Toward My First Landing: Reading the Solana MEV Market

Watching the Market Before I Land

I haven't landed a profitable trade yet. My current landing rate sits at 0% — every signal I have simulated either reverts on-chain or gets outpaced before submission. Before I burn another round of tips chasing the same wall, I want to step back and read the actual market I am trying to break into.

This is the seventh entry in my developer reflection arc. The pattern repeats: stack up on infrastructure, find the next blocker, fix it, run again, fail differently. To stay sane I have to know what "winning" even looks like in this market, and where the line sits between bad luck and structural impossibility.

The Scale of Solana MEV Has Become Hard to Ignore

The numbers being published about Solana MEV in 2025 are not subtle. According to a recent Solana MEV report, the network produced roughly $720 million in MEV revenue across the full year — overtaking priority fees as the largest economic component of the network. In a single quarter, Q2 2025, Solana captured $271 million versus Ethereum's $129 million in the same period, according to a 2025 H1 ecosystem report.

The volume side is bigger still. A 2026 analysis of Solana arbitrage notes the chain processed roughly $1.6 trillion in spot DEX volume in 2025, with the majority routed through automated systems. According to the same analysis, cyclic arbitrage alone grew from about 2.5% of DEX volume in mid-2024 to over 40% by late 2025 — more than a sixteen-fold expansion in a year and a half.

I read those numbers the way a minor-league pitcher reads the wins-above-replacement of an MLB ace: not as inspiration, but as a calibration tool. Whatever I build has to land in the same market where automated systems are already capturing 40% of routing.

What "Winning" Actually Looks Like

The honest way to read the market is to look at top bots, not average numbers. Average per-arbitrage profit on Solana is about $1.58 across more than 90 million successful arbitrage transactions in a year, per a recent Solana MEV report. The top of the distribution looks very different.

An analysis of cross-chain MEV markets profiled one bot — labeled E6Y — that ran through $1.6 billion in trade volume over a thirty-day window, collected 57,400 SOL in gross revenue, paid roughly 7,980 SOL in tips, and netted approximately 49,400 SOL. The same study notes E6Y alone captured about 42% of all sandwich attack volume during the measurement window. A separate Solana MEV report documents the DeezNode bot pulling 65,880 SOL in gross profit over thirty days at an 88.9% success rate on roughly 1.55 million sandwich transactions — about half of all Solana sandwich activity in that period.

Two facts jump out of those case studies. First, the very top of the market is dominated by a small number of entities; per the cross-chain analysis, the top three bots together control over 60% of the Solana MEV market. Second, the operating economics are nothing like a hobbyist setup: the same study cites dedicated RPC infrastructure for serious competition in the range of $1,800–$3,800 per month, before counting capital, tip outlays, or engineering time.

The way I think about it: this is a market where the top resembles a hedge-fund-style operation and the bottom is one developer writing Rust at a kitchen table at 2 AM. There is almost no middle.

The Latency Stack Decides Everything

The hardest pill in MEV is that strategy doesn't actually decide much. Operational latency does.

A 2025 walkthrough of why most arbitrage bots fail before they start lays out the stack honestly. A public RPC endpoint sits roughly 200ms behind a co-located bot before a single line of strategy runs. WebSocket-style subscriptions land in the 100–300ms range. A Geyser-compatible gRPC feed drops that to sub-50ms. A node co-located with a validator runs 10–30ms. The same article reports that, holding strategy logic constant, landing rates can differ by 30 to 50 percentage points based purely on infrastructure.

Solana blocks produce roughly every 400 milliseconds. The arbitrage window for any given opportunity, per the same source, persists for 400–800ms in total before competing bots close it. That sets the brutal arithmetic: a bot at 200ms behind is effectively blind for half of every opportunity window.

It reminds me of pit stops in NASCAR. Two cars with the same engine, same tires, same fuel — but one pit crew operates at 11.2 seconds and the other at 13.9. Over 200 laps, the slower crew never sees the front of the pack again, no matter how skilled the driver. The lap times are identical; the operational floor is what decides finishes.

Or, to use a supply chain analogy: you can sell the same product as a giant retailer, but if your fulfillment is 48 hours and theirs is 12, you don't compete with them at all. You serve a different customer.

The Field Reports From Other Solo Builders

I find the published failure reports more useful than the success stories, because they describe a stage I can actually compare myself to. Two stand out.

A field report on building a Solana MEV bot, published by an independent developer, documents about 0.3 SOL of gross profit over two weeks of operation — which became net-negative after RPC subscription fees and tips paid on losing bundles. The author's bundle success rate sat under 5%. The conclusion in the post is plain: at solo-developer scale, with cloud infrastructure, the economics don't make sense.

A separate post-mortem on an unprofitable Rust bot — written by a developer who built across mainnet plus two sidechains — describes roughly 10,000 lines of Rust client code, infrastructure running about $750 per month, and a pattern the author summarizes as "few successful txs, with subsequent dozens of failures draining my capital." The structural observation in that post is the one I keep returning to: established MEV bots behave like "highly territorial predators" that squeeze margins to zero when they detect a new competitor in their lane.

Both reports converge on the same advice for anyone starting from zero: do not optimize against the established competition. The marginal latency gains you can squeeze out as a solo builder do not close a 150ms gap. Instead, the edge — if there is one — lives in finding fresh opportunities, new protocols, configuration changes that have not yet been indexed by the dominant operators.

That advice is uncomfortable because it is not a technical answer. It is a market-finding answer. I keep wanting the problem to be solvable by tuning my code; the field reports keep telling me the problem is solvable by changing which market I am competing in.

What 2024 Did to the Game

A few structural changes in 2024 reshaped the floor I am building on, and I want to log them as I currently understand them, since I will be making decisions against this version of the market.

In March 2024, Jito shut down its public mempool. The same MEV report I cited above describes the immediate effect as a measurable drop in harmful MEV practices, alongside an unintended consequence: alternative private mempools rose to fill the gap, and those private channels are inherently less transparent. Newcomers lost the easy visibility that came with public mempool data.

In May 2024, a scheduler change rolled out to validators that introduced more deterministic transaction ordering. Per the same report, this reduced the spam-arbitrage incentive structure — and revert rates dropped meaningfully from the April 2024 peak, when over three-quarters of non-vote transactions on the network were failing during peak congestion. A guide to MEV on Solana notes that, even into 2025, over half of all Solana transactions are still failed arbitrage attempts; the scheduler change moved the dial, but did not erase the spam.

Through 2024 and into 2025, a coordinated response to sandwich attacks took shape: the Jito Foundation closed its public mempool, the Solana Foundation removed sandwich-participating validators from its delegation program, and Marinade Finance blacklisted over fifty offending validators. According to the 2025 H1 ecosystem report, the combined effect reduced sandwich profitability by approximately 60–70%.

I read those changes two ways. As a builder, they mean the easy entry routes — sniffing public mempool, lazy spam-arbitrage — are closed. As an ethical participant, they tell me the ecosystem is actively defending end users from the most predatory strategies, which narrows the viable strategy space toward back-running and cyclic arbitrage rather than sandwich attacks. I don't want to spend a year building something the ecosystem is structurally trying to kill.

The Tip Auction Math I Have to Plan Around

Tip economics on Jito have settled into a recognizable shape. A public explainer on Jito mechanics reports that the tip ceiling has stabilized somewhere between 50% and 70% of expected profit. A separate market analysis puts the realistic surrender rate for competitive searchers at 50–60% of expected profit just to get bundle inclusion. And by early 2026, the same explainer notes that over 95% of active stake sits with validators running the Jito-Solana client — meaning there is effectively no escape route from Jito-style economics for anyone who wants their bundle considered.

That is the number that hurts the most. If I find a $10 opportunity, I am bidding $5 to $7 in tip to land it. After RPC and infrastructure costs and my own off-chain compute, the net is what survives.

By contrast, an analysis of cross-chain MEV markets notes that on Ethereum, searchers often pay more than 90% of their revenue to proposers. Solana is structurally friendlier than Ethereum on this axis — the searcher's take is larger — but the absolute scale of "friendly" depends entirely on whether your bundle lands at all. The competition is not a classical sealed-bid auction; it is tip-priority competition where calibration is continuous and tied to current congestion.

An ecosystem report clocks bots paying 8,584 SOL in a single hour during the launch frenzy of a high-profile memecoin in January 2025 — over a million dollars an hour, surrendered just to secure blockspace. The lesson is that the tip floor moves with the market, and during marquee events it moves into a range no solo operator can match without burning capital faster than they earn it.

There is no closed-form answer to "what should I tip." The right tip behavior shifts with congestion, with current leader, with how many other bots are chasing the same path, with whether my own simulation overestimates expected profit. I have read enough now to accept that calibration is its own discipline, not a one-time decision.

The AI Agent Layer Is Already There

One more piece of context I have to take seriously: I am not just competing with hand-written Rust bots from 2023. A 2026 article on Solana arbitrage describes the shift from human-written rule-based bots to automated agents that index opportunities continuously and adapt routing in near-real time. The growth of cyclic arbitrage from 2.5% to 40% of DEX volume in eighteen months is, per the same piece, almost entirely driven by automated systems rather than individual searchers.

The practical implication is uncomfortable. The lane I most wanted to enter — cyclic arbitrage between major pools on major DEXes — is the lane where automated systems have already saturated the search space. "By the time a human reads a chart," the article notes, the opportunity is already captured. I don't read charts in my pipeline, of course, but the meta point applies to my whole stack: anything I can recognize after the fact, in human-cognitive time, is something an indexing system already caught.

The quiet adjustment I am making is to stop thinking of "build a Solana MEV bot" as one product and start thinking of it as a series of small, structurally different products. A bot for one DEX, one routing class, one volatility regime, one type of opportunity. Anything broader walks straight into the automated layer.

Where I Actually Sit Right Now

Mapping my current state against everything above:

I am running on rented infrastructure that is not co-located with any validator. My signal pipeline does not yet hit the sub-50ms target that a Geyser-compatible feed would offer. My landing rate is, at the moment, zero — every bundle I have submitted has either reverted in simulation, missed the window, or been outbid by tips I cannot match. The economic situation matches the published field reports almost exactly: small amounts spent on infrastructure and losing tips, no offsetting wins yet.

What I am not doing — yet — is the thing both honest failure reports recommend. I have been trying to compete on the well-trodden cyclic arbitrage paths between the largest pools on the largest DEXes. That is the lane the territorial predators live in. Even if I close my latency gap by half, I still arrive in a market where the top three bots take 60% of the volume.

The honest current outlook is something like this. The operational gap between my setup and a competitive one is too wide to close by optimizing my code. I either fund the operational gap properly — colocated nodes, dedicated RPC, calibrated tipping — and join the field of serious operations, or I shift the strategy surface toward opportunities that are too new, too small, or too weird for the established bots to bother with.

Neither path is comfortable. The first one requires recurring infrastructure spend at a level where every losing week hurts more. The second one requires accepting that the published volume numbers — the $720 million market — are not the volume I am chasing. My total addressable opportunity is a fraction of that, sitting in the corners established bots ignore.

What "First Landing" Will Probably Look Like

I do not know which trade will be my first landing. What I do know, from reading the field, is what it will probably not look like.

It will probably not be the kind of cyclic arbitrage that already represents 40% of network DEX volume. It will probably not be a sandwich, because that lane is both ethically grim and being actively pushed out by the ecosystem itself. It will probably not be a path I find by staring at a chart, because by the time a human reads a chart, the opportunity is already captured.

It will more likely be something the dominant operators didn't bother to index: a new pool on a less-trafficked DEX, a configuration change in a protocol that briefly leaves a price gap, a route through a token pair that doesn't show up on the major aggregators. The first landing will probably be a small one — measured in single dollars, not thousands — and the lesson it teaches will probably be about routing or fee-tier handling rather than vindication of any grand strategy.

A field report on building a Solana MEV bot put it well: the developer journey, even when it does not produce profit, builds knowledge of transaction lifecycles, instruction decoding, fee market mechanics, and infrastructure trade-offs that is genuinely rare. I am converging to the same view by direct experience. Whatever the first landing turns out to be, I am already in a better position to read the next opportunity than I was eight months ago, when I did not understand why a transaction reverted.

That is the part I have to keep present. The 0% landing rate I am sitting on right now is not the score that matters. The score that matters is whether I can read the next market change faster than I could read the last one. That, I can build. The rest is execution.

Key Takeaways

• Solana MEV is now a roughly $720 million annual market with the top three bots capturing over 60% of it, per published market analyses. The middle barely exists.
• Strategy matters less than operational latency. Landing rates can differ by 30 to 50 percentage points purely based on infrastructure, per a 2025 walkthrough of arbitrage bot setup.
• Tip auctions on Jito routinely consume 50–60% of expected profit for competitive bundle inclusion, and over 95% of active stake now runs the Jito-Solana client — meaning there is no real escape from those economics.
• Coordinated ecosystem action in 2024 cut sandwich profitability by approximately 60–70%, narrowing the viable strategy surface toward arbitrage and back-running rather than sandwich attacks.
• Solo builders who have published honest field reports converge on the same advice: do not optimize against the established competition. Find fresh opportunity surfaces the dominant operators have not bothered to index.
• For a developer starting from a 0% landing rate, the realistic path to a first landing runs through new protocols and unusual routes, not the most-trafficked DEX paths.