Netting against share of flow
Every line on this page is a counterfactual simulation over real trades. The inputs are August 2026 flow on Robinhood Chain, the mechanism is hypothetical because Nokturn did not exist then. This is a backtest, never a measurement, and the word matters enough that it is printed on the chart rather than said once in a script.
What we quote
27 to 33%
Between counterparties, at a realistic early share of 10 to 20%. Not 50,05% and not 63,76%.
The slope itself
21,4% to 50,1%
From 5% of flow to all of it. The shape is the argument, not any single point on it.
Traders per batch, early
2,5 to 3,3
Off hours, at 10 to 20% share. The netting is real and the batches are thin. Both are true at once.
An earlier version of this curve said 21 to 29% at the same share levels. It used the gross definition, which still counts a bot trading back and forth against itself as netting. The counterparty definition throws those out, and the number came back higher, at 27 to 33%. The method was tightened and the result improved. Both curves are drawn above so that claim can be checked rather than taken.
The curve is also a diagnostic, not a showpiece. If netting on mainnet lands far below the line at the share we actually have, the thing at fault is the solver or the batch duration, not the market.
The numbers behind the chart
| Share of flow | Between counterparties | Gross | Traders per batch |
|---|---|---|---|
| 5% | 21.43% | 31.92% | 1.94 |
| 10% | 27.19% | 37.99% | 2.47 |
| 15% | 30.43% | 41.17% | 2.90 |
| 20% | 33.39% | 44.41% | 3.29 |
| 25% | 35.73% | 47.06% | 3.63 |
| 30% | 38.06% | 49.51% | 3.95 |
| 50% | 42.81% | 54.99% | 5.09 |
| 75% | 46.65% | 59.79% | 6.35 |
| 100% | 50.05% | 63.76% | 7.47 |
Source: our own indexed queries over August 2026 flow, off hours, 45 second batches, against the v1.0 allowlist. The nine queries are permanent and public on the Dune dashboard, so the inputs can be rerun. The mechanism on top of them cannot, because it did not exist in August.