- Interpolate a value dependent only on low level features of the position (like MG-EG). For example, count material, blocked pawns, attacked squares, ...
- Interpolate a value dependent on existing, more complex values. For example, the complexity, kingDanger or space parameters.
The advantage of using a quadruple score construction is that the overhead is smaller when a large number of parameters is involved. Naturally, this concept will only work as many Score parameters as possible make use of the then-new dimension.
Please comment your Ideas for usages of QScore, if any.
Also, comment on my idea presented below.
I would like to try using the additional dimension to interpolate using the "complexity" ("initiative") value. When this value is negative, it reduces the MG/EG value by that amount (up to at most 0). In my understanding, this is basically a fortress heuristic.
Ideally, using this "fortress-indicator" as a Score dimension would reduce bonuses like "Outpost" or "Mobility" (bonuses that claim significant positional advantages, but do not on their own lead to a win) in situations that are very fortress-y.
In contrast, bonuses like "unblocked passer" or a middle-game material advantage of +2 have a much larger potential to be converted even if the position looks like a fortress based on the heuristics. Previously, these bonuses would simply be large enough to overcome the complexity penalty. Increasing their effect situationally may allow for larger overall complexity adjustments.