Auto-translated
The eternal, crucial, and burning questions of Heroes 3 continue to plague the minds of the community: how awesome are Titans, really? Is the Dragon a rare bird, and if so, will it even make it to the middle of the Dnieper River?
Frankly, it's surprising that after four pages, there still hasn't been a clear and understandable answer to the original poster's question (either vague or not quite accurate). I'll have to correct that.
Let's consider two different approaches to unit control separately: AI and human.
1. Auto-battle.
The Dragon wins if it survives until the 5th turn, but this won't happen very often due to the significant damage variance of the Titan. To calculate the probability of such an event, we need to know the damage distribution programmed into Heroes; this is unknown to us. There was a suggestion to model it as uniform, but this is not only an assumption but also an approximation, since the distribution is discrete (remember that all Heroes discard the fractional part and only operate with integers). Then, the final distribution of the total damage from 4 Titan attacks can be calculated using formulas from probability theory, but believe me, it's not easy. There will be convolution operations, which are extremely difficult to calculate in our case; we will have to deal with a geometric representation of probability in order to solve the problem. We can consider the distribution as discrete and uniform, build a tree of outcomes, but this approach is even more cumbersome. And yet, using other (perhaps somewhat approximate, I'm too lazy to check everything thoroughly) methods, it is possible to approximately calculate the probability of the Dragon's victory; I found it to be approximately 0.6. There is no point in calculating more fanatically, because the initial assumptions are already too strong. Conclusion: in auto-battle, the Dragon is stronger. For those who are tired of all the complicated words and numbers, I will give a simple interpretation in "Heroes" terms. If we ignore the probability of the Dragon dealing maximum damage 4 times in a row and assume that the damage distribution is symmetric with respect to the average (which is almost certainly the case), then in auto-battle, dragons are stronger than titans by exactly +1 attack and defense, i.e., on snow, they will be equal.
2. Under the careful guidance of a human.
We respond to the "defense" of the Titans with the "defense" of the Dragons, and to the retreats – with fly-ins, but only starting from the 4th turn. If the Dragon's morale is high after the fly-in, we click "wait" and get the right to strike first on the 5th turn, which means winning with a probability similar to that in auto-battle – approximately 0.6. If the morale drops for the Titan, it wins almost certainly, which means the final distribution of forces is approximately 70 to 30 in favor of the Titans.
Which of the two approaches to assessing the relative strength of units is closer to the "Heroes" reality, you decide for yourself; depending on this, the answer to the question of who is REALLY stronger changes.
After all, it is quite obvious that a battle between two groups of level 7 units with zero-level heroes and no magic is something like a "spherical horse in a vacuum".
P.S. Let's hope that at least someone will be able to read all these "many letters".