It doesn't take much time to cut off a note-on.

mixmkr

Anderton
 
I don't think there's any way you NEED a measure, you probably only need a millisecond or two (even attack times of 1 ms are audible). However, by making it a measure, it's easier to move things around.

That statement seems to conflict with

Granted, 99% of the time the sequencer will start up okay.

 
No, they don't conflict. Most of the time things will start up okay but if not, a few milliseconds of space would probably be enough to let the computer settle down.
 

With Battery 3, on every playback, the first beat on the current now-time always sounds weird. I always move the now-time to a earlier point if I want to hear it play properly. I suspect that it's a buffering issue.
 

This monkey started with playing live gigs and an Atari running the sequences. To this day I don't think Sequencers have improved much. 
 We ALWAYS left the first measure before the music started. We put all the PG and CC stuff after the first 1/4 note in this first measure. If you put a PG event at zero it was a train wreck of sounds. 
And don't forget, we are talking midi tracks, not audio. 
Seems there is no problem with audio starting at zero. 
 
The question to the calkwalk people would be: 
Do all the events that will happen upon hitting play get loaded into a memory buffer while the transport is sitting at the start? 
And then the other question is, how quickly do soft synths respond to the data? It is my understanding that those infamous midi buffers we need to keep ramped up to around 1,000 need a few nano seconds to dump their load. The software does therefore possibly need time to "think" to keep the events happening in time. 

I believe you are right in that at that instant you hit play a few things have to go down before the sequencer can start playing back, midi and audio. And having midi note data right on the first measure can sometimes get corrupted or not play etc. If there is any chance of that then starting at measure 2 or 3 is a good option as it simply eliminates any possibility of anything going wrong before the first midi data is sent or audio is heard.
 
It certainly seems to have a much stronger effect on midi data in general. eg if send a large chunk of controller data at the start of measure 1 then some notes may be overlooked in the process. At measure 3 all the midi controller data will be sent followed by the note on and you will hear it. (even it is slightly late)
 
Many will work much more so with audio and therefore can have no issues with this but those who are using a lot of either external or internal VST midi data then this can be an issue.
 
Turning internal VST midi data into audio early is probably another way around it. But it should not pose a problem starting on measure 3. It ensures many things can happen comfortably before the music starts. If you are working with charts etc you just have to make sure the measures are correctly numbered on the charts to match the sequencer.
 
Another good reason to start on say measure 3 is a situation where you might want to advance a midi track and make it play slightly early to compensate for any latency in that particular synth may impose. (VST or hardware, they all need some time to make a sound) Not only for latency reasons but you might just want a track to sound a little more urgent by advancing it against the grid timing. I fine tune midi track timing all the time. The music can sound a lot better just for it. If you start an arrangement on bar 1 you will never hear those first notes. But by starting on measure 3 especially if you cue a bar prior you will always hear that first note if it is there. Advanced or not.
 

I almost always have MIDI drums, typically Toontrack, and I don't know of any of their patterns (for the most part) that don't have a *note on* on the very first hit of the measure.  I use their patterns for countoffs, and of course songs and I have never lost a note.  Same with Z3Ta2, Rapture, or DimPro...the other synths I use quite a bit.  Additionally, I will almost start a song in MIDI to create and then replace MIDI tracks if needed....so I use MIDI quite a bit and Sonar's MIDI was basically the main reason I migrated to Cakewalk.
 But...I'm the lucky guy in the studio!  :-D
 
Again... Maybe need to start a new thread to see what Cake says.  I value their opinion most on this question and am also ready to be corrected....   but then again...I'm the lucky one!!

mixmkr
I almost always have MIDI drums, typically Toontrack, and I don't know of any of their patterns (for the most part) that don't have a *note on* on the very first hit of the measure.  I use their patterns for countoffs, and of course songs and I have never lost a note.  Same with Z3Ta2, Rapture, or DimPro...the other synths I use quite a bit.  Additionally, I will almost start a song in MIDI to create and then replace MIDI tracks if needed....so I use MIDI quite a bit and Sonar's MIDI was basically the main reason I migrated to Cakewalk.
 But...I'm the lucky guy in the studio!  :-D
 
Again... Maybe need to start a new thread to see what Cake says.  I value their opinion most on this question and am also ready to be corrected....   but then again...I'm the lucky one!!

Some of this may also be a holdover from the days when sequencers were driving external hardware. If you had program change commands and initial controllers on a lot of tracks, as well as some note-on messages and maybe even some poly aftertouch, I think a Mac Plus would not only have choked on the data, but might have needed a Heimlich Maneuver to recover.

Its very important question that was answered by the experience forum members, but I think leaving a measure will give the computer enough time to gather its resources so the project can start smoothly on time. There is no hard fast rule but its a standard practice that has stood the test of times. Either way all the recommendations given  are solid in my humble opinion.