Know Your Enemy: Mapping Solana's Top MEV Bots

The Reality Check I've Been Putting Off

Writing the bot has consumed my attention for months. Today I'm forcing myself to stop typing and look up. Who am I actually competing with on Solana? What does the top of the leaderboard look like? If I can't answer that, I'm sparring with shadows.

Sun Tzu's old line — know your enemy and yourself, and you can fight a hundred battles without disaster — keeps rattling around in my head. I know my own bot intimately. I know its bugs, its gaps, its small wins. What I don't know yet is the shape of the field. Today's job is to map it.

What the Solana MEV Map Actually Looks Like

The first thing to internalize is that Solana MEV is a real market, not a side hobby. According to Helius's Solana MEV Report, arbitrage bots alone executed more than 90 million successful trades over a one-year window, capturing $142.8 million in profit. Sandwich extraction over a 16-month period landed somewhere between $370 million and $500 million, per the Accelerate 2025 conference talk on the state of Solana MEV. That same talk cited annual MEV revenue on Solana of $720.1 million in 2025.

To put that in American terms: this isn't a food-truck operation. This is somewhere between a regional supermarket chain and a Fortune 500 business unit, depending on how you count. And almost all of it flows through a small set of wallets.

I think more carefully about that "small set" before continuing. According to Solana Floor's deep dive based on Flipside Crypto data, one wallet — labeled "E6Y" in the analysis — captured 42% of all sandwich activity in a 30-day window. The next two bots together took roughly another 18%. The top three controlled more than 60% of the entire sandwich market.

That's not a market with healthy competition. That's a market with three NASCAR-grade race teams and a thousand kids in soapbox-derby cars, all on the same track.

The 42% Bot: How One Wallet Eats the Sandwich Market

Let me sit with the leader for a minute. The same Solana Floor report records that the E6Y bot pushed $1.6 billion in 30-day sandwich volume. Monthly profit was around 49,400 SOL — translated by the analysis to roughly $300,000 a day, with daily bundle counts running as high as 26,000.

There's also a documented single trade where this bot pulled 8,400 SOL of profit off one memecoin transaction, according to the same source. That's a one-shot payoff bigger than most American workers' annual salaries, off a single sandwich.

Reading numbers like that produces a particular kind of vertigo. Not awe, exactly — more like the feeling of standing on a subway platform when an express train blows past. The thing is moving so fast that the air pressure alone tells you not to step closer.

The vertigo is useful, though. It corrects a fantasy I've been carrying without quite admitting it: the fantasy that with enough cleverness, I'm going to outflank the top of the market on its own turf. That's not how this works. You don't outflank the New York Yankees by being slightly better at swinging a bat in the same stadium. You either bring institutional resources or you find a different game.

DeezNode and the Validator Moat

The second giant I keep reading about is the bot running through a program address commonly labeled "vpe" — widely associated with a validator operator that goes by DeezNode. The PANews long-form report and Helius's MEV report together paint a striking picture.

Over a 30-day window from early December 2024 to early January 2025, the vpe bot executed 1.55 million sandwich transactions, averaging 51,600 per day, per Helius's data. Total revenue in that window was 65,880 SOL — roughly $13.43 million. Jito tips paid out by the bot totaled 22,760 SOL, about $4.63 million. Average revenue per sandwich came in at 0.0425 SOL, or $8.67. Annualize the run rate and you're looking at roughly $160 million per year.

But the staggering number isn't the revenue. It's the success rate: 88.9% of those 1.55 million sandwiches landed profitably, according to the same data. For comparison, the SolanaMevBot.com landscape report documents that pre-2023 arbitrage bots saw 98% of their transactions fail. The April 2024 peak across the entire network had 75.7% of non-vote transactions reverting. And here's an operator landing nearly nine of every ten attempts.

How? The PANews report fills in the picture: DeezNode operates its own validator, ranked 92nd by stake, with 811,604 SOL delegated — about $168.5 million worth. They control roughly 19.89% of Marinade's mSOL delegations. Their stake surged from around 308,000 SOL on November 13, 2024 to over 802,000 SOL by December 9 of the same year, the same source documents.

What that means in practice: this operator doesn't just see the order flow. They produce blocks. When they're the leader for a slot, they're not racing anyone — they're writing the rules for that block. The "race" is over before the starting gun for everyone else.

The American equivalent is a casino that also owns the building. Most poker players spend their careers trying to beat the table. A small number of operators own the property, set the rake, and run the cameras. The game looks similar from the outside, but the math underneath is different.

The Arms Race I Can't Win Head-On

If knowing the players was the first sobering lesson, knowing the infrastructure is the second. The RPC Fast infrastructure guide lays out roughly what tier of hardware top operators run: AMD EPYC-class CPUs, 256 to 512 GB of RAM, enterprise NVMe storage. Daily data volume around a terabyte. A dedicated MEV-grade RPC node, the same source notes, runs $1,800 to $3,800 a month.

Latency is the more brutal axis. Standard WebSocket subscriptions deliver block updates in 50 to 300 milliseconds, per the same guide. Yellowstone gRPC, which exposes validator memory directly, gets that under 50 milliseconds. Jito's ShredStream provides another 50 to 200 millisecond advantage over standard gossip propagation. Same-datacenter colocation shaves off 5 to 15 milliseconds. A LAN-local RPC connection drops to under a millisecond.

Solana's target block time is 400 milliseconds. Within that window, every fraction of a millisecond is a competitive position. The difference between a 200ms feed and a 50ms feed is the difference between catching a window and watching it close.

Worse, the public mempool that was once the equalizer is gone. In March 2024, Jito shut down its public mempool service, as documented in the RPC Fast guide, citing concerns about sandwich abuse. After that, private mempools spread — DeezNode reportedly proposed a "DeezMempool" arrangement to validators. Blind sandwich attacks, where the bot can't see the victim transaction in advance, rose from around 1% of all sandwiches to roughly 30%, per the same source. That tells you the visible flow has mostly moved into private channels.

If you wanted to design a market structure to lock out small developers, this is roughly what you'd build. High capital requirements for hardware. Direct validator relationships for flow. Private mempools that you can't see into. Sub-50ms latency requirements that demand co-location with the chain's hot path.

This is the part where I have to be honest with myself: I'm not going to win the millisecond race against operators with $168 million in staked SOL and a top-100 validator rank. Not this year. Probably not any year, with the resources I have.

Where the Cracks Are — Niches for the Rest of Us

Reading all of this could push you to quit. It almost did, for an evening. But then I started looking at the arbitrage data instead of the sandwich data, and the picture shifted.

According to Helius's data, the average successful arbitrage on Solana over a year captured just $1.58 in profit. The total was more than 90 million successful trades. The single highest arb in that window was $3.7 million. So there's an enormous long tail of small opportunities that, individually, no whale would bother chasing — they're hunting the median $8.67 sandwich at scale, not the $5 arb.

It's the difference between a Costco buying a million pencils at half a cent each and a regular shopper grabbing a single notebook at the corner store. Costco wins on scale and supplier relationships. But there are tens of millions of shoppers and only a handful of Costcos, and not every aisle in the country is built for bulk pickup.

Another crack: tip ratios. Extropy Academy's cross-chain analysis notes that arbitrage bots pay 50–60% of their profits as tips to validators, while sandwich bots pay only 5–20%. Sandwiches are a higher-margin business per dollar of profit kept, which is part of why they get the validator-side attention they do. Arbitrage is a more even playing field — validators don't have a strong reason to preference one arb bot over another the way they do with a private sandwich flow.

A third crack: long-tail tokens and event-driven windows. The Jito "mev wif me" post documents one extreme example — during a WIF token event on January 10, 2024, a single trader exchanged 86,739 SOL for 17 million WIF tokens, and the price moved from $0.14 to $3.99. An arbitrage bot called "2Fast" cleaned up $1.9 million of profit in the ensuing chaos. The tips paid across 50 blocks (about 24 seconds) totaled 2,286 SOL. These event windows aren't predictable months in advance, but they aren't owned by the top three sandwich operators either.

A fourth crack, easy to miss: success rate, not just raw speed, is what separates pros from amateurs. The 88.9% success rate of the vpe bot doesn't just come from being fast. It comes from disciplined trade selection — knowing when not to fire. A solo bot that fires too often, even on a fast feed, ends up paying transaction fees to lose. That's the trap most newcomers fall into. The discipline to wait, simulate, and reject ninety-nine opportunities for the right one is a skill that doesn't require institutional infrastructure.

How Bot Behavior Distorts the Network Itself

One more piece of context I keep returning to: this competition isn't a sealed-off side game. It bleeds into network behavior that affects every Solana user.

The SolanaMevBot.com landscape report notes that at the April 2024 peak, 75.7% of non-vote transactions on Solana reverted — the network was basically a stadium full of bots spamming attempts and failing. Pre-2023 arbitrage bots, per the same source, consumed about 60% of block compute while failing 98% of the time. That's catastrophically inefficient, and it pushed the Solana team to ship the Agave 1.18 scheduler update in May 2024, which the same report credits with meaningfully reducing spam failures.

The TrumpCoin launch on January 20, 2025 produced another striking moment: per Helius's data, bots paid 8,584 SOL in tips in a single hour — about $1.5 million in less than an hour, just for the privilege of being included.

That last number explains a lot of what I see in the validator economics. The Accelerate 2025 talk referenced Solana's total fee plus MEV annual revenue at about $1.4 billion. Validators are not running their stacks for fun; they're running them because the tip economy underwrites a meaningful slice of their yield. That's also why private arrangements between validators and top bots aren't going away. The incentive structure favors them.

For me, this means I have to think of network-side effects as a constraint on my own design, not just as background color. If half the network's compute is being used to spam-and-fail at high volume, my own simulation accuracy and pre-flight checks aren't a nice-to-have. They're the entire game.

What This Changes for My Own Build

So what's the takeaway when I sit back down at the keyboard tomorrow morning?

First, I'm crossing one approach off the list permanently: I'm not building a generalized sandwich bot. The sandwich market is owned. Three operators with validator-grade infrastructure and private mempool access take more than 60% of the volume. Anything I build in that lane gets out-raced before my transaction lands at the leader.

Second, I'm leaning further into arbitrage on the long tail. A $1.58 average means there are millions of micro-opportunities flowing past every day. Even capturing a small slice of that is real money for a solo operator. The bar isn't "beat the top arb bot to the once-a-year $3.7M opportunity." The bar is "consistently catch enough $5 to $50 opportunities to cover infrastructure and keep iterating."

Third, I'm taking the success-rate lesson seriously. Discipline beats speed at my tier. If I can build a system that fires only on high-confidence opportunities, with realistic simulation of slippage, fees, and adversarial reordering — and just doesn't fire otherwise — I think there's a real path. Not to becoming the next E6Y. But to running a profitable small operation that learns something new every week.

Fourth, I'm staying realistic about the validator moat. I'm not going to run a validator. I'm not going to outbid an operator with $168 million in stake for slot priority. What I can do is study how top operators behave in public on-chain data, learn from their patterns, and avoid stepping into their direct line of sight when I can. The fight I want is the one they aren't paying attention to.

Implications for the Solo MEV Developer

Zooming out, the broader picture is that Solana MEV has matured into something resembling institutional finance — concentrated, infrastructure-heavy, and increasingly gated by private relationships. The romantic image of a lone dev outracing the field with elegant code is mostly fiction at the top of the market. The real fight at that tier is about capital, colocation, and validator alignment.

But institutional finance, in the U.S. analogy, didn't kill smaller participants. It just changed what they can profitably do. Independent options traders, prop shops, and quantitative funds all coexist with Wall Street giants because they hunt different prey. A solo MEV builder occupies an analogous space on Solana: not racing the big sandwich operators, but feeding on the long tail of opportunities the giants don't bother chasing.

The next quarter for me is about building enough discipline and observability into my own stack that I can identify which corners of the market are genuinely accessible — and ignoring the rest. That's the field I can actually fight on. Today's job was to admit that out loud.

Key Takeaways

• The Solana MEV market is highly concentrated: per Solana Floor and Flipside data, one bot (E6Y) captures 42% of sandwich activity, and the top three control more than 60%.
• The dominant sandwich operator (the vpe bot, associated with DeezNode) runs its own validator with 811,604 SOL staked — institutional-scale capital, not garage-developer stuff, per the PANews report.
• Jito's public mempool was shut down in March 2024; private mempool access and validator relationships now determine who sees the best flow, per the RPC Fast infrastructure guide.
• Infrastructure is expensive: $1,800–$3,800 per month for dedicated MEV-grade RPC nodes, and sub-50ms latency is table stakes at the top, per the same source.
• The arbitrage long tail remains accessible: with more than 90 million trades averaging $1.58 each, per Helius's report, there's room outside the top operators' focus — but only for bots disciplined enough to fire selectively and absorb the lessons of high-success-rate operators.