Valraven rate is estimated at 1 in 4, stated to be capped at +/-1 valraven at any given time. Thus far, all indications are that Valraven rate is the same between bounty and raid/invasion (with invasions just being hidden and giving the sigils on a win). The more sigils you start with, the more likely it is you’ll complete twice that many battles by the end of the event if you get sigils from every possible Valraven.
What I’m uncertain of is the math they use to ensure this plus one/minus one limitation, and it may skew things slightly.
I just did some tests in raid, my alt completed 10 battles and has seen 4 valravens at:
1 - expected 0-1 ravens, have 1
3 - expected 0-1 ravens, have 1 + 1
8 - expected 2 ravens, have 2 + 1
10 - expected 2-3 ravens, have 3 + 1
If this pattern repeats for the -1 (allowing you to get 0 ravens until the average would have you expect 2 or more), you could go be at battle 8 before you see your first raven. If you remained at -1, you should be guaranteed another raven by 12, and 16 etc. This should also let you catch up and get multiple consecutives, RNG willing, and still be able to go to +1 overall, not just for the local stretch, but again, I’m not privvy to the exact implementation. But this specific implementation would also lower the overall average expectation of battles played by roughly 1 battle across the board (being -1 when you finish your last sigil would hurt you slightly more than being +1 would help you, so the average slightly trends to the former) and have the minimum and maximum amount of battles for any given pool of perfect play be [initial sigils * 2] +/- 6 rather than [initial sigils * 2] +/- 2, which would be the case if raven appear was clamped at within 1 of expected, rather than +/-1.
You need x18 avg multiplier across all battles, median or better Valraven RNG, kill every raven, and win every battle. If you do x24-30 during the earlier rounds (3-4 mythic bounty troops, including the captain), you have just a bit of breathing room on RNG and maybe losing a battle or missing a raven. But if raven RNG really allows +/-6 battles over any given extended stretch, as it now seems like may be the case, you could end up needing an average of x20 to finish all reward tiers with a t6 if you are far on the bad side (remaining -1 ravens until the end), which would require 3 bounty troops for at least part of your repetition cycle. Still possible, but much more annoying, especially considering anyone on the other could coast with 2 bounty troops.
tl;dr: Valraven RNG may allow for a bit more range in minimum and maximum expected battles than originally anticipated.