Is Funding Rate Arbitrage Profitable After Fees? A Historical Test

Funding-rate arbitrage profit: what did both legs actually earn after four fills?

A quoted positive funding rate is income on the short position’s settlement notional, not profit on your entire account. This bridge keeps the original completed funding-inclusive study through August 2026. It scales each of the latest complete 30-day starts to $1,000 committed capital; it is not a fresh October trade or a newly optimized rule.

Buy spot quantity q and short the identical base-asset quantity. The spot leg earns q×(spot exit−spot entry); the perp leg earns q×(perp entry−perp exit). Their sum is basis P&L. Add actual intervening funding credits/debits, then subtract both entry and both exit allowances. Unlike an example based on a constant advertised rate, this ledger can show negative funding income and a losing hedge.

The original allocation uses one unit of spot capital, short collateral equal to the opening perp notional, plus entry costs. Thus roughly half the committed capital buys spot and the other half backs the short. You cannot divide the full funding credit only by a tiny margin balance and call that an account return. Fees here are 0.10% spot and 0.05% perp per fill, with 0.02% adverse allowance on every fill.

Verify: Complete 372 base-cost dual-leg starts

How much spot, margin and spare capital does a funding hedge need?

Capital locked in spot is not immediately the same as collateral available to a derivatives wallet. A fast price rise can make the short lose mark-to-market before selling the appreciated spot. Transfers, venue failures and withdrawal suspensions can prevent the intended hedge from meeting a margin demand in time. Equal quantity neutralizes much of directional exposure; it does not neutralize those operational paths.

The column below reconciles the original opening commitment exactly. Extra reserve is not embedded in the headline $1,000. If you reserve another $200 without deploying it, your denominator becomes $1,200 and the same dollar net P&L has a smaller return. Do not silently keep a $1,000 denominator when discussing additional safety capital.

Our margin-breach flag uses the original simplified 0.5% maintenance assumption and exit-fee reserve on observed hourly mark prices. Any breached path has no realizable endpoint return in the source; the counterfactual arithmetic cannot be sold as obtainable profit. Actual tiers, intrahour spikes, insurance, cross-margin and liquidation penalties require a separate venue model.

When should you reject or close a funding-rate arbitrage idea?

Reject a funding-rate arbitrage idea when expected settlement income cannot cover both legs’ entry/exit costs, a plausible basis reversal and its required collateral buffer. Recompute with actual settlement intervals and executable quotes for equal quantities. Annualizing one unusual settlement assumes persistence; this historical study does not establish that assumption.

Stop evaluation—not necessarily market-close at any price—when a leg is unavailable, quantities differ, funding history is missing, collateral cannot be transferred, or the proposed net yield requires leverage that your loss buffer cannot carry. Establish a contingency for an orphaned leg before entering. No Telegram alert automatically closes a hedge or protects it from liquidation.

In this specific frozen sample, 0/150 seven-day starts and only 21/129 thirty-day starts ended positive at the base allowance. That does not prove all carry strategies fail; it demonstrates why a settlement rate alone is insufficient. More favorable fees or basis entries are different scenarios that must be frozen and tested separately.

How recent are the arbitrage prices and completed funding settlements?

The last verified price-only context in this frozen edition is September 19, 2026 UTC; its complete funding-inclusive backtest ends in August 2026. The dated price table is not a current quote, live liquidation level or arbitrage signal. Editing this article on October 3 does not advance either data cutoff.

The complete funding-inclusive experiment still uses September 2025–August 2026. On September 20 the September monthly funding archive and requested daily funding file were unavailable; the public futures funding REST endpoint was unavailable from this server location. We cannot honestly extend the combined experiment by replacing actual settlements with a current rate or zero. Recent price data alone do not make the old return sample current.

The newest completed observations within that verified input range appear next. Historical explanatory charts are now selected by latest eligible date, not earliest date or median return. Until real September settlements are available, this page does not claim a September funding-inclusive backtest.

Verify: Recent mark inputs and funding availability checks

Which completed Binance arbitrage cohorts can you actually verify?

Every available market in the latest complete prescribed Monday cohort within the frozen September-2025–August-2026 funding input is shown, including losses. These historical completed cohorts are not today’s funding opportunities.

The answer: a positive funding rate is not enough

Funding-rate arbitrage was not an automatic income stream in this experiment. After both legs’ fees and slippage, 0/150 complete seven-day starts were profitable; that rose to 21/129 (16.3%) at thirty days and 36/93 (38.7%) at ninety days. The test buys BTC/ETH/SOL spot and shorts equal coin quantities on Binance, with a separately funded short margin account. It measures net return on all committed capital—not advertised APR on one leg.

The returns below include the change in the spot–futures gap and hypothetical entry/exit costs. We do not cherry-pick the highest displayed rate or close at the best future point. Every eligible Monday starts the same prescribed strategy.

The strategy can be directionally hedged while still exposed to basis changes, margin pressure, negative funding, execution failure and exchange custody. “Market-neutral” does not mean risk-free. We neither recommend a venue nor assume derivatives are available in your jurisdiction.

With $1,000, how much capital is actually earning the advertised rate?

This model buys about one unit of spot notional and reserves roughly another unit as isolated collateral for the equal-quantity futures short. It also funds opening costs. About half the total account is therefore spot exposure and half is margin. A funding return quoted against the short’s notional is not a return on the entire budget.

We hold equal coin quantities, not simply equal dollar labels. Let q=1/spot entry price. The futures short initial notional is q×futures entry price. Initial total capital equals spot purchase principal + short collateral + both opening costs. No idle-cash lending or borrowed spot is included.

Budget scaling is proportional in this model; real $100 or $1,000 accounts may face minimum orders, transfer charges and different spread/slippage. The experiment does not show that a small account should undertake a complex hedged trade.

How to separate funding income, basis P&L and costs

Net P&L = spot price P&L + short-futures price P&L + realized funding receipts − both-leg entry/exit costs. When spot and futures move equally in currency units, the two price legs offset. When the gap changes, they do not. Positive receipts can coexist with a negative total.

Funding is accumulated settlement by settlement as actual rate × coin quantity × mark-price reference at that timestamp. Positive rates pay the short; negative rates charge it. We do not multiply the entry rate by all future settlements, and settlement intervals come from the source rather than being assumed identical forever.

Fees and slippage here are transparent assumptions, not your exact tier or a historical fee audit. The following sensitivity changes slippage on every fill while keeping source prices, entry times and base fees unchanged.

How many days until fees are recovered—and can that profit disappear?

There is no guaranteed number of days until a funding hedge covers its costs. In this frozen Binance BTC/ETH/SOL experiment, first break-even is the first sampled UTC day boundary where closing both legs would have non-negative modelled net P&L, including accumulated funding, current basis and closing allowances. It can become negative later and is not an exact intraday crossing.

The median waiting time below is conditional on reaching break-even within the chosen horizon. The count that never reached it is essential: quoting only the successful waiting times would make an unreliable strategy look predictably fast. Sampling is daily, not an exact intraday first-passage time.

BTCUSDT: where the latest complete 30-day carry went

Latest complete 30-day net-return case for this market, starting 2026-07-27T00:00:00Z. Funding contributed +0.29% of total starting capital, the two price legs contributed +0.01% and costs consumed +0.21%. Net: +0.09%.

Top: spot and futures normalized to their own entry price. Middle: the accounting bridge; funding plus basis minus costs equals net. Bottom: the short’s margin surplus under the stated model, expressed as a share of total capital. These are separate concepts and none is an annualized promise.

ETHUSDT: where the latest complete 30-day carry went

Latest complete 30-day net-return case for this market, starting 2026-07-27T00:00:00Z. Funding contributed +0.24% of total starting capital, the two price legs contributed +0.00% and costs consumed +0.21%. Net: +0.02%.

Top: spot and futures normalized to their own entry price. Middle: the accounting bridge; funding plus basis minus costs equals net. Bottom: the short’s margin surplus under the stated model, expressed as a share of total capital. These are separate concepts and none is an annualized promise.

SOLUSDT: where the latest complete 30-day carry went

Latest complete 30-day net-return case for this market, starting 2026-07-27T00:00:00Z. Funding contributed +0.22% of total starting capital, the two price legs contributed +0.01% and costs consumed +0.21%. Net: +0.02%.

Top: spot and futures normalized to their own entry price. Middle: the accounting bridge; funding plus basis minus costs equals net. Bottom: the short’s margin surplus under the stated model, expressed as a share of total capital. These are separate concepts and none is an annualized promise.

Why a hedged position can still fail

A rising market may lift the spot holding while draining the isolated short’s margin. A spot gain in a separate account does not automatically transfer collateral. Funding payments help or hurt the short’s balance, but do not eliminate that path risk. None of the accepted base-cost observations crossed our fully funded short boundary; that is a sample result, not evidence of immunity.

Actual maintenance tiers, mark-price construction, transfer delays, partial fills, delisting and exchange restrictions can matter. We check an explicit fixed-maintenance proxy; we do not claim this is a venue bot’s executed return. Cross-exchange arbitrage adds transfer, venue and synchronization risks and is not tested here.

Before evaluating a displayed opportunity, separate a current rate quote from a paid settlement, verify spot and contract identity, estimate the entire round trip, and measure available collateral. Waiting longer can spread fixed entry costs, but it also extends exposure to rate reversal and operational risk.

Verify: Binance: funding arbitrage bot and risk explanation · Current cross-exchange funding and calculator

Methods, field definitions and reproducible evidence

Buy spot and short equal coin quantity at each eligible Monday UTC open. Hold 7/30/90 days with no hindsight exit. Inputs: BTC/ETH/SOL Binance daily spot plus hourly USD-M trade and mark prices and actual settlement rates, September 2025–August 2026. Mark data lack June 29, 2026 UTC in the frozen monthly files; the shared ledger has 48 incomplete and 57 immature candidates, excluded from results.

q=1/S0; collateral=q×F0; initial capital=1+collateral+spot entry cost+future entry cost. Net=q×(Send−S0)+q×(F0−Fend)+sum(q×mark at settlement×rate)−all four execution costs. Base spot fee 0.10%, future fee 0.05%, slippage 0.02% each fill. Internal settlement timestamps only; entry and exit settlements excluded. ROI=net/initial capital, not net divided by margin or spot alone.

Short wallet starts at collateral; opening costs are separately funded. At each hourly mark high, equity=collateral+funding+q×(F0−mark). Compare to (0.5% maintenance+future closing cost rate)×q×mark. Breached rows retain counterfactual_net_return but net_return=null; never include them as achievable returns. First break-even samples daily opens after any internal settlement. No leverage optimization, tax, cash yield, lending, actual fill reconstruction or calibrated risk forecast.

All displayed percentages are historical descriptions, not calibrated probabilities of your next trade. Starts overlap and assets share market shocks. Counts are not independent trials; no causal effect, statistical significance or execution guarantee is claimed. Missing and pending outcomes remain separate from zero. Inputs were acquired retrospectively, not preserved as real-time investment knowledge.

The complete files load only when you choose a download. Download spot.json.gz and derivatives.json into one folder, save four_decision_studies.py, then run: python four_decision_studies.py INPUT_FOLDER OUTPUT_FOLDER. Python standard library only; no exchange key or live trading is used. The input directory contains 591 discovered target markets, not 591 complete markets for every analysis. Per-market coverage and per-window eligibility are published.

JSON carries original float precision; CSV contains the same observation fields. UTC timestamps, missing values, cost scenarios and model variants must be retained when reproducing a table. Hashes protect this frozen edition against accidental changes; they do not prove economic completeness.

Verify: Full study JSON · Observation CSV · Summary CSV · Candidate and missing ledger · Daily source bundle · about 19 MB gzip · Hourly derivative inputs · about 5 MB · Per-market daily coverage · Reproduction program · Preset protocol · Hashes and source coverage · Chart coordinates

Frequently asked questions

Sources and verification

• Binance public-data documentationPrimary data format and checksum documentation; our simulation is not a Binance trading record.
• Binance spot market-data fieldsDaily OHLCV; candles are not execution, order-book or personal cost-basis evidence.

Use evidence without outsourcing your decision.

Inspect the market tools and see exactly which alerts CoinNudge delivers.