Approaches

Every strategy family and approach directory in the repository, with its documentation. Each approach page renders the MDX notes stored in that directory — a learning resource for how the strategy works and what the evidence showed.

Afterstate learning

1 approaches

Judge the position a move leaves behind rather than the move itself — and build training data in which every column the player could have chosen has been measured, not guessed.

1 written · 0 draft · 1 total · family guide

Baselines and diagnostics

7 approaches

The measuring instruments — simple reference players, cross-checks between the two rules engines, and per-move counters that say what a policy is actually doing.

7 written · 0 draft · 7 total · family guide

Constructive and reservoir policies

7 approaches

Instead of taking the points that happen to be available, deliberately build a structure that will pay off later — and keep it alive through the rows that keep pushing up from underneath.

7 written · 0 draft · 7 total · family guide

Depth-4 search and long outcomes

4 approaches

The reference search sees four of your own moves; these experiments judge a move by playing on for twenty-five, then try to keep only the moves that survive that longer look.

4 written · 0 draft · 4 total · family guide

Fair expectimax

11 approaches

Look a few moves ahead, take the best column on your own turns, and average honestly over every disc the game might deal — no hoping, no fearing.

11 written · 0 draft · 11 total · family guide

Hand-built heuristics and shallow search

15 approaches

Give points to the things about a board that a human would notice — low stacks, open columns, gray discs you can reach, numbers that are nearly ready to clear — add them up, and play the column with the best total.

15 written · 0 draft · 15 total · family guide

Lifetime objective

15 approaches

Judge a move by how much longer the game will still last, rather than by how many points it scores right now.

15 written · 0 draft · 15 total · family guide

N-tuple networks and learning from play

12 approaches

Instead of hand-writing what makes a Drop7 board good, learn it — from small patches of cells, and from millions of games the program plays against itself.

12 written · 0 draft · 12 total · family guide

Oracle and curriculum research

7 approaches

Let a planner cheat — show it the hidden numbers and the discs still to come — then try to teach a player who cannot cheat.

7 written · 0 draft · 7 total · family guide

Terminal policy iteration

6 approaches

Choose between moves by actually playing each of them forward many times and seeing which one leads to the longest, highest-scoring games — then change the reference policy's mind only when the evidence is overwhelming.

6 written · 0 draft · 6 total · family guide

Tree search

4 approaches

Instead of examining every column to a fixed depth, grow the look-ahead only where it looks promising, guided by quick simulated playouts.

4 written · 0 draft · 4 total · family guide

Value and policy learning

12 approaches

Instead of searching ahead, train a model on past games to judge a board or pick a column — and learn why that kept failing.

12 written · 0 draft · 12 total · family guide