Origins
Discussions around “the best dice set” in Kingdom Come: Deliverance 2 are common, but rarely supported by quantitative evidence. To address this, I developed a statistical dice simulator capable of evaluating dice sets under controlled, reproducible conditions and optimal play strategies.
Features
Scope & Constraints
The simulator currently evaluates sets of six identical dice. Mixed‑type sets are excluded because optimal play strategies differ significantly when dice have heterogeneous probability distributions.
For example:
If three dice have a high probability of rolling 1s and the other three do not, an optimal strategy would delay scoring the high‑probability dice to maximize the chance of clearing all dice.
Supporting mixed sets would require a more complex strategy engine, which is planned for future iterations.
Deterministic Simulation Framework
The simulator uses a deterministic seeding model inspired by the MIDI Machine:
A user‑provided string is hashed.
The resulting hash seeds a PRNG.
Independent PRNG streams can be assigned to different turns or different simulation modes.
This ensures that results are reproducible and comparable across runs.
Statistical Outputs
For each dice set, the simulator computes key descriptive statistics for an optimally played turn, using a strategy that scores:
All remaining dice, if possible
Combos when available
Otherwise, a single 1 or 5
The output includes:
Mean
Median
Minimum
Maximum
25th and 75th percentiles
These metrics provide a clear picture of both expected performance and distribution shape.
Optimal Passing Strategy Search
The simulator includes a mode that brute‑forces the best parameters for deciding when to pass a turn. It evaluates combinations of:
Dice Threshold — pass when the number of remaining dice is ≤ threshold AND
Minimum Score Requirement — pass only if the current score meets or exceeds this value
For each dice set, the simulator reports:
The maximum achievable mean and its corresponding parameters
The maximum achievable median and its corresponding parameters
Visual charts illustrating distribution and performance
This allows players (or researchers) to compare strategies based on risk tolerance and desired outcome metrics.
Performance & Multithreading
The brute‑force mode uses multithreading due to the significantly higher number of simulations required.
Standard simulation mode remains single‑threaded for determinism and simplicity.