Cutoff

A Benchmarker’s cutoff is a cap on the number of their solutions that can earn rewards, it is used when determining qualifiers. In TIG, a “cutoff” is attributed to each Benchmarker, this limits the number of their solutions which can qualify for block rewards. The cutoff acts as a hard defense, ensuring that Benchmarkers balance their Proof of Work, this protects the protocol against manipulation of the Pareto Frontier.

Calculation of the Cutoff

For each challenge $x$, a Benchmarker $i$ is attributed a cutoff, denoted ${cutoff}_{i,x}$. Benchmarkers are capped by their ${cutoff}_{i,x}$, that is, Benchmarker $i$ can’t have more than ${cutoff}_{i,x}$ qualifiers for challenge $x$. A Benchmarker’s cutoff is recalculated every block:

Case 1: If Benchmarker $i$ has less than 10 active solutions in each challenge:

Case 2: Otherwise:

The first component of the cutoff provides a minimum failsafe that stops Benchmarkers from submitting a small amount of unbalanced proof-of-work. Without this, a Benchmarker could sidestep significant imbalance penalties by spreading activity across multiple aliases, possibly exploiting free compute.

If a Benchmarker has more than 10 solutions in each challenge, then the second component of the cutoff is active. Note that a Benchmarker $i$‘s ${moving\_challenge\_factor}_i$ is a moving average of their challenge factors across all challenges. If there are $n$ challenges, then at block $t$, ${moving\_challenge\_factor}_i$ is calculated as:

where $w$ is a weight initially set to $0.8$. Note that $\langle \text{challenge\_factor}_i(t)$ is the average of the challenge factors of Benchmarker $i$ from the previous block.

This second component of the cutoff ensures that once a Benchmarker has solved at least 10 instances across all challenges, they can qualify for rewards, and the 1.2 multiplier allows their cutoff to grow gradually. We use a moving average to avoid situations where a sudden upward shift in the Pareto frontier would instantaneously reduce a Benchmarker’s qualifiers to zero, imposing an overly harsh penalty.