Question: Blitz time control

[Spinojara]

Exactly how is the appropriate time control calculated for ccrl blitz? It says on the website that it is 2'+1'' on an intel i7 4770k. And then that time controls on other processors are calculated using a benchmark of stockfish 10. How is this calculated? I assume it is some nodes per second thing.

[Graham Banks]

I do know that our incremental time control in comparison to repeating time controls was discussed here:
https://talkchess.com/viewtopic.php?t=82553&start=60

[Gabor Szots]

We start from SF 10 benchmark (start the engine and issue 'bench'), based on Graham's machine which gives approximately 2050. If that benchmark provided the same value on my machine, I would use a time control of 2+1. However, my machine is somewhat faster and I get benchmarks around 1600-1700. That corresponds to a multiplier of around 0,8 so it gives me a time control of 96s+0,8s.
Initially I used that 96+0.8. However, I noticed that some engines (mostly very old WinBoard ones) cannot handle increments that are fractions of a second. Therefore I decided to use 1s as the increment. To compensate for the extra time the engines receive by using 1s instead of 0,8 I use a slightly smaller base time, usually 90 s.

[Spinojara]

We start from SF 10 benchmark (start the engine and issue 'bench'), based on Graham's machine which gives approximately 2050. If that benchmark provided the same value on my machine, I would use a time control of 2+1. However, my machine is somewhat faster and I get benchmarks around 1600-1700. That corresponds to a multiplier of around 0,8 so it gives me a time control of 96s+0,8s.
Initially I used that 96+0.8. However, I noticed that some engines (mostly very old WinBoard ones) cannot handle increments that are fractions of a second. Therefore I decided to use 1s as the increment. To compensate for the extra time the engines receive by using 1s instead of 0,8 I use a slightly smaller base time, usually 90 s.

Exactly what I was looking for. Thank you!