Adversarial results

Every number, and the command that reproduces it.

These results keep the mechanism’s favorable and unfavorable cases together. Every number below comes from a named KNOT test that can be rerun from the repository.

forge test --match-contract MEVProtectionTest -vv
forge test --match-contract EconomicViabilityTest -vv
forge test --match-contract RouterRealismTest -vv

Constructed round-trip fixture

The same synthetic reserve sequence, run with and without the federation:

	Attacker P&L, 5-token stake
Independent pools	+13.5843, profitable
Under Knot	-0.0298, a loss
Difference between fixtures	13.6141

The KNOT path is lower at every sampled size, from 0.1 to 20 tokens. This proves the mechanics of that constructed sequence; it is not a historical replay or a production-profit forecast.

Routing

The router fixtures check quote ordering, not organic order flow:

Samples routed between the two members	40, all land on a member by construction
Deep member vs an identical outside pool	0 wins · 40 ties · 0 losses

The external control has the same reserves and fee as the deep member, so a tie is the only honest result. These tests do not establish how a production router or real flow would respond.

Same-member sandwich fixtures

The attacker sells the exact output inventory acquired in the front-run, so both portfolios close in currency0.

Fixture	KNOT attacker delta	Hookless control
Skewed member	-1.347578	not the control used for this regression
Balanced member	≈+0.393701	≈+1.803279

KNOT dampens the balanced fixture but does not eliminate the sandwich profit. The skewed fixture closes at a loss. These are bounded simulation results, not a universal MEV claim.

Skew sensitivity

Skew	Isolated-quote reduction
1× (balanced)	0 bps
2×	4,299 bps
4×	6,674 bps
6×	7,466 bps
8×	7,862 bps

An earlier version divided retained output tokens by hypothetical input-token fees. That ratio was dimensionally invalid and has been removed. A normalized fixture can compare both values in one numeraire, but it still would not measure routing loss or realised LP returns, so it is not used as a headline claim.

Stateful

Twelve sequential extraction attempts with state carried forward: the quote for a fixed size falls from 1.2678 to 1.2406. The test advances the member and aggregate books with the gross input on every step, matching executeSwap. In this sequence the fixed-size quote tightens as reserves move.