Hi, I recently saw this pop up in my recommendations on Steam; while I appreciate the concept of the game, I worry the way creatures are assigned a definitive depth level may have odd, even unwelcome, outcomes on the difficulty curve. Consequently, I am curious about how you intend to prevent the worst of the possible effects described below from happening (as game development may still be in a stage early enough that potential problems can be avoided before they are even encountered).

If I understand the pitch correctly, a creature is “fixed” at a given depth level as a potential opponent either if it was the first to reach this depth, or if it loses there. In other words, the opponent a player will encounter at a given depth is selected among a pool composed of the first creature to reach this depth, those it defeated, the creatures defeated by the creatures defeated by the pioneer creatures, and so on and so forth. The stated objective is that difficulty increases with depth, with no developer interventions necessary.
If we ignore for now the way strategies interact with each others, and model creatures as more or less “dangerous” in the absolute (a pretty big “if”, but hear me out), the consequences of this selection method on difficulty are twofold.
First, the initial difficulty of a given depth level is determined by the pioneer; the implication is that if the pioneer is especially good, it create a spike in difficulty for the following players: there will be an increase in difficulty with depth, but there is no obvious guarantee that this increase will necessarily be gradual. Worst case scenario: a mad genius plays at initial launch and block the second depth level with a virtually invincible juggernaut. While there is statistically little chance of something so drastic happening, given the growing number of depth levels, it seems unpredictable difficulty spikes are unavoidable in the proposed set-up.
Secondly, on the other hand, the first creature to reach a given depth is the more dangerous possible opponent there: the next additions to the pool lost to it, or to something that lost to it, etc. As a result, every addition to the opponent pool is weaker than the last; if opponent selection among this pool is random, the average difficulty of a given depth level gets lower and lower, only bound by the difficult of the depth below it... which is also decreasing. Thus, there may be a “difficulty decay”, as depth levels get swollen with weaker and weaker potential opponents; this is admittedly better than the alternative, a “difficulty creep”, that could easily result from an other way to assign depth levels to creatures. Perhaps giving a defeated creature 50% chance of getting stuck at one level and 50% chance of ending up one level deeper might do the trick, but that's a simplistic hunch that sounds too good to be true...

All that being said, the difficulty an opponent presents isn't actually absolute, but depends largely on which general strategy beat another; while this allays some of my difficulty-related worries mentioned above, it also opens up another can of worms, as the repartition of creatures along depth level will thus be influenced by the metagame. Theoretically, the final version of the game may have an optimal strategy, offering the best win/loss ratio, toward which most experienced players converge and which fills all depth levels after a certain points. If I remember correctly, this situation happened in the history of the famous collectible card game Magic: the Gathering, despite the experience of its developers; whatever you think a healthy metagame looks like, you may not stumble upon it at the first try.
A useful tool to deal with unbalance is simply to patch the game, to make some weak strategies better or the strong ones less dominant. However, as a consequence, at a given depth level the impacted creatures will get weaker or stronger, and thus their power level may become inadequate to the depth level they were assigned to pre-patch.

My tentative conclusion is that it may be necessary for NPC creatures to adjust their depth level based on how well they fare against players on average, “migrating” upward or downward (if the depth level has been founded by a player) in search of a 50% win/loss ratio -adapting to balance changes and smoothing difficulty spikes. More drastically, NPCs may instead (or also) disappear, either after a certain time elapsed or if unbeaten or uncontested for too long, so that if they are indeed unbalanced, at least they won't be for too long, leaving opportunities for players to fill back the niches anew; not very elegant and somewhat frustrating, but the occasional culling of creatures would have the advantage of diminishing the amount of data you'd have to store, if that's a factor worthy of consideration.
That being said I've only thought about your game for a very short time based on what I could read about it on Steam, so I clearly don't really know what I'm talking about; I would like to read more about the mechanics put in place to ensure a relative regularity of the increase in difficulty with depth.