You're right it isn't a good solution. Its the function c*I*A (c > 0). Not what we want is it? What we want is c*A + n where c < 0 and n > 0. This would be a "Dual wield is most efficient at 1 attack profile and less efficient as A approaches infinity".
And we also want to lock dual wield to be obviously less powerful than it is now correct? But I would say at some point dual wield is balanced for its return on investment based on the number of profile attacks. So lets assume at some point dual wield is a wash between the alternatives and we want that wash to exist in our new dual wielding system.
I'm going to pick an arbitrary point along that line and say at 2.5 profile attacks, dual wield in current mordheim settings should be about equal to dual wielding in our new custom awesome system we are creating.
Since the current system fancies a 1/x curve for what we are trying to balance and the new system is a simple -c*X + n line. We have to pick a C and an n. The N is the return on investment when you have 1 profile attack, and C is the diminishing returns of that investment (the slope at which the benefits of dual wield are reduced as more profile attacks are added).
if a shield gives a 1/3 chance of dodging a blow (roughly, not counting crits), we'll go ahead and assume that dual wielding should cancel this and increase the potential damage to 50% more than normal. So at A = 1, N = 1.5. Our arbitrary point in space (2.5 attacks) is going to be fixed as well and that return on investment is 3.5/2.5 = 7/5 = 1.4.
We have 2 points - (1,1.5) and (2.5,1.4).
y = -.066666*A + 1.566666
At 1 A: y = 1.500, A*y = 1.500
At 2 A: y = 1.433, A*y = 2.866
At 3 A: y = 1.366, A*y = 4.099
At 4 A: y = 1.300, A*y = 5.200
So lets try and fit something in.
Proposition: For each profile attack, roll d6. If you roll a 4, 5, or 6, you may make one
additional attack with your offhand weapon. For each additional 6 you roll, you get an additional bonus attack as well. Very complicated yes, but I'm in the surreal realm of anything goes at the moment (Tweaking comes later).
Now for the evaluation of whats occuring:
1 A: You have a 50% chance to attack twice. Odds are you have 1.5 profile attacks!
2 A: You have a 25% chance of attacking twice (missing both dual wield rolls), and a 13.8% (5/36) chance of attacking 4 times (rolling 6,6 or 6,5 or 6,4), and a 61.2% chance of attacking 3 times. Odds are you have 2.888 profile attacks! (look how close that is to the A*y value, which is 2.866)
3 A: You have a 12.5% chance of attacking 3x (missing 3 dual wield rolls), 57.8% chance of attacking 4x, 26.4% chance of attacking 5x, and 3.2% chance of attacking 6x. Odds are you have 4.2 attacks on average.
4 A: Too complicated to compute, i used a spreadsheet with countif's on 3 A to simplify my life and i'll just guess that you'll have somewhere above 5 attacks but less than 6
So with this change we've effectively changed dual wield from (1->2, 2->3, 3->4, 4->5) to (1->1.5, 2->2.9, 3->4.2, 4->5.X)
But this is far to complicated for mordheim and is unreasonable. Try suggesting this to players and they'll look at you with that "Huh?" look.
We can simplify it pretty easily though. How about you only get your offhand attack on a 4+? It effectively adds "half an attack" to your profile for dual wielding. Pretty simple as well. But it is another dice roll you have to deal with in the hand to hand phase (Yuck)!.
I dunno, lots of food for thought here.
If you want to make it really simple and balance dual wield vs shields just say shields give a 5+ unmodifyable save before crits are assessed. It'll cancel 1/3rd of incoming attacks. The dual wield would go from 2 attacks to 1.33 attacks, from 3 attacks to 2 attacks, and from 4 attacks to 2.66 attacks. At 2 profile attacks the shield would completely nullify DW.
Did I go off on a tangent? Somebody smack me back into reality.