Three benchmarks, two questions
We publish three benchmarks. The first asks how fast we produce a plan and how good it is. The other two ask a harder question: whether a plan that scores well is also a practical plan, one that can be executed as scheduled.
More constraints for a practical plan.
Google OR-Tools
July 2026
Strategic mine planning, against an open-source solver. How quickly can a full ten-year plan be produced, and how good is the plan? Benchmarked on the same orebody, the same constraints and the same assumptions.
View the benchmark →MineLib: McLaughlin
August 2026
What a published solution loses when it has to be executable. A published solution for a 2.1 million block gold deposit, replayed under the physical constraints a real operation works under.
View the benchmark →MineLib: zuck_large
August 2026
The same test on a second deposit, where the result was very different.
View the benchmark →We have run the executability test on two deposits with an identical method. On McLaughlin the published solution lost 58.4% of its stated value. On zuck_large, 29.5%. That spread is the finding: a plan built without physical constraints was not designed to survive them or to fail them, so how it fares once they are applied is essentially random, and nothing in the reported NPV tells a planner which case they are looking at.
All three benchmarks are our own results. Each page sets out its method, what it does not show, and where the underlying data comes from.