Starting to think my plan of analysis is not possible, please help!

Thanks in advance for bearing with me:

I have 120 participants, all randomly assigned to one of three conditions leaving 40 particpants in each condition.
Lets call the conditions seller A, seller Z and Buyer.
I need groups, preferably more than 40, so it is fine for one person to be in several groups as long they do not end up with oneself.
However, how the groups are comprised is critical; first of all, all groups need exactly three participants. But most importantly; half of the groups needs to have two seller A's and one buyer, while the other half needs to have two seller Z's and one buyer.
For each group I need one value; the sum of all three values of all three members on the variable 'price'.

I have been thinking of some form of stratified random sampling, and later combining the price variables. Or making a new variable which assigns the scores of two other particpants (A or Z + buyer) to that participant.
None of the syntaxes I've tried worked however, and I've pressed all the buttons I could think of... The problem seems to be the fact that I need groups with very particular rules. Need help badly!