting a 5 minute fight. Then it is reasonable to be prepared for anything between 4-6 minutes. I'll show the sequence reconstructions as computed by Rawr and give an interpretation of them. In some cases you'll see some suboptimal suggestions which I'm still working on improving, but it's good enough to get a general sense.

Starting with 4 minutes the optimum sequence is:

So we start with Heroism burst at the start. One possible consideration when running this in real situation is threat which is not currently considered in the reconstruction. Goal here is to burn mana as fast as possible to get the pot/gem cooldowns going. Further AB burst until we have enough room for Evocation and a bit more to create a sizeable mana gap in anticipation of switch to dpm cycle which is mana negative. Main purpose of the switch to dpm cycle at this point is to prepare for the next AP+IV phase. After enough mana is recovered (or more importantly not burned with AB spam) it switches back to AB spam, and ends with AP+IV burst. You can see here a clear bug at the end as it runs out of mana during AP+IV, while it could move the final AB dpm cycle in front of it (this is a bug in the reconstruction, the cycle solution is guaranteed to have enough mana, it's just a matter of arranging the different cycles in temporal relationship).

Next if we look at 4:30 fight we get:

We have a similar structure. Starting with AP+IV burst, evocation, brief switches to AB dpm cycle to conserve mana and end with heroism burst. One thing that changed is the alignment of trinket. 4:30 allows for 3 uses of a 2 min trinket, so we see a change there. If we are planning for 5 minutes then in the 4:00 scenario we probably wouldn't realize in time that fight will be shorter than expected and we would use the trinket whenever it was ready. So that would be one source of suboptimal sequence as a result of unexpected change in fight duration. Other than that the structure is more or less the same. The AP+IV and Heroism bursts are switched, but that choice is more or less arbitrary.

Going on to 5 minute fight:

Again basically the same structure. For 5:30 we have:

First we notice the suboptimal placement of AB dpm cycle, more of it should be in front of Heroism. Another thing to notice is that this sequence is not really compatible with the plan we should use for 5 minute fight. The suggested order of major cooldowns is IV+Hex=>Heroism+AP+Hex=>IV+AP+Hex. We could rearrange the sequence into AP+IV+Hex=>IV+Hex=>Heroism+AP+Hex. That way we have a more similar structure, but we also see that longer fight duration allows a better variant of AP+IV desyncing and using both with the trinket. In 5 minute fight the IV is not ready yet when we have to use the trinket. Longer duration here allows us to wait with the trinket until IV is ready. Similar to the previous cases it is good to be aware of these nuances. As the fight goes on and we get better estimation of fight duration we can still change our plan. With this rearrangement the plans are more or less the same until the trinket is ready for the second time. At that point we have to make a decision on whether to delay with trinket to pair it with IV or just use it whenever it is ready. In my view it is better to play safe and assume shorter fight and use the trinket as soon as possible. We're balancing between cooldown stacking and the potential option that we won't even be able to use the cooldown if fight ends sooner than expected.

For end let's look at the 6 minute fight:

We see that this is more or less the same situation we had at 5:30 as no new cooldowns become available in this time frame.

Looking at all the examples we see that the big decisions concern cooldown stacking. Should I wait with the cooldown to stack it or use it now when it's ready. Other than this everything stays the same. Start with high mana burst to start the gem/pot cooldowns as soon as possible. After all mana cooldowns are going enter conservation mode anticipating the future bursts. This part is what requires the most experience to do right and has to be performed on the fly taking into consideration your actual external regeneration that is not under your control. In practice it's usually better to err on the side of conserving mana. As arcane it is very easy to burn mana. Even with the highest mana pools you can burn all your mana in about a minute. So until you're almost at the end just make sure that when mana pot/gem is coming ready you have enough room to use them as soon as they are available. After managing your mana becomes second nature then you can start worrying about the other decisions about how to stack cooldowns.

Overall I hope you got the impression that arcane is very flexible in terms of adjusting to uncertain fight duration. The plan remains more or less the same in all situations and as a result you can reestimate the remaining duration of the fight on the fly when you have a better estimate in hand. The key idea to take from this is to use AB spam on all cooldowns if appropriate and use your fillers for mana management in preparation of those bursts.