KaspaKaha: "1/ Appreciate the public follow-up @emdin - and real credit for correcting the o"

KaspaKaha@KaspaKaha·2d

1/ Appreciate the public follow-up @emdin - and real credit for correcting the ordering point. That was the substance, glad it’s settled. We’re a team of MEV researchers; transparency is the job. So here’s everything, on the table 👇 2/ Where we agree - which is most of it: sandwiching on Igra is real, persistently attempted, and as attempted it’s been unprofitable. Higher carrier feerate wins the same-block slot. Your findings and ours line up. No fight there. 3/ But “0/12 net-negative → not profitable” has a gap, and it’s the crux. That measures the sandwiches that were attempted, at the sizes those attackers happened to pick. Sandwich P&L is convex in front-run size - one optimum; too small or too large both lose. 4/ So the same victim swap flips negative→positive at the right size. The result shows those attackers sized wrong and left EV on the table - not that an optimally-sized sandwich on those victims is unprofitable. “Not profitably attempted yet” ≠ “not profitable.” The claim needs the per-victim optimum, not the observed sample. 5/ Our number, since you asked - no deflection. 446K KAS total, never a sandwich figure: • Arbitrage ~99% · Liquidations ~1% · Sandwiches ~0 (we agree with you) By chain: Igra 365,149 KAS (78%), Kasplex 22%. 6/ Definitions, so it’s checkable: • Completed sandwich = same-Kaspa-block front+back-run, both land, value = backrun − frontrun − fees → ~0 net-positive (matches your 12). • Arb extracted value = realized WKAS on a completed cycle, net of carrier fee + gas. • Venues: Igra L2 + Kasplex L2. 7/ So both hold: front-running resisted-as-executed (agreed); arbitrage MEV live and ~99% of realized extraction (the data). On sandwich profitability, we read it as a sizing result, not a mechanism result - and we’ll show the optimization against your ZS dataset. That’s the engagement worth having. - a team of MEV experts. Data’s on the table.

Pavel Emdin@emdin

Hey @KaspaKaha, TL;DR this is wrong info, as I already mentioned in my DM to you several days ago. Results of our research which we'll publish soon: ~6% of frontrunning attempts win the carrier race and land (11 confirmed sandwiches over ~25 days), and 0% of those extract anything from victims. Which is... quite good. For the public, this is my DM message, I copy it here since it contains no private data and important for the context: "We independently confirmed your narrow technical point: our doc conflates probabilistic inclusion with fee-prioritized ordering once transactions are selected. We’ll correct that, thanks for noticing. We also completed a chain-wide measurement covering last months, all 37 ZS pools, 10k swaps joined to their Kaspa carriers, and approximately 14.5 million Kaspa blocks. We found 12 genuine same-Kaspa-block sandwiches across five attackers. All 12 were net-negative for the attacker. Our data confirms that while sandwiching is real and persistently attempted on Igra, and that higher carrier feerate wins the same-block position in the cases we observed, it does not currently support your claim re “live and continuous at scale” as profitable extraction. Over the full measured window, only 12 landed chain-wide, and none was profitable. It also fully supports our previous claim re single-digit per-attempt success rates -- empirical number is ~4.5%. We're going to publish these results next week. Your public app displays 404,362 KAS of extracted MEV without a visible methodology or breakdown. To reconcile the findings, could you share: - how much of the 404,362 KAS is attributed specifically to Igra; - the split between arbitrage, liquidations and sandwiches; - the observation period and covered venues; - your definition of a completed sandwich and extracted value; - one representative Igra example that we can independently reproduce. We’re not asking for your detector or full dataset as we need enough evidence to understand whether your Igra claim measures profitable extraction, attempted execution, or mechanism-level exposure. If your Igra signal reproduces against our chain-wide dataset, a broader engagement makes sense."

English

1.3K

ꓘodingIsFun@KodinglsFun·2d

@KaspaKaha @emdin Igra's position: "In practice, on our live network, sandwich bots fail and lose money." Researcher's position: "In pure mathematical theory, the trade parameters allowed for profit if a bot was smarter." Reality is winning for now, but will that remain?

English

131

Pavel Emdin@emdin·2d

@KodinglsFun @KaspaKaha Our position is on Ethereum and alike, attacker who sees the victim and pays lands it deterministically. On Igra, the same attacker achieves the required co-inclusion ~6% of the time which is measured, regardless of the size and trade params.

English