We want to determine the IQ of an SDG robot.
We subdivide the IQ into sub-IQs for subactivities.

Buying IQ:
How good is the robot at making buying decisions?

The robot is offered a set of derivatives to buy.
Half of them are too cheap (CHEAP) and half too expensive (EXP).
The robot buys set BUY of derivatives.
BuyingIQ = intersect(BUY,CHEAP) - 2 * intersect(BUY,EXP)

Offering IQ:
What is the quality of the derivatives offered by the robot?

The robot creates a set S of derivatives.
The robot should only offer profitable derivatives and they
should not be too expensive.
OfferingIQ = 
SUM over all derivatives in S:
if price > break-even-price then 1 - (price - break-even-price) else 0

RawMaterial IQ:
What is the quality of the raw material created by the robot?
The raw material should be symmetric and use the worst weights.

Input: derivative
Output: raw material rm
(if symmetric then 1 else 0) - ( fsat(rm,Jmax) - fsat(worst,Jmax))
Jmax is the maximum assignment.

Cannot use test cases: 
derivative for which the worst raw material is known.
Need symmetry checker.
Need to find the best assignment (slow!).
Determining the IQ is time consuming (NP-hard)!
Is the a better definition of RawMaterial IQ that is polynomial?


Finishing IQ:
What is the quality of the assignments?
Input: raw material rm
Output: assignment J
FinishingIQ =
if fsat(rm,J) = fsat(rm,Jmax) then 1 else fsat(rm,J)-fsat(rm,Jmax) /* is negative */ +
if fsat(rm,J) >= E(Z(rm,bmax)) then 2 else 0
Rationale:
Against rational players fsat(rm,J) >= E(Z(rm,bmax)) must hold for an intelligent player.

Use test cases: raw materials for which 
best assignment is known and for which best
expected value is known.
Need quality checker.




RobotIQ = BuyingIQ + OfferingIQ + RawMaterialIQ + FinishingIQ