The agents may show both tactical as well as strategic behavior.

from Merriam-Webster

tactical:
(1) : of or relating to small-scale actions serving a larger purpose (2) : made or carried out with only a limited or immediate end in view b : adroit in planning or maneuvering to accomplish a purpose

strategic:
of great importance within an integrated whole or to a planned effect

Tactical behavior includes using a weakness in most or all other players.
They may have forgotten to have checked for a certain border case.
Or they may not yet have implemented a certain feature.

Strategic behavior includes implementing a strong solver and not worry about "tactical" issues.

We see both tactical and strategic behaviors in every semester.

The goal of the game is to find the best algorithms and best implementations
in a given domain, in our case Boolean MAX CSP.

Tactical behavior is used by some of the agents because it is easier to implement and profit from,
as long as the other agents have not protected themselves against such tactical maneuvers.
The tactical behaviors serve as useful maintenance requirements generators in the cycle:
observe the history, find the reason for losing, formulate new 
maintenance requirements for the agent, plan, implement and test.
However, the  tactical behaviors don't contribute so much to getting strong algorithms.

Brent suggested using a static benchmark to discourage too much tactical behavior.
The static benchmark measures the performance of the SOLVE component of the agent
on a set of problems. The average satisfaction quality reached by an agent on the benchmark
serves as IQ for the SOLVE component of the agent.

How is the static benchmark created? It will be created automatically from
problems that appeared in the most recent game.
It will consist of two kinds of problems:
(1) Problems that were solved well and created a profit for the solver = acceptor.
(2) Problems that were not solved well and created a loss for the solver = acceptor.

The results of both (1) and (2) will challenge the weaker solvers.

Both the number of points achieved in the competition as well as the average
satisfaction quality achieved on the benchmark will be used
to determine the letter grade for "agent software development performance".

The question is:
Is it necessary to control tactical behavior of the agents?
Or should we just let the game without the benchmark discourage
tactical behavior? With good strategic behavior you can win much more
than with tactical behavior.

We also have the component market: The strategically thinking software developers
can contribute to the component market and get their work recognized that way
by having their software used in other agents.