I estimate the value of an Alternator with 2 forms to be close to ((ideal/practical value of) form 1+(ideal/practical value of) form 2)/2 (Use the ideal values ​​of the pieces if both forms are colorbound when separated but not colorbound if they become one & alternate) if the weaker form is just a strictly weaker version of the stronger form. E.g, an NW/N Alternator would be worth about (5+3)/2=4. For an Alternator where the moves of the 2 forms don't overlap, it'd gain some bonus. So, my Alacrious Alternator might be stronger than Fabulous FIDEs.

For pieces that promote upon capture, your opponent probably won't let that happen, so their value should be closer to their unpromoted form, while for pieces that demote upon capture, I estimate their value to be (value of undemoted form+value of demoted form)/2.

That seem reasonable guesses. The average value could be biased a little bit towards the stronger version, though, as the player handling the piece is likely to keep it in that form most of the time. In fact the value difference between the components becomes like a positional term such as mobility; for the latter you also don't have to average over the entire board (with equal weight to all squares), because players will avoid the really bad squares. A Knight in the corner has only 1/4 of the mobility of an edge in the center. But for tactical considerations it is worth a lot more than a quarter of a Knight, as it can be moved to a much better location at the cost of one or two tempi. For the practical value of a Knight it hardly matters that the corner squares are so poor; players just never put it there voluntarily, and in the rare case they had to do it for capturing something, they will immediately move it away again, and just incorporate the cost of the lost tempi as part of the tactical exchange. With alternators it is similar; a piece that alternates between, say, Wazir and Queen would never be kept in Wazir form for very long; changing it back to its Queen form would likely be the move with the highest priority, similar to developing a still undeveloped piece.

For morphing on capture the situation is less clear, as in general you cannot capture at will (and survive).

In the Mad Morphers design #2 I tried to avoid such complications by only morphing between pieces that are very similar in value. Which is why I discarded Mat Winther's Elk (a shade-based R-N alternator). Archbishop and Chancellor are very similar in value (0.25 Pawn difference, and which is actually best might depend on game phase). I wanted to have some diversity in the morph condition (shade, move, capture), but morphing on capture (as in micro-Shogi) only makes sense for pieces that are usually not recaptured immediately. And the super-piece seems the most likely to be used last in any tactical exchange.

The Happy Hoppers will still need some work; what I inittially envisaged is way too strong. Since a Mao tested as half a Knight I figured that combining it with a Lame Alibaba would bring it back up to a full Knight's value. The Diagram also seems to overestimate the value of a Mao, so I wasn't too concerned it valued the Lame Squirrel much higher. But the empirical value of the Mao could be suppressed by the fact that you can block two of its moves on the same square. The Diagram might not account enough for opponents specifically targeting to do do that. Anyway, by testing with Fairy-Max the empirical value of the Lame Squirrel turned out to be rook-like. So I am trying Mao + Lame Alfil now, a pair of which appears to be slightly worse than a Bishop pair. (Moa + Lame Dabbaba should be similar.) For the Cricket the mgQcQ3 appears to be a good Rook replacement, it tested as ~0.5 Pawn stronger than a Rook.

A Cannon has about the same value as a minor, so if the super-piece is not much stronger than a Queen the army that otherwise has nA[W-F], mRcpR, mgQcQ3 should be well balanced w.r.t. the classical Betza armies. The nA[W-F] is not a hopper (but not all pieces of the Colorbound Clobberers are color bound either...), but I included it to also have a piece the value of which goes up towards the end-game. When lameness no longer counts for much its value should become rook-like, while the value of the Criclet and Cannon would drop. Unfortunately the regular version of Fairy-Max cannot handle locust capture. So I cannot test the Killer Kangaroo. I thought about using the Long Leaper of Ultima, but I did not like the weakness of it in attacking edge and (especially) corner squares. In Ultima it is only a major piece by having stalemate count as a win. So I added capture to an edge square as extra move. And for simplicity abandoned the multiple capture, which seemed to contibute only little, as you would hardly have the opportunity. The Diagram estimates the piece to be stronger than Queen, but I am not sure I can trust this. It is not very clear how locust capture should be evaluated, and logic says that in any given position a Queen should have more captures (of the same pieces) as the Killer Kangaroo, as the latter's captures could be blocked. The KK might have choice of destination squares, though, making it easy to dodge retaliation, so the question is how much this is really worth. If that turns out to be too much, this could be remedied by requiring it lands immediately after the capture victim.