I checked the formula for prayer fumble, because it looked strange to me with wisdom 45 still getting prayer fumbles.

spell_util.c line 707

if (item == spellNormal && op->type == PLAYER && s->flags & SPELL_DESC_WIS
            && RANDOM_ROLL(0, 99) < s->level / (float) MAX(1, op->chosen_skill->level) * 1)

First we have a multiplier *1, this is only processor cost.
Second it looks like, we have always at least 1% chance to fumble, when we roll a 0?
Third when we have something like integer/float the result is integer, so the float must be rounded/truncated somewhere. I am not 100% sure if it is before the operation or after.

So it looks like only when wisdom(stats) level is higher than wisdom(spelllevel) we have no fumble or the 1% depending on the round/truncate behavior above.

But i am sure, when you have level 110 prayer and your wisdom skill is 110, you have 1% fumble.

I had some other ideas. Currently we have int against pow, int is nice but pow is better. It's same with cha and wis.

What if only int goes to spell fumble and pow is damage and mana? Something like

spelllevel-(int*2)+-itemspellfumble.

This could be same for priest with charisma. Tabernackle likes nice guys. And prayers work better. 

prayerlevel-(cha*2)+-purity value.

So each class has a damage boost attribute and a hit chance attribute.

Could also be interesting to use the hit chance attribute for crit damage. 

like
int 20% grants 20% critchance
cha 20% grants 20% critchance (tabernackle likes you) 
dex 20% grants 20% critchance.

So int, cha and dex gives also damage but by using critchances.

And finally we have ranged classes, they mostly need their dex and instead of using dex for direct damage boost, dex is used for critdamage boost.

We can do the same on throwing weapons. Dex is crit and str is direct damage boost.

Don't know how we can build in a secondary stat for sling, bow, crossbow? Sling and bow could be strength bonus too, but crossbow not. Perhaps crossbow becomes a higher bonus on critdamage, bow a medium bonus, sling slowest, but sling becomes the most bonus for strength, bow medium, crossbow nothing?

And can someone put this thread in suggestions? 

Actually the problem with SF in general is that it's not really been developed to have any purpose other than to prevent a player from casting spells. So basically it has a narrow range of effectiveness and once you have maxed (negative) SF any additional gain to the stat is useless. So to combat that I'd have two suggestions

1. Have more powerful versions of spells with a higher SF requirement. So that would mean that Pow would increase your spell damage in general while Int would give you access to more powerful spells.
2. Have any excess SF be converted into a double cast. So for example if you have 150% of the SF required to cast the spell it would change to 1 spell of 100% success rate and a 50% chance to simultaneously cast it again.

I'm assuming skill level in your formula is the spell level (e.g. Lightning is a lvl 50 quest so spell level would be 50) rather than the player's level in the skill as otherwise players would become less good at casting the more they train in the skill. Considering that Firestorm is a lvl 15 req quest and it's clone Icestorm I think is lvl 30 or 35 iirc I don't know if that would cause issues in implementation. Same would apply to Firebolt, Lightning and Frostbolt.

Just working through the numbers to get a lvl 50 spell to fumble 12.5% of the time requires 15 int. On the other hand to get a lvl 110 spell to fumble 12.5% of the time requires 75 int. So the big question is would the damage output of a lvl 110 spell be the same as a lvl 50 spell with 60 extra pow.

From a player viewpoint I'd be tempted to add guilds and attunements to the calcs but from a server processing viewpoint a simple calculation is faster. I'll note that as a slight optimisation you could store a value of skill level - B in the skill object so that if the constant is always 10 then a lvl 50 spell would have a lvl 40 SF value. Secondly would it be faster if instead of doing a d100 ((spell level - int)/2) you did a d200 (spell level - int)

I'll note that there is nothing that says spell level has to be directly linked to quest level to obtain the spell. So you could in theory have a quest that is unlocked with lvl 50 Ma that has as a reward a lvl 15 spell.

Once relatively quick version could be to change firebolt into minor, lesser, small, medium etc versions. Each version would have higher base damage and SF than the previous one. I've not really thought out the numbers too much so it's a bit off the cuff but I would suggest that each level has 1.2x the damage, mana cost and SF. So that would mean that a Tier 4 spell (Medium) would have 72.8% more damage but would cost 72.8% more mana than the Tier 1 (Minor) version and if the Tier 1 version used 110 SF base then the Tier 4 version uses 190 SF. So to get the same fumble rate you need an extra 40 int.

Theoretically extra effects could be changed so that for example the size of the cone in a storm could also depend on the spell level. You could in theory have an omnidirectional magic bullet etc. Having said that it would take a lot more work than just changing dmg, mana and SF values where you could in theory just loop through the object/array and copy new values in a cloned object

As far as priests lacking a SF stat you could compensate for it by giving a higher weighting to their equivalent stat so that for example mages are SF+int*2 while priests are cha*4