Maximizing Skill Usage

[nitrogrlie]

What have others found to be the best method for optimizing their characters skills usage whether it be for DPS or for HPS?

I was looking at the various events that are fired during a spell cast (UNIT_SPELLCAST_SENT -> UNIT_SPELLCAST_START -> UNIT_SPELLCAST_SUCCEEDED -> UNIT_SPELLCAST->STOP) and the associated time in milliseconds between these events to see if there is a clear indicator of when I can start cast the second spell, but not really. SPELL_UPDATE_USABLE might work but the problem is that it isn't tied to a specific spell or unit/player so there is no way to know who and what the event is associated too.

Anyways, I guess my real question is, is there a better method for maximizing skill usage other than:

double nextCastTime = currentTime + spellCastTime - networkLag;

My problem with the above is that networkLag is only updated once every 30 seconds, but I think the server allows you to pre-cast a 2nd spell before the cooldown is up officially within the lag threshold. I wonder if you can even go faster than that.

Additionally, I added a check for the event UNIT_SPELLCAST_DELAYED (this occurs when your cast time is pushed back as a result of getting hit while casting) which increments nextCastTime by +0.5 seconds.

Anyone able to offer any suggestions to improve this method?

[DrGonzo]

Save current mana level, cast spell, wait 500ms, loop while checking for mana to decrease, wait another 500ms for good luck. Not the most elaborate, but pretty effective.

[amadmonk]

None of my bots have needed to account for lag in order to still "play" better than any human I've ever grouped with.

You may be over-engineering it

[nitrogrlie]

None of my bots have needed to account for lag in order to still "play" better than any human I've ever grouped with.

You may be over-engineering it

You haven't played with me
You are probably right, but it's so trivial to do, why not?

[lanman92]

Obviously not TOO trivial, as you came to a forum for an answer :P

[amadmonk]

Yeah, the thing is that if you *don't* account for the lag delay (the way I'm doing it today doesn't account for it), checking when to cast is as simple as looking at the cast/channel ID.

If you DO account for it, then spell casting has to become a mini state machine complete with a scheduler. It's at least an order of magnitude more complex, for not a huge amount of real world gain...

[nitrogrlie]

Obviously not TOO trivial, as you came to a forum for an answer :P

No, not at all. I came to the forum to find out if there is a superior method that anyone employs then the one I use which I posted. The implementation isn't difficult, it's knowing what constraints you are working with.

Yeah, the thing is that if you *don't* account for the lag delay (the way I'm doing it today doesn't account for it), checking when to cast is as simple as looking at the cast/channel ID.

If you DO account for it, then spell casting has to become a mini state machine complete with a scheduler. It's at least an order of magnitude more complex, for not a huge amount of real world gain...

But it is really just a single double based of the timestamp and known casttime or global cooldown conditions. I execute a skill/spell and (as long as it increments the global cooldown) I know that I cannot user another skill/spell on the GCD for some predetermined time. This is important to me because in PvP when my bot sees an enemy player heading to an area that's out of LoS from me (projected movement vector based on last few position reads) I can estimate when the player will get there and thus choose between a spell that can hit them before they get there versus one that will be too late cause the cast-time is too big.