Originally Posted by

**Ms1**
It may seem better, yes. But it was discussed and dismissed.

Such formulas make sense if we try to answer questions like "how hard is it to reach this level of excellence?"

We are interested in a practical economical result, not the volume of efforts. We believe that normally the cost function is linear.

20% of efforts gave 80% of results? The user does not care.

I don't understand. The formula changes the order of ranking. It is the same as geometric mean when benchmarking.

A practical example

Code:

codec1: c_time 100MB/s d_time 1GB/s ratio 45% -> score 10+2+450 = 462
codec2: c_time 1GB/s d_time 10 GB/s ratio 47% -> score 1 + 0.2 + 470 = 471.2

But the codec 2 is close to 10x better:

compare to geometric formula c_time^0.5*d_time^0.5*(size/10,000,000)^1.5

α,β,γ should satisfy α+β <= γ

/10,000,000 doesn't change order

Code:

codec1: 10^0.5*2^0.5*45^1.5 = 1350
codec2: 1^0.5*0.2^0.5*47^1.5 = 144

so 9.3x better