At least 1 in 5 Solana swaps fail, costing $11.3M in fees

01 · One secondOne block, 130 failed tries

Press swap in a crypto wallet app and the trade usually goes through. The new coins show up a moment later. The fee is so small that most people never look at it.

On Solana, at least 1 in 5 swap attempts fail, and the sender still pays the fee. Most of those failures come from bots, and many of the bots are racing each other.

The race is easy to see inside one block, the batch of trades and payments the network adds two or three times a second. On 15 September, one Solana block held 130 failed attempts to trade through the same pool. A pool is a pot of two coins that anyone can swap against. This one held WIF, a memecoin, and USDC, a dollar coin, on the exchange Orca. The attempts came from 36 senders, and the busiest of them tried 32 times. Nearly all were selling WIF into that pool, the last step of a trade meant to end with more money than it started with.

One loop through that pool did go through. A loop is a chain of swaps that ends in the coin it started with. This one swapped about $80 of USDC into SOL, Solana's own coin, then the SOL into WIF and the WIF back into USDC, and came out $1.15 ahead. The bot then paid a network fee of 8 cents and a $1 tip to a service that rushes trades into blocks, and kept 7 cents. The failed attempts paid 48 cents in fees between them and got nothing.

Every try in that block followed the same plan. The winner's three swaps all went through. For the other 130, the trade came too late, so the whole try was cancelled and only its fee was paid.

That was one block. We went through every swap Bitquery records on Solana in the 11 months to early October: 11 billion swap attempts, of which at least 2.4 billion failed. Most of the failures came from bots, and the largest group of them chased the same price gaps, the trade known as arbitrage. They paid $11.3 million in fees for trades that never happened. The arbitrage loops we could price made $50.8 million before fees and tips.

02 · The raceWhy do Solana swaps fail?

A swap is a trade of one coin for another. On Solana, most swaps go through pools on exchanges such as Raydium, Orca, Meteora and PumpSwap. Every trade moves a pool's price a little, so for a moment the same coin can cost less in one pool than in another.

Arbitrage is buying where a coin is cheap and selling where it is dear. On Solana a bot can do both inside one transaction, a bundle of steps the network runs all at once or not at all. Bots read new blocks from fast data feeds, such as Solana gRPC streams, and react in a fraction of a second. Many of them build their loops with Jupiter, an aggregator that finds a route across pools, and set a minimum return. Others call the pools from their own programs and check the result on their own. Either way, if the loop would end with less than it started with, the whole transaction is undone.

That undo is a failed swap. Usually, by the time the network ran the trade, someone else had already taken the gap. Nothing moves except the fee. Solana charges a base fee of 5,000 lamports for each signature, plus any priority fee the sender offers to get in faster. A lamport is a billionth of a SOL. A failed swap paid about half a cent on average.

A miss is that cheap, so bots send many copies of the same trade, often from several wallets and through several routes. The first to land takes the gap and the rest fail. Winners also pay tips to services that deliver trades to validators, the machines that build each block. When the tip travels inside the trade, a failed swap pays no tip. A winning one does.

03 · The countWhat is Solana's transaction failure rate?

Over the whole period, Bitquery's Solana DEX data holds 11 billion swap attempts, each a transaction with at least one swap in it. 2.4 billion of them failed, 22% of the total. The share moved from month to month: among full months it was lowest in May and highest in March, when 28% of attempts failed.

Counted swap by swap instead of attempt by attempt, the share is 29%: a trade routed through an aggregator such as Jupiter shows up once for the route and again for each pool it touched, and 58% of failed attempts used an aggregator, against 11% of those that went through. We count attempts here.

These counts miss one kind of failure. When the pool itself turns a trade down, for example because the price moved past the limit the sender set, no swap is recorded. We read 30 random blocks in full on a public Solana node, a server anyone can ask for the chain's own history: 593 of 1,460 failed swap attempts were missing from our records, 41%. Most were trades on PumpSwap, where Pump.fun coins trade once they leave their launch curve. Counting them, the real share of failed attempts is nearer 3 in 10; the sample puts it between 27% and 39%. Our other counts of failures come from the records, so each one is a floor.

The missing failures fall on the smaller wallets. In the blocks we read, most of the failed tries by wallets with under 100 swaps that day were missing from our records, but few of those by the busiest bots, the wallets with 10,000 or more.

04 · Your swapsWhy did my Solana swap fail?

Usually because the price moved before the trade landed. Every swap carries a slippage limit, the most the price may move against the trader before the swap is cancelled, and wallet apps show the failure as "slippage tolerance exceeded". That happens to people far less often than to bots, but more often than our records show. For one day in each month, we sorted every wallet that sent a swap by how many swaps it tried that day. In our records, wallets that sent fewer than 10 saw 0.9% of them fail. Wallets that sent 10,000 or more saw 62% fail, and those wallets, about 190 on an average day, sent 83% of the recorded failures on those days.

A small wallet's swap tends to fail inside the pool, though, and those are the failures our records miss. So we took 119 random wallets with one to nine recorded swaps on a sample day in March, June or September and listed everything each one signed that day on a public node. 38 of their 342 swap attempts failed, 11%, where our records show none. Most were trades on PumpSwap, and most of those stopped at the trader's own slippage limit, usually on a memecoin sale. The failures bunched up: 10 wallets had all of them, often trying the same sale again and again. The rest got every swap through that day.

05 · The sendersWho sends the failed swaps?

More than 4 million wallets sent at least one failed swap, but the failures were not spread among them. The 100 busiest sent 64% of the recorded failures. The busiest one alone sent 70 million of them and failed twice for every swap that went through.

Of the 20 busiest senders, 15 were loop bots like the ones in section 01. Most of the others traded one way between SOL and dollar coins, a pattern that fits bots trading against prices on a CEX, a central exchange where coins trade off the chain.

06 · The attemptsWhat were the failed swaps trying to do?

Most recorded failures were loops. At least 53% of them started and ended in the same coin, the shape of an arbitrage try. That is a floor: when a trade fails partway, the swaps after the failure point are never recorded, so some loops look cut short.

The most common error, in half of the recorded failures, was code 0x1771 (6001 in decimal), which wallet apps show as "slippage tolerance exceeded". In Jupiter's program it means the route would have returned less than the minimum the sender set, and on our sample days 94% of those errors came from trades that started with a Jupiter route. For a loop, that means the price gap the bot was chasing had shrunk or closed before its trade ran.

Counted by swap, BisonFi had the highest failure rate of the 10 busiest exchanges and Pump.fun the lowest. The failures the records miss fall mostly on PumpSwap, so its true rate is higher than the chart shows.

07 · The clockWhen do Solana swaps fail?

The failure rate changes through the day: it was highest in the early afternoon, UTC, when 27% of swap attempts failed, and lowest in the morning.

08 · The billHow much do failed swaps cost?

Failed swaps paid $11.3 million in network fees, priced at each minute's SOL price. That was 9% of all the fees that swap attempts paid; the rest came from swaps that went through.

On average a failed swap paid less than half the fee of one that went through. The fees go to the validators that build the blocks, and part of each base fee is burned. Bots pay priority fees to go first, though what a priority fee buys can be very little.

These totals are a floor. The trades a pool turned down are missing, and in the 30 blocks we read they paid 24% of all the fees that failed swaps paid. Bots also send trades that stop inside their own program before any swap runs, and those never show up as swaps. One loop bot, EDZQ…fYr2, had 10,182 failed attempts on 15 September by a public Solana node's count, and fewer than half of them reached a pool. Most of the rest gave up first, with an error its program names NoEdge: no path that pays.

The bill peaked in September 2026, with 14,869 SOL, about $1.6 million.

09 · The prizeHow much do Solana arbitrage bots make?

We counted 276 million loops that went through. Three in five were built with aggregators such as Jupiter; the rest went to the pools without an aggregator. The ones we could price made $50.8 million before costs.

Costs took a fifth of that. The loops' own network fees came to $2.5 million, and the failed swaps of the wallets that ran loops cost $8.5 million. That left $39.9 million before tips.

A few wallets took most of it. The 10 wallets that made the most took 41% of the profit before fees, and the 100 that made the most took 89%. Everyone else who ran a loop, more than 300,000 wallets, shared what was left. At the other end, more than 800 wallets that ran at least 100 loops lost money once their fees and failures are counted. The busiest loop wallet of all, HuTs…pXiP, ran 8.5 million loops and ended $3.0 million down before fees.

Some single loops are huge. The largest we found turned less than a dollar of USDC into $0.7 million on 1 May, through pools that had priced a little-known token far out of line, and a public Solana node shows the same gain in the bot's account. The average priced loop made 19 cents before fees.

Fees can swallow the gain. On 15 September one Jupiter loop made 7.39 SOL and paid 6.00 SOL of it as its network fee. A high priority fee buys a better place in the block, so a bot that sees a big gap can bid most of it away and still come out ahead. Arbitrage pays thin margins on other chains too: on five EVM chains we counted 139 million trades that made 69 cents each. Month by month, the profit swung widely: it was highest in September and lowest in June.

The table lists the ten wallets that kept the most after their fees and failed swaps.

Tips cut the gain further. They travel as separate transfers, so they do not show in swap records. On 3 sample days in June, July and August, loops paid between 3% and 20% of their gain in tips to Jito alone, one of several services that take tips.

So who wins? The bots that win often enough to cover their misses, with ten wallets taking 41% of the profit before fees, and the validators, who collect a fee on every try, won or lost. Someone who swaps a few times a day mostly gets through. When such a trade fails, it is usually a memecoin sale that hit its own price limit.

10 · MethodHow we did this

We used Bitquery's records of DEX swaps on Solana from 30 October 2025 to 7 October 2026, for every exchange Bitquery reads, from the Solana DEX data. Successful and failed swaps are kept apart, and a swap inside a failed transaction keeps the error the network returned. The full history is also sold as files on the Bitquery Data Store.

A swap transaction is a transaction with at least one swap in it, counted once however many swaps it holds. A loop is a successful transaction whose swaps start and end in the same coin, either one aggregator route (Jupiter, OKX or DFlow) or two or more pools that the same wallet traded in a row without an aggregator. Swaps through one pool and back, and transactions where two wallets trade, are left out as the work of volume bots. Profit is the coin that came back minus the coin that went in, priced at each minute's SOL price when the loop starts in SOL, and at one dollar for USDC and USDT. Fees are the network fees each transaction paid. A wallet is the address that signed and paid the fee.

We checked the work four ways. We worked out every figure a second time with differently written queries and got the same numbers. We compared real transactions with a public Solana node: the winning loop in section 01, 40 more loops and 17 failed transactions all matched, to within the small deposit a new token account needs or a share paid to another account in the same transaction. For two bots we counted every transaction on the node for a whole day, and every one in which a swap ran was in our records. Last, we read 30 random blocks in full and the whole day of 119 random small wallets on the node. That is how we found the failures our records miss.

11 · EvidenceTransactions and wallets cited

FAQ

Why do Solana transactions fail?

Most failed swaps on Solana come from bots. Many are arbitrage bots that lost a race: a bot sends a trade that only pays if prices stay where they are; if another bot gets there first, the trade would lose money and the transaction cancels itself. Bots send many copies of the same trade, so most copies fail.

Why did my Solana swap fail?

Usually the price moved past your slippage limit before the swap landed, and the swap was cancelled. Wallets such as Phantom show this as "swap failed" or "slippage tolerance exceeded". In a sample of 119 random wallets that swap a few times a day, 11% of swap attempts failed, most of them memecoin sales on PumpSwap.

What does Jupiter error 0x1771 (6001) mean?

It is Jupiter's "slippage tolerance exceeded" error: the route would have returned less than the minimum you set, so the swap was cancelled. It was the most common error among recorded failed Solana swaps, in half of them.

Do I pay a fee if my Solana swap fails?

Yes. The network fee is charged whether the transaction succeeds or fails, because the validator still did the work. A tip paid inside the transaction is not charged when it fails, because the whole transaction is undone.

What is a Solana arbitrage bot?

A program that watches the prices of the same coin in different pools and, when they drift apart, buys in the cheaper pool and sells in the dearer one inside a single transaction. If the gap is gone by the time the transaction runs, it fails and only the fee is lost.

How much do failed transactions cost on Solana?

Failed swap transactions paid at least $11.3 million in network fees over the 11 months we studied, 9% of the fees that all swap transactions paid. Our records miss some failures, so the real bill is higher.

How much do Solana arbitrage bots make?

The loops we could price made $50.8 million before costs. After their own fees and the failed swaps of the same wallets, $39.9 million was left before tips. The 10 wallets that made the most took 41% of the profit before fees, and the top 100 took 89%.