Risks of Innate Magic

The Bane raised some questions about the downside/save mechanic I've suggested for racial spell-like abilities and the innate abilities of the Gifted. Now, I may have over-thought the rule, and I'm certainly considering other options, so it's worth discussing; but first, I need to explain how I got to where I did.

First was the inspiration: I like the way the original Fighter and Magic-User seem to have just two abilities each, one base and one scaled, and wanted a class system based on that principle. I replaced the thief and cleric, added the charmer, and decided if I added another supernatural class, I'd have three mundane and three supernatural, which I could also match up to the six attributes to make a neat, symmetrical system:

Mundane Class	Supernatural Class
Fighter (Str)	??? (Con)
Trickster (Dex)	Magic-User* (Int)
Charmer (Cha)	Spirit-Worker* (Wis)

"A supernatural class based on Constitution" sounded to me like a class that "powered" its own magic, as opposed to using intangible forces (Magic-User) or asking spirits/gods for aid (Cleric/Spirit-Worker.) Hence, the Gifted. And in keeping with my stinginess and insistence on reusing existing rules, I figured the Gifted's powers should be paid for in hit points. Also, I wanted them to trade power variety in exchange for for frequent power use, to make them more distinct from Magic-Users.

The problem is, in D&D, characters start with fewer hit points than in Microlite 20. So, I knew I couldn't go with a flat per-use spell point cost, like M20's 1 + 2*spell level hit point cost. Hence, the mechanic I used reduces the risk considerably:when Gifted character uses a power of level X:

1. roll for the power's damage or effect on target (one or more d6s);
2. roll a Con save;
3. if character fails the Con save, remove all damage/effect dice that are greater than spell's level;
4. total remaining dice for hit point cost.

So, a gift does not always cost hit points, only on a failed save, and even then, a 1st level spell is only going to have a 1 in 6 chance of costing a hit point. The double test does seem a little clunky, though, so I've thought about merging the Con save and spell level test in the third step: every d6 rolled is compared to Con/3 (I'm assuming this is already recorded.) Each die that is higher counts as 1 hit point of damage.

Then again, maybe it's better to switch to treating a failed Con save as indicating fatigue, using the sleep deprivation rule (1 failed save = 1 point of damage, confused, easier to distract; Con save again every 24 hours or next use of power; 2nd failed save = another point of damage, penalty on all actions; continued failed saves = continued damage.)

Or maybe other kinds of risks, as long as there's some kind of risk. I'm sure I'll have more to say later.

* I still debate shortening Magic-user and Spirit-Worker.