Speed trumps position until speeds plateau. Then position usually trumps speed. This from a more amateur racer where it isn't flat out for the last hour before the sprint. For pro races, since most sprinters are virtually the same in top speed, it's about position and how fresh you are (aka how much of your top speed you can hit).
There was an informal enormous ride back in the day, the SUNY Purchase Tuesday Night Sprints. A lot of people would show up, realistically 60-80 on a regular big day, over 100 on a huge day (the ride
patrons would break up the group into two). A lap was 2 miles, took about 7-8 minutes, and we'd be there 2-2.5 hours. I'd typically contest 10-12 sprints, 15-16 on a really good night (pretty much every sprint). 1 mile neutral, 1 mile anything goes. I usually went with friends/teammates and we'd do leadouts so guys would go as soon as we passed the bus stop (beginning of the 1 mile to the line). Since it started on a proper shallow downhill we'd rapidly be up to 40 mph, slowing a bit as the road undulated, then sprinting on a false flat downhill.
I learned there were four aspects to the sprint.
First, top speed. With better top speed, a rider could start his sprint from further back, aka "shelter". In many sprints top speed was enough to win a sprint. However, the fastest 3-6 riders were evenly matched in top speed, and for those sprints top speed no longer counted as much. A key figure to keep in mind - 1 mph is 1.5 feet per second, so a 1 mph difference in speed in a 10 second sprint is 15 feet. 0.1 mph difference is 1.5 feet. When sprints are won/lost by just a few feet, the top speeds are virtually identical, almost imperceptibly different.
Winning by a bike throw is crazy close speeds, like winning by 2" means 0.01 mph difference in speed. This means a good bike throw, gaining about a foot, is worth not quite 0.1 mph in a sprint. (Similar top speeds makes sense - wind resistance increases exponentially so the fastest riders tend to go to a similar top speed - this is basically the case with the Kilo time trial on the track, where top speed differences are minor so now the emphasis has been on getting up to speed, aka the jump, which leads me to...).
Second, the jump. If two riders have similar top speed, and, again, this is somewhat common, then accelerating to that speed quicker is critical. Once up to speed, if both riders max out within, say, 0.1 mph of one another, you'll be pretty evenly matched. If one rider got to speed quicker, and they both jumped basically at the same time, the better jump sprinter will win.
Corollary to the jump - you have to be able to optimize your cadence, and in a road sprint that means you need to shift during your sprint (unless you launch at pro speeds, then you're already in the 11T). At SUNY Purchase I had a secret weapon for a while - a right side bar end shifter. This enabled me to jump in a lower gear then shift up through the gears as my speed climbed. Most racers used downtube shifters so they were stuck in whatever gear they jumped in. I could beat one such rider, a far superior racer and better sprinter, who we all know as Eric Minn (of Zwift). I'd jump in a lower gear than him to match his phenomenal jump, then shift through the gears to hit a better top speed. When STI came out and he got them, I pretty much never beat him again. He could out jump me no matter what and I couldn't pass him because we had similar top speeds. The only way I could beat him was if I jumped super early (so his jump didn't get him past me) or if he got boxed in, but generally he was too crafty and smart for either of those.
Third, position. The above two are important but only if you have decent position going into the sprint. Do you have to be 2nd or 3rd? It depends on wind.
* If it's a headwind, being the first to jump is pretty risky - everything important will be compromised. Your jump suffers, your top speed suffers, and there's a proportionately larger draft benefit for those behind you. I've seen decent sprinters do well from 15-20 riders back, if not further. At the same time, if one rider jumps super hard and the rider on his wheel lets a gap go (maybe a non-sprinter, even a teammate), it's very hard to claw back that gap. Position works both ways in a headwind, but generally if it's all sprinters up front you can wait in a headwind sprint. If it's very narrow then you need to be far enough front, but if it's a wide road you can be really far back. At SUNY if it was a headwind I'd either jump really, really, really late (if I was 2nd wheel behind a leadout guy) or I'd "lose" my leadout's wheel, let him keep the group together, and jump from 15 or so back.
* If it's a crosswind, and you understand how to use a crosswind, then position is absolutely critical. Be 2nd or 3rd wheel, jump to the sheltered curb side, and hold that curb. Anyone trying to pass you will face the same wind as you, and just as they pass you they'll be sheltering you all of a sudden, giving you a little respite. I saw one rider, a non-sprinter but with good 1 min power, lead out a current multi Masters National Champ (track sprint, crit, some others). I turned away because I figured my non-sprinting friend would get annihilated. He ended up winning the sprint, to my absolute disbelief. Crosswind, and the sprinter just could not get around the non-sprinter (Tour of Nutley, Cat 3s).
* If it's a tailwind then all bets are off. Biggest gear, biggest wheels, and if you have them then lead out early. Only those with similar equipment will be nearby. Top speeds will be unnaturally high. The rider with the compact crankset or the 12T small cog will be seriously disadvantaged. Speed becomes key, and gearing and wheel selection is critical. You have to have a gear you can push even at ludicrous speeds, and too low of a gear will blow you up quicker ("contraction fatigue"). Big wheels sail really well in cross-tailwinds, and I've picked up a lot of speed sprinting in those situations. SUNY Purchase often had a cross-tailwind sprint - combined with a very slight maybe 1% downhill final 200m the fastest sprints usually topped out at 46 mph coming off 35-38+ mph leadouts (take about 8 mph off for the slower headwind days). On those days my leadout guy/s knew to go flat out (i.e. they're actually sprinting to the point where they pull off) and I'd jump really early, banking on holding top speed long enough to get to the line. Those sprints were the most fun.
Fourth, the throw (and other techniques). If you don't know how to throw your bike properly, you'll give up places and possibly wins. In the link above there's a picture of a rock solid pro that almost won a stage, but because he doesn't throw his bike properly, he got 2nd. Later he says in an interview that if he'd won the stage it would have changed his career. He did a track type throw, which basically involves extending your arms. But a road sprint throw involves putting your butt way behind the seat, thrusting the bike forward. Peter Sagan is really good at this. When Taylor Phinney got tied for bronze in the world RR (U23 RR,
picture in this article), it was because the Canadian guy really threw his bike well. If Taylor had throw his bike better ("properly") he'd have had the bronze to himself.
Other techniques optimize the jump, the sprint, etc (like shifting while sprinting, or rocking the bike, or sprinting on the drops) but those relate more to the first 2 things, speed and jump. The bike throw is unique in that it gets you closer to the line immediately with very little effort.