8bit0 🏴‍☠️

onchain pirate

[8bit0 🏴‍☠️ @eightbitsim]

https://t.co/abe633uuzU

[8bit0 🏴‍☠️ @eightbitsim]

Это ещё не конец.

[8bit0 🏴‍☠️ @eightbitsim]

This is one problem, not two. A wallet gets access to a signal and consistently lands ahead of other buyers reacting to it, through a particular builder. Those buyers aren’t the trigger, they’re the ones left paying more.

Across 1,200 rounds, 1,880 user buys landed behind this wallet in the same block and another 1,316 in the next slot. The wallet then sold roughly 3 slots later. Buyers paid a median 4.3% more per token than they would have if the wallet hadn’t traded in those slots 216 SOL in additional costs.

After a large buy lands, other buyers follow in the next slot 30 to 38% of the time. This wallet does so in under 1% of cases. It doesn’t need to predict the future. It needs privileged access to the signal and priority landing ahead of everyone reacting to it.

The result is simple the wallet gets better execution, while ordinary @solana users get worse prices. Is that acceptable?

[8bit0 🏴‍☠️ @eightbitsim]

@MostlyData_ @solana yes, and respect back to you. good to have this argued in public with data on both sides.

[8bit0 🏴‍☠️ @eightbitsim]

1/ @MostlyData_ respect your work a lot, nothing personal. “wide sandwich” has no proper definition, so let’s use your term, momentum trading. only question left, is it toxic for people trading on @solana?

[Umberto | Mostly Data 🧙‍♂️ @MostlyData_]

1/8 There are rumors of a new kind of sandwich on Solana.

However, a sandwich needs one thing: knowing your trade before it lands.

We looked at the on-chain data and found no sign of that. What we see is a momentum bet.

👇🧵

[8bit0 🏴‍☠️ @eightbitsim]

it can't, and it doesn't need to. take the most public signal there is: a kolscan top-50 wallet buying 0.7+ SOL. every copy bot on the chain sees that tx. this bot buys behind it in the same block 22% of the time on that builder vs ~5% for any other big buy, and lands within 2 txs 11% of the time there vs 2% on jito/BAM. same signal, different position.

[8bit0 🏴‍☠️ @eightbitsim]

@MostlyData_ your 14% matches mine (14.3%). in my sample those rounds lose 6.1 SOL in total. the other 86% make 44.0 SOL, paid by the users who buy behind it. a 14% miss rate is the cost of the bet, and the users cover it 7x over

[8bit0 🏴‍☠️ @eightbitsim]

just to clarify when nobody buys in its window, it loses money. it only makes money when users buy behind it, and the more they buy, the more it makes. nobody's claiming it knows future flow. it gets right behind a big buy, holds ~3 slots, and sells into whoever follows, seeing tpu flow + builder priority enables this.

[8bit0 🏴‍☠️ @eightbitsim]

apologies for the late reply, was gathering some info last night. the data points to a mixed copy-KOL setup (as mev teams colloquially call it, I'm sure you know it). funny thing, a lot of the people who RT'd your post run it themselves, so take some of the cheering with a pinch of salt

you can ask them for data too. I have limited sources, you have unlimited + @SolanaFndn leverage ser. if something looks odd, ask for block engine level data. and if you do get it, how do you verify it's true? kind of hard. which is the real point, push @anza_xyz to make block engines redundant, because you can't trust ANY of them. even with TEEs lol.

[8bit0 🏴‍☠️ @eightbitsim]

good point, I ran it. big buys get follow-on buying 70-77% of the time on every builder, so the flow is everywhere. the bot isn't. across 12k random slots, ordinary buyers follow a public big buy just as often on that builder as on jito/BAM (1.0x). the bot follows it 4.3x more often on that builder (oct 5-12), and 2.7x since oct 14.

[8bit0 🏴‍☠️ @eightbitsim]

Deleting the above, I crunched more data to build a more coherent argument. Linked here:

https://t.co/EjMPUjDsZH

[8bit0 🏴‍☠️ @eightbitsim]

1/ @MostlyData_ respect your work a lot, nothing personal. “wide sandwich” has no proper definition, so let’s use your term, momentum trading. only question left, is it toxic for people trading on @solana?

[8bit0 🏴‍☠️ @eightbitsim]

5/ there’s no public mempool, so before a tx lands only the leader’s TPU and the builder can see it. 1-2 positions is under a ms of block time, way too fast to watch it land and react. e.g. slot 457634942, trigger paid 0.00008 SOL fee, bot paid 0.0315 and sat right behind it

https://t.co/VeeJvYZKpi

https://t.co/l4cNyJoABZ

[8bit0 🏴‍☠️ @eightbitsim]

4/ and it’s just one builder. normal users land right behind a same-token buy ~20% of the time everywhere. the bot: 78% on that builder, 36% on jito/BAM, 19% elsewhere. all 22 wallets, and it pays ~11× more fee on those slots

[8bit0 🏴‍☠️ @eightbitsim]

@concernedlion_ @MostlyData_

[8bit0 🏴‍☠️ @eightbitsim]

@Bruce_allinsol @coinbase Thanks comrade 🤝

[8bit0 🏴‍☠️ @eightbitsim]

1/ @coinbase runs 6 Solana validators. 5 use the same block builder, 1 doesn’t. Look at the difference.

SOL taken by front-running bots per 1k leader slots, Sep 27-Oct 18:

Coinbase: 2.66
Coinbase 02: 2.67
Coinbase 03: 2.67
Coinbase 04 (4.2M SOL): 2.20
Coinbase 05, not on the builder: 0.42

https://t.co/6TfMVY5ycm

[8bit0 🏴‍☠️ @eightbitsim]

https://t.co/abe633uuzU

[8bit0 🏴‍☠️ @eightbitsim]

4/ So what are your criteria for choosing a validator?

Grayscale’s filing says the Trust “has no right to direct the Staking Provider in the conduct of validation activities.”

Do your criteria include making sure it doesn’t profit from toxic flow at your holders’ expense?

cc: @CoinbaseInsto @lewisjhan @Duncanmoir @andres21co @PeterMintzberg @Krista_Lynch_ @21shares @FranklnTempletn

[8bit0 🏴‍☠️ @eightbitsim]

3/ Questions for the ETPs staking SOL with coinbase

@FTDA_US’s filing names coinbase as staking provider. @21shares_us lists it as primary. @Grayscale, is it one of yours?

Are you aware of this? Are you profiting from it?

If so, we’d hope it reaches your holders. 21Shares’ own prospectus says block rewards and transaction fees “will not accrete to the Trust.” Franklin’s Staking Rewards cover consideration “in exchange for using… the Fund’s Solana.” Builder-side revenue isn’t that.