This subject has become a major issue in Private League 75, which has several gms that have been playing for a long time and post extensively in both the league and this community forum.
The use of less than 4 or 5 plays on either defense or offense,
intentionally, most likely results from those plays significantly standing out from the rest based on their average per play. As an example, given the choice of using a Cover 2 defense on 3rd & long that gives up an average of 8 yds versus a Double LB Blitz that yields 2 yds per play, there's an obvious advantage and many do not want to leave it up to the AI "Coach" to call the blitz. So, gms use rules to make specific defense calls, which can cause extreme results in the interface between several rules and the regular gameplan, with the use of 4 or 5 or less plays for the entire game.
Additionally, does play familiarity influence AI play selection? Very important to know this if it does.
Whether extremely limited play selection is intentional or not, I would like to suggest a penalty or control to curb it. Similar to when a team has too many players, for example, the AI steps in and cuts a player or sits a player (painfully one of the best) when the roster is above the allowed number. And, if a starter is played constantly, the fatigue factor comes into play and his ability/performance is decreased. I would think the same could be done for overuse of a play.
There is not much that can be done to balance play results so that the plays that tend to produce better results become more numerous and variable. With that in mind, like team rosters, you have your better players mixed with other of varying ability, which is all a part of the decision making process & strategy for a gm. It should be the same for the playbook and play selection, as you would have to use a mix of your best plays and others.
Here is the discussion thread on the subject from L75:
https://private75.myfootballnow.com/forums/thread/1/215?page=2#1389