Total for 1 October 2024–30 September 2026. Transactions are not a count of trades; company funding of these fees is not verified.
01 / September 2026 example
What would other chains charge?
September’s two V2 exchange contracts provide a monthly example for comparing the same activity across five chains.
September fees · two V2 exchanges
$861,006
USD value using daily POL prices
Contract calls
41.7m
Transactions sent to the two V2 exchanges
Average fee
2.06¢ / call
Includes successful and failed calls
Increase the activity
1× is September’s V2 total
1× September10× September
41.7m callssame work per call
21 matches against maker orders, across eight Polygon transactions. This sample is from 3 October 2026, separate from the September total. Each chain is priced for those same matches. A maker match means a match against one maker’s order; it is not necessarily a unique trade.
Estimated fees · USDNetwork demand¹ 0–120% · marker at 100%
$0$750k$1.5m
PolygonReference
$845,0452.03¢ per equivalent call
8.6%of gas budget
106.7m gas/s budget
Arbitrum
$1,312,4863.15¢ per equivalent call
91.5%of pricing target
10m gas/s target
Base
$388,7200.93¢ per equivalent call
4.6%of gas budget
22.9% of pricing target
HyperEVM
$211,0080.51¢ per equivalent call
261.3%of gas budget
»
Includes 60-second blocks
SolanaSample average²
$98,3030.24¢ per equivalent call
1.3%of compute budget
Estimated compute use
¹ Average demand before other apps. Solid markers: network budgets. Dashed markers: pricing targets, which affect fees and are not hard limits. Bars stop at 120%; » means more. Solana compute use is estimated.
² Solana uses an average modeled fee from 64 sampled September transactions. Actual fees may differ. Calculation.
Polygon’s fees were recorded on-chain. The other four are estimates, priced for the same 21 matches on 3 October. Method and assumptions.
Polymarket · Blockchain fee researchPrices: 3 October 2026
Detailed fees and network budgets
PolygonRecorded gas, new price
$845,045
estimated monthly fee
2.03¢per contract call
Network demand¹
8.6%of gas budget
»
106.7m gas/s · block budget
ArbitrumSame gas as Polygon
$1,312,486
estimated monthly fee
3.15¢per contract call
Network demand¹
91.5%of pricing target
»
10m gas/s · pricing target
BaseSame gas as Polygon
$388,720
estimated monthly fee
0.93¢per contract call
Network demand¹
4.6%of gas budget
»
22.9% of pricing target
HyperEVMFast + slow blocks
$211,008
estimated monthly fee
0.51¢per contract call
Network demand¹
261.3%of gas budget
»
3.5m gas/s · includes slow lane
Solana64-transaction average
$98,303
estimated monthly fee²
0.24¢per contract call
Network demand¹
1.3%of compute budget
»
100m CU/block · estimated use
¹ Average demand as a share of each network’s gas or compute budget, before other apps. Arbitrum and Base pricing targets affect fees; they are not execution limits. Base’s gas budget is 200m gas/s and its pricing target is 40m gas/s. Bars stop at 120%; » means demand goes further.
Pricing targetNetwork budget
Solana’s estimated monthly fee is about $98,300, compared with $845,000 on Polygon. HyperEVM’s $211,000 estimate would require 2.6 times its combined gas budget.
How the monthly estimates were calculated
The four EVM estimates reuse September’s 23,708,219,024,166 gas. We assume the same activity would use the same gas on each EVM chain. Arbitrum and Base also include a data-fee estimate from eight sample transactions. The Polygon estimate of $845,045 uses one gas-price quote and one token price from 3 October; the recorded $861,006 uses the fees actually paid and each day’s token price in September.
HyperEVM’s combined gas budget includes blocks produced every 60 seconds. We have not subtracted other apps’ demand or measured peaks within each day. September’s busiest daily average was 10.8% higher than its monthly average.
Solana: average-based monthly estimate. The 64 sampled September transactions contain 98 maker matches. At two makers per Solana transaction, the model uses 75 transactions and 248 charged signatures. With 5,000 lamports per signature, 250,000 compute units per transaction and a priority price of 1,000 micro-lamports per unit, their total fee is 0.00125875 SOL. Divided by 64, that is 0.00001966796875 SOL per equivalent contract call, or $0.00235764 at the 3 October SOL price. Multiplying by September’s 41,695,485 calls gives about $98,303.
The monthly Solana estimate applies this successful-transaction average to all V2 calls. Compute, transaction packing and native authorization are assumptions. Estimated network use includes execution and known signature costs; other scheduler overhead and shared-account limits can reduce capacity. Download the calculation · View the 64-transaction sample
Detailed September costs and capacity
EXACT SEPTEMBER V2 WORKLOAD
September’s estimated fees across five chains.
41,695,485 V2-destination calls used 23,708,219,024,166 gas, including 8,949 failed calls. Actual fees: 8,451,153.8769 POL, valued at $861,005.76 using September's daily prices.
Chain / budget measure
Monthly fee quote
Native token quote
Average budget use
Busiest daily average
Measured gas, estimated fees on other chains. These estimates use the full month’s measured gas and prices from 3 October. Arbitrum and Base also include an estimated data fee: the eight-case average per call, multiplied by the month’s exact call count. We assume moving the workload would not change its gas use. The count includes every call to the two V2 contracts, including calls that are not decoded settlements. Solana uses the 64-transaction average described above.
The broader dataset needs more capacity. The V2 calls above used 70.1% of the gas in the broader contract dataset, which used 33,812,715,107,019 gas in September. If gas use per transaction stayed the same, 10× that broader workload would use 122.3% of Polygon’s nominal monthly capacity. Its busiest daily average would reach 137.9%, before allowing for other apps.
These percentages compare demand with the whole network’s stated capacity, before other apps. HyperEVM includes its slower 60-second lane. Solana’s compute demand is estimated from the 64-transaction average; its busiest daily average uses the highest daily call count. Peaks within each day are not measured.
A separate check on Polygon’s fee settings: five block headers collected on 3 October report a target of 24m gas per block, or 16m gas/sec at the observed block rate. September’s V2 workload would use 57.2% of that target; the broader contract dataset would use 81.5%. At 10× the workload, those figures become 5.72× and 8.15× the target. This is a setting reported by those block producers. It is not a universal network target, and we have not verified that it applied in September. View the recorded target
Over two years, more than a billion transactions involving Polymarket’s contracts ran up 180.14 million POL in blockchain fees. At each day’s token price, that adds up to $18.12 million. The bill averages just 1.79 cents per transaction. A small charge, repeated often enough, becomes a large expense.
To see what that money paid for, take September 2026. Calls to the two V2 exchanges numbered 41.7 million and used 23.7 trillion gas, the measure of computation on Polygon. Their recorded fees were worth $861,006. A call is not a trade: some calls fail, while one settlement can match several orders.
Price that month’s activity using 3 October quotes and the comparison changes. Polygon comes to about $845,000, Base $389,000, HyperEVM $211,000 and Arbitrum $1.31 million. Solana comes to about $98,300, using an average modeled fee from 64 sampled September transactions.
But HyperEVM’s lower quote would require 2.6 times its combined gas budget. At ten times September’s activity, Base would use 45.7% of its gas budget but exceed its pricing target, which can push fees up. Solana’s estimated compute demand would reach 12.9% of its whole-network budget. The fee and the amount of work a chain can carry have to be read together.
The fee records do not identify how much Polymarket itself funded. Cross-chain figures are estimates at fixed prices. Capacity uses monthly averages before other apps; Solana’s fee and compute assumptions may differ from an actual deployment.
02 / What did the transactions do?
One transaction can match several orders.
Counting transactions alone can make chains look cheaper than they are. A single Polygon transaction can match several maker orders; a Solana implementation may split those matches across more transactions.
The eight Polygon transactions below contain 21 matches against maker orders. The Solana model processes the same matches in 13 transactions.
This 3 October sample illustrates the work inside a transaction. The monthly Solana estimate above uses a larger sample of 64 September transactions.
Batch 01
Batch 02
Batch 03
Batch 04
Batch 05
Batch 06
Batch 07
Batch 08
Each green mark is one match against a maker order. 21 matches across eight recorded Polygon transactions.
Eight observed batches · 21 maker matches · 5,359,204 gas
Chain / evidenceUSD fee per maker matchNative token feeWork equivalent
Measured: Polygon fees from transaction receipts. Repriced: Arbitrum and Base estimates use Polygon’s gas consumption and each destination’s data-fee oracle. Modeled: HyperEVM execution and a proposed Solana implementation.
Fees are quoted at 3 October 2026 prices. These eight transactions are a case study; they are not a representative average of all Polymarket activity.
TOKEN PRICES
How the model counts transactions and calculates fees
One batch is not one trade.
The observed maker counts are 1, 2, 1, 1, 2, 11, 1, 2. That is 21 maker matches in eight Polygon transactions. Aggregate taker events are not counted as extra matches.
The main Solana v0 model allows two makers per transaction. It would need 13 native transactions and 47 charged signatures for the same 21 matches. This is an estimate of how the matches could fit into transactions, not a measurement of a deployed port.
How fees are calculated
EVM: execution gas × gas price + destination data fee. Solana: 5,000 lamports per charged signature + requested compute × priority price.
The default Solana model requests 250,000 compute units (CU) per native transaction and pays a priority price of 1,000 micro-lamports per CU. That compute budget is an assumption. Partial fills, signature handling and atomic execution still need to be implemented and tested.
What a larger sample changes
A separate sample of 64 September batches, selected using a fixed seed, contained 98 matches. With the same two-maker packing rule, it would need 75 Solana transactions: 1.172 per source batch, compared with 1.625 in the eight-transaction sample. The ratio changes with the mix of work.
Solana v1 permits larger transactions. The advanced controls let you test that format while keeping the execution work unchanged. We have not verified how the transactions would serialize or tested a Polymarket port.
03 / Two years of fees
Two years across 34 contracts.
1 October 2024 – 30 September 2026 · UTC
This history covers 34 contracts, more than the two V2 contracts in the September comparison. The datasets contain overlapping transactions, so their totals must not be added.
180,140,563.575084492425268417 POL in native fees; $18,119,619.03 valued with daily token prices.
We count transactions sent to any of the 34 listed contracts, or with saved logs from those contracts. We have not verified whether Polymarket funded the fees. Download the exact monthly ledger
Earlier sample: 86,605 receipts
The earlier sample contains 86,605 unique transaction receipts that match the study’s contract rules. Their fees total exactly 13,880.855003885266457810 POL. This total covers only those receipts. It does not measure fees for the full period or verified Polymarket company spending.
Users, relayers or other parties can pay these contract fees. To identify company spending, we would also need a verified list of who funded each payer. The dollar figures value tokens at market prices; they do not show what anyone paid to buy those tokens.
Among transactions sent to the listed contracts or recorded in their logs, that day accounted for 161,984.849 POL, or 29.86% of the month’s token fees. The 03:41–04:44 UTC interval alone accounted for 51,357.110 POL.
The day's gas-weighted effective price was 1,061 gwei; the peak interval averaged 3,144 gwei.
Checked against two providers.
Both RPC providers returned the same fee evidence for block 63,939,674: 104 attributed transactions paid 62.032779239477440160 POL, at effective prices of 2,646–4,560 gwei.
This checks one block during the peak against two providers. The full-month totals were checked separately against the complete indexed dataset.
What the earlier sample missed.
The 120-block November sample visited 6 November only at 12:36 UTC, outside the peak interval. That day explains 76.7% of its native-fee underestimate.
The GhostHunter paper reports about 2.35m POL across its reverted match-order set. Our broader contract-and-date census finds 1,961,077 failed calls costing 2,348,906.983241292600142594 POL, within 0.05% of that rounded fee figure.
Scope and valuation matter.
The check covers V1 from 15 August 2025 up to, but not including, 28 April 2026, and V2 from 28 April up to, but not including, 6 May 2026. Valued at daily token prices, these fees total $226,075.40. Including 6 May adds 1,441 calls and 81.03 POL. The audit includes both versions.
Attack attribution remains unverified.
Our count is 0.44% above the paper’s headline figure. We counted every failed call to those contracts without decoding which function each call invoked. A similar fee total does not confirm which calls were attacks, who funded them, what caused them or whether anyone profited.
Adjust transaction volume, growth, failed attempts and network fees. These controls update the detailed sample calculator and projections. They use the small transaction sample, not the full September gas total. The comparison at the top keeps its original assumptions.
Model growth, fee pressure and Solana implementation
EIGHT-BATCH PLANNING SCENARIO
Estimate fees and capacity for your workload.
This calculator repeats the mix of work in the eight-batch sample. You choose the monthly volume; the starting value is an assumption, not a historical count.
%
2.63mmaker matches in month 11.93source batches / sec at assumed 5× peak
Chain
Month 1 fees
12-month fees
Native tokens / month 1
Capacity check at peak demand
The calculation holds network fee inputs from 3 October 2026 and your chosen token prices constant. Fees and transaction acceptance can change when a network is busy. These estimates do not guarantee throughput or include the cost of migrating.
Adjust fee pressure, retries and Solana implementation
%
%
×
The retry model assumes a failed EVM attempt uses half the execution gas of a successful attempt, but the same amount of data. A failed Solana transaction recorded on-chain pays the full modeled fee. Dropped transactions are excluded. Data fees and Solana’s base fee per signature stay fixed. These estimates exclude operating costs, migration costs and money tied up in Solana accounts.
The defaults assume this workload can use 50% of each network budget and that peak demand is 5× the average. You can change both; neither is a measurement of Polymarket traffic or spare network capacity. The Solana shared-account check assumes every transaction writes to one account. It includes known signature costs, but leaves out write-lock, instruction-data and loaded-account overhead. For v1, the equivalent compute-unit priority bid is converted into a total fee in lamports.
Compare finality, reliability and HyperCore trader fees
FROZEN RESEARCH / PERFORMANCE AND RELIABILITY
Throughput. Finality. Reliability.
Block time, execution capacity and settlement finality are different measurements.
HyperCore and HyperEVM run different kinds of work. The comparison above estimates HyperEVM contract-settlement fees. To estimate HyperCore / HIP-4 costs, you also need the value of trades, the maker/taker mix and market deployment costs. HyperCore’s order throughput does not measure HyperEVM capacity. Fee schedule
HyperCore / HIP-4: model outcome trading fees
Estimate traders’ closing fees
This calculator estimates fees on closing trades in an outcome market quoted in USDC. Enter the USDC value of the closing trades that incur a fee. Count each fee-paying party’s trade value once; do not automatically charge both sides of every trade. This input is not total trading volume or a transaction count. Opening positions can be fee-free, and settlement charges need a separate calculation. These are fees paid by traders, not gas paid by an operator. The five-chain comparison above excludes Polymarket trading fees.
The starting values of 1m USDC and 50% maker share are examples, not measured Polymarket activity. Rates use the tier-0 spot reference with no staking or referral discount; account rates vary. This model does not pay maker rebates; negative maker fee rates are treated as zero. Only a fee scale of 1 has been checked against live fills; other settings are estimates. The model does not convert the 21-match sample into a USDC trade value. It excludes builder charges, settlement fees, deployer staking capital and market limits. Amounts are shown in USDC.
A live fill provides one check: a seller paid 0.08096058 USDC on a trade value of 60.23853 USDC. That matches a 7-bp base rate × 2 deployer multiplier × 0.96 referral factor. The buyer paid zero. The overview initially described zero fees, and the deployer page says “testnet-only.” Those descriptions conflict with the observed mainnet metadata and fills, so this model uses the observed fees.
Reported incidents, failed transactions and throughput claims
Polygon
Reported events include the July 2025 Heimdall halt, incorrect milestone recovery, and September 2026 producer disruption. The September report describes about 20 minutes of degraded service, but its component labels are inconsistent. The July Heimdall recovery time has not been verified. A design intended to prevent reorganizations does not prove that none occurred.
11m payments/sec refers to off-chain channel benchmarking. It is not on-chain Polymarket settlement throughput.
The notice for Hyperliquid’s 3 October maintenance lists a 20-minute scheduled window and about 10 minutes of expected downtime. The actual halt duration has not been verified. HyperCore order throughput does not measure HyperEVM contract capacity.
Failures are not outages.
A contract revert still consumes gas; it does not imply chain failure. Full-period failed transaction counts and fees require the completed census.
The newest outage listed in Solana’s incident feed is from February 2024, before this study’s period. The feed therefore cannot establish whether later incidents occurred. We do not use it to estimate annual uptime or fork rates. View incident dates and source checks
05 / OPEN THE EVIDENCE
Check the data and calculations.
Download the measured fees, model inputs and calculations below. The files distinguish recorded activity from estimates for other chains.
Network observations and fee quotes were collected on 3 October 2026. The 30-day and 90-day prices are simple averages of daily USD observations taken near noon, ending on that date. They are not weighted by trading volume and do not forecast future prices. Changing the price window changes the dollar valuation, but not the number of tokens required.