If you encounter problems or errors increase past time. If you have resources and tools but nothing to use them on increase future time. If your state remains too much the same (ie boredom) increase present time. These should feed into each other. If you stop to analyse or plan for too long a time nothing is happening so you should increase doing. When you do stuff it is either harmful, neutral or positive. If it is positive you can do more of it. If you do something harmful you need to understand it and trigger analysing. If your wheels are spinning empty you need to come up how the activity is meaningful to you. When you are done analysing you either find A) something to positive impact your situation (trigger present mode) B) something negative to avoid (trigger future mode how your life will be without that component). When you have dreamed something desirable you either a) now have a problem of not having means to attain it (trigger past mode) b) have a clear action to follow through (trigger present mode).

These do not translate into concrete divisons but give guidelines how to switch between the mindsets and why.

hypothesis testing, amongst other practices, is regarded as the key to learning. It's a process of mining the past for clues about the future, in the present.

Framing those 3 time perspectives in terms of procedural knowledge, rather than Zimbardo's framework or the arbitrary terms of 'past, present and future' may be the optimal way of thinking about them, other than for communication.

One of the most cited works in this area, Chi et al. (1981), examines how experts (PhD students in physics) and novices (undergraduate students that completed one semester of mechanics) categorize and represent physics problems. They found that novices sort problems into categories based upon surface features (e.g., keywords in the problem statement or visual configurations of the objects depicted). Experts, however, categorize problems based upon their deep structures (i.e., the main physics principle used to solve the problem).

Their findings also suggest that while the schemas of both novices and experts are activated by the same features of a problem statement, the experts’ schemas contain more procedural knowledge which aid in determining which principle to apply, and novices’ schemas contain mostly declarative knowledge which do not aid in determining methods for solution.

You're framing this in terms of yourself, but the most important area I'd say appears to be omitted. Learning about problems not involving you can teach you a lot about what you should do. The knowledge of others vastly exceeds your own. I guess this might be past since you wouldn't hear about it until it happened usually. Future and present planning is best determined by past analysis, but the reverse is not true. You cannot learn much about the past from the present or the future. The present also will not teach you much about the future, but the future will help you with the present. This to me presents a hierarchy of past>future>present.

I think you need to further refine your definitions. In reality, there is no "present", only asymptotically close past and future. And, of course, you can't do or change anything in the past - all you have is a choice about how future you uses future seconds.

However, this makes your hard question harder, not easier. It's not just 3 buckets you need to balance, it's a continuous function of distance, importance, and expected return on learning (studying the past and planning the future) vs action (executing plans).