Indie vs Institution — The MEV Competition Reality

When You Realize You're Not Racing the Person Next to You

I keep imagining MEV like a NASCAR race. Cars on the same track, same rulebook, the faster driver wins.

That mental model is broken.

The race I thought I was in — a few hundred people writing arbitrage bots in spare bedrooms — turns out to share a track with teams that have their own pit crews, telemetry contracts with the track itself, and engineers earning seven-figure salaries to shave a tenth of a second off lap times. The trophy doesn't go to the best driver. It goes to the team that owns enough of the race that the other drivers are racing against pit crews, not against the same kind of person.

I want to write down what I'm seeing now, because the shape of this game matters more than any one technical detail in my bot. Every assumption I make about strategy starts from the answer to one question: who am I actually competing against?

The Infrastructure Gap Isn't a Number — It's a Wall

The first thing that surprises me is how much running infrastructure costs at the top end. Dwellir, which sells RPC services and therefore has a financial interest in talking up infrastructure, still publishes numbers that line up with what I see elsewhere. Their tiering goes from free public RPCs (functionally useless for arbitrage) up through self-hosted nodes at $750–$2,800 per month, and notes that professional MEV operations run "five to six figures monthly on infrastructure before any trading capital," according to Dwellir's MEV infrastructure breakdown.

Five to six figures a month. Before they buy a single dollar of inventory.

For context, that's the cost of a fleet of food trucks running 24/7 — except you can't see the food trucks, and they only sell their product to one customer (the chain), at whatever price the chain decides. That's not a barrier to entry I can climb with hustle. It's a different category of business.

I run a paid RPC subscription and a handful of adjacent services. That's where I am: a corner-store budget operating in a market where the corner-store budget is a rounding error on the next tier's electricity bill. This isn't a complaint, it's just calibration. I know what I'm holding.

Latency Isn't About Code. It's About Geography.

I used to think latency was a code problem. Tighter loops, less overhead, fewer allocations. After spending weeks tuning my Rust pipeline, I now know that's only the surface layer.

The real latency story is geographic. The same Dwellir analysis breaks down arbitrage capture rates by total latency: bots responding in under 30ms capture an estimated 80–90% of opportunities, while bots in the 100–200ms range (where most retail setups live) capture only 20–40%, and anything over 200ms falls below 10% capture rate, per Dwellir's figures. The most striking detail: one quantitative trading operation found "400ms of node latency was costing them 40% of potential arbitrage captures."

Forty percent of opportunities, vanished into wire time.

The fix wasn't a better algorithm. It was moving servers physically closer to validators. Co-location, the article explains, takes latency from "50ms to sub-10ms" — roughly a 5x improvement from physical proximity alone.

Think about that. The single biggest performance lever isn't in my code, my language choice, or my data structures. It's whether my server is in the same building as the validator I care about. I can write the most beautifully optimized arbitrage loop in the world; if I'm in a consumer data center on the wrong coast, I'm racing against people who put their machines in the same rack as the people validating the blocks.

This is the part that took me a while to accept. The competitive frontier is real estate.

Who Actually Controls the Game

The concentration numbers are bracing. On Ethereum, the top two block builders capture "over 90% of block auctions," and searchers pay "more than 90% of their revenue to proposers," according to ExtropY Academy's 2025 arbitrage analysis.

Read that second number again. A searcher who clears $1,000 of profit doesn't keep $1,000. They keep maybe $50. The other $950 goes to the entity sequencing the block. That's not a competitive market with vigorous price discovery. That's a toll road where the toll booth sets the price after you've already driven onto the road.

On Solana, the picture is similar in shape but different in detail. ExtropY's data shows the top three bots control roughly 60% of sandwich attack volume, and one bot — known by the address prefix "E6Y" — captured about 42% of that volume on its own during the observed period. Validators running the Jito-Solana client control over 92% of network stake, per the same source.

The concentration isn't just about money. It's about who can see what's happening before everyone else, and who gets their transactions in front. A small number of entities have those views.

For an even sharper data point: the same analysis estimates the top Solana arbitrage bot generated approximately $300,000 in daily net profit over a 30-day period. The average arbitrage transaction on Solana, meanwhile, nets about $1.58. The distribution isn't a bell curve. It's a power law where the top operator pulls in more in a day than I'd make in five years of average-transaction-level returns at the same volume.

They Don't Just Compete Better — They Own the Layer Below

This is the part that reframes everything for me.

When I first read that Sol Strategies, a publicly listed company, spent $18 million in November 2024 acquiring validator node operators across Solana and adjacent ecosystems, and then another $35 million Canadian dollars in March 2025 to buy out the operators of Laine and StakeWiz.com, per industry coverage at Futunn, I had to read it twice.

They didn't compete with the validators. They bought the validators.

This is what the institutional advantage looks like in its purest form. I optimize my searcher code; they buy the entity that decides which transactions land. I co-locate my server next to a data center; they own the data center. The competitive game I think I'm playing — write better code, find better routes — sits two layers above the game they're playing.

It's the difference between opening a coffee shop across the street from Starbucks, and being Howard Schultz. One of those is competition. The other isn't a competition at all.

The Economics Have Been Squeezed Almost Flat

Even for the operators who do compete on the searcher layer, profit margins have collapsed. The MEV team blog mteam.space notes that bribes to block builders and validators "often exceed 99.5%" of profits for sandwich and arbitrage strategies, with recent months pushing that figure above 99.99% on highly competitive short-tail strategies.

99.99%. That means a $10,000 gross profit becomes a $1 take-home for the searcher. The other $9,999 goes upward.

Market reporting from late 2025 suggests margins on short-tail sandwich strategies have collapsed near zero, with a significant share of operators running at breakeven or net loss. The leading short-tail bot — known as jaredfromsubway.eth — still generates millions in profit, but the same mteam.space writeup notes the team behind it pays "upwards of $1 million" in employee salaries.

So even the bot generating millions isn't a hobbyist operation. It's a fund with payroll, benefits, and engineers who could be at Google or Citadel and chose this instead. That's the floor for serious short-tail MEV. If I want to play that game, I need to be that.

I don't need to be that.

Where Individual Developers Still Have a Real Shot

Here's the part that keeps me at the keyboard.

The same mteam.space analysis draws a distinction that's saving my project. There are two MEV games, not one.

Short-tail MEV is competition to hit a known opportunity faster. Latency, infrastructure, builder relationships. This is where institutions own everything and bribes eat 99% of profits.

Long-tail MEV is competition to find opportunities that aren't yet known. Niche protocols, unusual economic mechanisms, event-based strategies, weird token transfer behaviors. The same article notes long-tail bribes typically run between 0% and 80%, and that "individual researchers targeting niche strategies face substantially lower competition."

This is the indie game studio's version of the AAA studios' market. I'm not going to outspend Rockstar to build the next Grand Theft Auto. But there's a real, sustainable category of games made by small teams that find audiences AAA studios never targeted. Stardew Valley wasn't built by an institution. It was built by one person who saw something the institutions missed.

That's the model for individuals in MEV. Not competing on speed at known opportunities, but on discovery of unknown ones. The window between "I find a new pattern" and "everyone finds it" is the window where individual developers can be profitable.

Pawel Urbanek's writeup at pawelurbanek.com describes tooling exactly aimed at this — Rust-based EVM tracing, querying signature databases with about 800,000 event signatures and 2 million method signatures, looking for patterns that institutional tools (which focus on short-tail) don't surface. The performance is workable on a developer's machine: first-run traces in about 24 seconds, cached traces around 2.6 seconds. Those aren't numbers that scare anyone with a hobbyist hardware budget.

What's instructive is what those tools look for: unknown method signatures (likely undocumented bot logic), unusual token transfers, storage changes that don't match common patterns. The hunt is qualitative. It's the opposite of latency racing. It rewards patience, weirdness-spotting, and willingness to read transactions until you see a pattern.

What I'm Actually Choosing

So when I look at my project against this backdrop, the question isn't "can I beat a Wall Street prop desk at sandwich attacks" — that question has an obvious answer, and the answer is no. The question is whether I can find places on the map where those teams aren't looking.

A few categories that keep coming up in the sources, and that I'm thinking about:

• New DEX forks before they're widely covered by institutional bots
• Small liquidity pools with unusual triangular paths, where the obvious routes don't apply
• Tokens with custom transfer mechanics that break the standard AMM assumptions other bots rely on
• Layer-2 chains where latency advantages compress
• Cross-chain arbitrage where technical overhead filters out single-chain specialists
• Niche events tied to specific protocols' economic mechanisms

Each of these is a different bet. Each requires research instead of speed. None of them require co-location.

The honest version of where I am: I don't yet have a profitable strategy in this space. I've been burning gas chasing patterns that turn out not to be patterns. But the competitive shape of the game tells me that's the right area to be burning gas. Speed-chasing is a losing bet for me; pattern-finding might not be.

What Failing Spam Tells Me About Everyone Else's Position

One detail that grounds me: on Solana, more than 50% of transactions are unsuccessful arbitrage attempts, according to Helius's introduction to Solana MEV. Separate analysis cited by TokenPost suggests MEV bots occupy approximately 40% of Solana block space while contributing only about 7% of fees.

That's a lot of failed bots. The reason they keep failing is cheap: Solana fees are low enough that you can throw spaghetti at the wall thousands of times a day for very little. The institutional bots eat the meaningful opportunities, and the long tail of individual bots burns fees in failed attempts.

I don't want to be in that statistic. The spam-and-pray model only works at the bottom because Solana hasn't priced it out yet. It's not a strategy; it's the absence of one.

The strategy I'm trying to build is closer to the opposite: fewer transactions, more thinking, more research per opportunity. If I'm going to lose to institutions on speed (and I will), the only sane move is to refuse the speed game and find a different one.

What Competing With Institutions Actually Means

Here's the reframe that's helping me sleep.

I'm not competing with institutions. I'm sharing a pool with them. The pool is huge — total MEV extracted across chains since 2020 exceeds $7.2 billion per Dwellir's compiled figures — and the institutions own the deep, fast lanes. They're swimming in the lap pool with timers and coaches.

But the pool has shallow ends, weird corners, unmonitored stretches near the diving boards. Those are still part of the pool. The water is still real. The competition isn't who can swim fastest in the lap pool; it's whether I can find a productive stretch of water where I'm not racing the swim team.

The realistic outcome for an individual developer at my stage isn't beating the institutions. It's discovering that the question "am I beating them?" was the wrong frame. The right frame is whether the corner I've found is worth the time I'm spending in it, and whether what I learn here compounds into something the institutions can't easily replicate — which is taste, patience, and the willingness to read transactions that don't matter to anyone with payroll to make.

What This Means Going Forward

If you're an individual developer looking at this market and trying to decide whether it's worth your time, the honest answer is that it depends on which game you think you're entering. The race for short-tail MEV is over for individuals — not because individuals lack skill, but because the game has been won at the infrastructure layer by entities that, in some cases, literally bought the rails. Walking into that race expecting to outrun a NASCAR team in a rented sedan is not a strategy; it's a learning experience that costs you money.

The long-tail game is different. It's slower. It rewards the kind of work that doesn't scale with capital — reading transactions, spotting weirdness, building research tooling, learning protocols deeply enough to find their soft spots. That's a game where being an individual is actually a structural advantage, because nobody at an institution is paid to sit and notice the strange thing that happened in one obscure pool last Tuesday. They're paid to grind the lap pool. The corners are mine if I'm willing to spend the time.

What I don't know yet is whether the corners I'm finding will eventually compound into anything sustainable. That's the open question I'm carrying into the next stretch of work.

Key Takeaways

• Infrastructure cost is a wall, not a hill. Professional MEV operations spend five to six figures monthly on infrastructure before any trading capital, per Dwellir. Retail setups don't approach this and shouldn't try.
• Latency is geography, not code. Co-location moves latency from roughly 50ms to sub-10ms — a structural advantage individuals can't match by tuning algorithms.
• The top of MEV is extreme concentration. Top two Ethereum builders take 90%+ of block auctions; top three Solana bots control ~60% of sandwich volume. Searcher bribes can exceed 99.5% of profits on short-tail strategies.
• Some institutions own the layer below searchers. Sol Strategies has spent tens of millions buying validators outright. This is not searcher-vs-searcher competition; it's vertical integration.
• Long-tail MEV is a real, distinct game. Competition shifts from execution speed to discovery. Individual researchers focused on niche protocols, unusual mechanics, and undiscovered patterns face substantially lower competition and far less brutal bribe economics.