Hier der Beitrag auf Discord:

Why is there such confusion?
There are two types of parameter tuning:
1) What Kiudee method does via playouts
2) Optimising the parameters for highest speed (Which is what my "Tuner program" does.)

Example: Tilps suggestion to add threads=2 to Demux backend opts, produces an extra 10000 nps.

Currently being tested are the following parameter combinations:
```{
  "seconds_per_move": 10,
  "options": {
    "backend-opts": ["backend=cuda-fp16,multi_stream=true,threads=2,(gpu=0),(gpu=1),(gpu=2),(gpu=3)"
    ],
    "multi-gather": ["true"
    ],
    "backend": ["demux"
    ],
    "threads": [2,3
    ],
    "nncache": [50000000,20000000,30000000,40000000
    ],
    "minibatch-size": [816,408
    ],
    "max-prefetch": [37,160,32,48,64
    ],
    "cpuct": [2.0179,1.745,0.95
    ],
    "cpuct-base": [50000,38739
    ],
    "cpuct-factor": [3.8542,3.894
    ],
    "fpu-value": [0.4089,0.33,0.25
    ],
    "temperature": [1.4229,1.359
    ],
    "smart-pruning-minimum-batches": [300,0
    ],
    "max-out-of-order-evals-factor": [2.4,2,3
    ],
    "smart-pruning-factor": [3.6141,1.33
    ],
    "time-manager": ["alphazero(alphazero-time-pct=23.06290)",legacy
    ],
    "verbose-move-stats": ["true","false"
    ],
    "ponder": ["false"
    ],
    "move-overhead": [0
    ]
   
  }
}```
And there are 9216 configuration combinations of the above.

Lothar