Intraday ETF Arbitrage Trading Strategy Analysis
Identify and execute ETF arbitrage opportunities with Sourcetable AI. Calculate NAV discrepancies, track premium-discount spreads, and analyze creation-redemption mechanics automatically.
Andrew Grosser
February 24, 2026 • 15 min read
Introduction to Intraday ETF Arbitrage
March 2020: SPY hit a 4% premium to NAV during the morning session as bond ETFs like LQD traded 5% discounts. Market stress created the widest ETF arbitrage gaps since 2008. ETF arbitrage represents one of the most sophisticated yet accessible trading strategies available to institutional and retail traders. When an ETF trades at a premium or discount to its net asset value (NAV), arbitrageurs step in to profit from the discrepancy. This price difference—often just 0.05% to 0.50%—creates hundreds of opportunities daily across thousands of ETFs.
The challenge isn't finding these opportunities. It's analyzing them fast enough. You need real-time NAV calculations, basket composition tracking, bid-ask spread analysis, and execution cost modeling. Traditional Excel spreadsheets require complex formulas linking multiple data sources, constant manual updates, and sophisticated macros that break when market data changes sign up free.
Sourcetable transforms ETF arbitrage analysis from a formula-heavy spreadsheet nightmare into a conversational experience. Upload your ETF holdings, basket compositions, and pricing data. Then ask questions in plain English: 'Which ETFs are trading above NAV?' or 'Calculate arbitrage profit for SPY with 0.02% transaction costs.' The AI instantly analyzes thousands of data points, calculates spreads, models execution scenarios, and generates visual comparisons.
This guide explains how intraday ETF arbitrage works, why Sourcetable outperforms traditional spreadsheet analysis, and how to implement this strategy with AI assistance. Whether you're an authorized participant executing creation-redemption arbitrage or a trader capitalizing on intraday price inefficiencies, you'll discover how to analyze opportunities faster and execute with confidence. Try Sourcetable for free at sign up free.
Why Sourcetable Excels at ETF Arbitrage Analysis
ETF arbitrage demands split-second decisions based on complex calculations. You're comparing real-time ETF prices against calculated NAV, factoring in creation-redemption costs, transaction fees, bid-ask spreads, and execution slippage. In Excel, this requires linking to external data feeds, building basket pricing models, creating custom functions for spread calculations, and constantly refreshing data.
Sourcetable replaces this complexity with natural language commands. Import your ETF universe—say 500 ETFs with their basket compositions, weights, and real-time prices. Instead of writing VLOOKUP formulas to match constituent prices, calculating weighted averages for NAV, and building conditional formatting to highlight arbitrage opportunities, you simply ask: 'Show me ETFs trading more than 0.15% above NAV with daily volume over 1 million shares.'
The AI understands arbitrage terminology. It knows that premium means ETF price exceeds NAV, that creation units typically come in 50,000-share blocks, and that transaction costs include brokerage fees plus market impact. When you ask 'What's the net arbitrage profit for QQQ?', Sourcetable calculates the premium, subtracts estimated transaction costs, factors in the bid-ask spread, and returns the net profit per share—all in seconds.
Traditional spreadsheets break when basket compositions change during quarterly rebalances. Sourcetable adapts automatically. Upload the new basket composition file, and the AI recalculates all NAV values using updated weights. No formula fixes, no broken cell references, no manual verification. The system maintains data integrity while you focus on identifying profitable trades.
Visualization matters in arbitrage trading. You need to see premium-discount patterns across time, compare spreads across similar ETFs, and identify which market conditions create the best opportunities. Sourcetable generates these charts instantly. Ask 'Show premium-discount trend for SPY over the last 5 days' and get an interactive chart showing intraday patterns, volume correlations, and statistical outliers. No pivot tables, no chart wizards, no manual axis formatting.
Benefits of ETF Arbitrage Analysis with Sourcetable
ETF arbitrage offers unique advantages for sophisticated traders: market-neutral returns, high win rates, and scalable execution. The strategy works because ETFs must track their underlying baskets, creating predictable price convergence. When discrepancies emerge, they rarely last long—making speed and accuracy essential. Sourcetable delivers both through AI-powered analysis that eliminates manual calculation delays.
Real-Time NAV Calculation Without Formula Complexity
Calculating NAV requires pricing every constituent security, applying correct weights, and summing the basket value. For an ETF with 500 holdings like SPY, that's 500 price lookups, 500 multiplication operations, and continuous updates as prices change. In Excel, you'd build a SUMPRODUCT formula referencing external data feeds, create error handling for missing prices, and set up automatic refresh macros.
Sourcetable handles this complexity invisibly. Upload your basket composition file with tickers, shares per creation unit, and current prices. Ask 'What's the current NAV for SPY?' and the AI calculates instantly: (Price₁ × Shares₁ + Price₂ × Shares₂ + ... + Price₅₀₀ × Shares₅₀₀) / 50,000. It accounts for cash components, accrued dividends, and adjustment factors automatically. When constituent prices update, NAV recalculates without manual intervention.
- Indicative NAV (iNAV): Published every 15 seconds by ETF sponsors; SPY iNAV reflects the real-time value of S&P 500 components; a 0.5% premium means SPY shares trade $0.50 above every $100 of underlying value—a direct arbitrage opportunity.
- Creation/Redemption Mechanism: Authorized participants (APs) can create ETF shares by delivering the underlying basket, or redeem shares to receive the basket; this arbitrage mechanism normally keeps premiums/discounts within 0.1–0.2% in liquid ETFs.
- Basket Composition Timing: Creation basket published daily before market open; if basket changes overnight and the AP hasn't updated their hedge, they face basis risk during execution—timing mismatches are a key operational risk.
- Bond ETF Complexity: Corporate bond ETFs like LQD have iNAV based on dealer quotes that lag actual tradeable prices; during the March 2020 liquidity crisis, LQD traded at a 5% discount because bond market prices were unreliable—the ETF was actually price discovering for the underlying bonds.
Automated Premium-Discount Screening Across Thousands of ETFs
Profitable arbitrage requires monitoring hundreds or thousands of ETFs simultaneously. You're looking for spreads exceeding your cost threshold—typically 0.10% to 0.25%—with sufficient liquidity for execution. Building this screener in Excel means creating a master table with ETF prices, calculated NAVs, premium-discount calculations, volume filters, and conditional formatting rules.
With Sourcetable, screening becomes conversational. Import your ETF universe with pricing data. Then ask: 'Show me all ETFs with premium greater than 0.20% and average daily volume above 500K shares.' The AI instantly filters 2,000+ ETFs, calculates premiums as (ETF Price - NAV) / NAV × 100, applies volume criteria, and returns a ranked list. Want to narrow further? Add: 'Only show equity ETFs in the technology sector.' The AI refines results immediately without formula rewrites.
- Liquid vs. Illiquid Underlying: ETFs tracking liquid large-cap stocks (SPY, QQQ) maintain premiums/discounts below 0.05%; ETFs tracking illiquid assets (micro-cap, frontier market, high-yield bonds) routinely show 0.5–3% premiums/discounts—the spread reflects underlying liquidity.
- Market Stress Widening: In normal markets, SPY premium/discount averages ±0.01%; in March 2020, it reached ±0.5%; screening for ETFs showing premiums/discounts above 3× their 90-day average identifies stress arbitrage opportunities.
- Commodity ETF Contango: USO (crude oil ETF) rolls futures monthly; in contango markets (near-month cheaper than far-month), rolling from front to next month locks in a loss; USO's 2020 performance included -45% from contango roll costs alone.
- Leveraged ETF Reset: 3× leveraged ETFs reset exposure daily; a 3× ETF loses more than 3× the underlying over multi-day periods due to volatility drag; 2× S&P 500 with 30% daily vol loses (1-2×0.30²) = -18% annually from volatility drag alone.
Transaction Cost Modeling for Accurate Profit Estimation
Raw premium-discount spreads don't equal profits. You must subtract transaction costs: brokerage commissions, bid-ask spreads, market impact, and for creation-redemption arbitrage, the fees charged by authorized participants. These costs typically range from 0.05% to 0.15%, meaning a 0.12% premium might yield only 0.02% net profit—or a loss if costs are higher.
Sourcetable models these costs automatically when you provide parameters. Tell the AI: 'Calculate net arbitrage profit for QQQ assuming 0.01% commission, 0.03% bid-ask spread, and 0.05% creation fee.' It computes gross premium, subtracts all cost components, and returns net profit per share and per creation unit. For a $350 QQQ trading at 0.18% premium with those costs, you'd see: Gross profit $0.63, costs $0.315, net profit $0.315 per share, or $15,750 per 50,000-share creation unit.
- Arbitrage Spread Threshold: SPY bid-ask spread of $0.01 on a $470 price = 0.002%; plus commission of $0.005/share round-trip = 0.001%; minimum arbitrage premium to be profitable = 0.003% (0.3 cents per $100). Most retail traders can't access this threshold.
- Short Selling Costs: The short leg of ETF arbitrage requires borrowing shares; SPY borrow rate is 0.02–0.05% annually (near-zero); niche ETFs can have borrow costs of 5–15% annually, eliminating the arbitrage economics entirely.
- AP Access Requirements: Only registered Authorized Participants can create/redeem ETF shares at NAV; retail traders must trade in the secondary market, meaning they can't directly arbitrage—they can only trade the premium/discount, not eliminate it.
- Slippage on Basket Execution: Creating SPY requires buying all 500 S&P 500 stocks simultaneously; market impact on 500 names at once can cost 2–5bps on a large creation; small premiums are not worth executing due to market impact costs.
Historical Pattern Analysis for Strategy Optimization
Understanding when arbitrage opportunities emerge helps you position for the best trades. Premiums often widen during market opens and closes, around major news events, or during periods of high volatility. Analyzing these patterns in Excel requires historical data imports, time-series calculations, pivot tables for aggregation, and custom charts.
Sourcetable makes pattern analysis effortless. Upload historical premium-discount data for your target ETFs—say, minute-by-minute data for the past month. Ask: 'What time of day does SPY show the largest average premium?' The AI aggregates data by time, calculates average premiums for each minute, and identifies that 9:31-9:45 AM shows premiums averaging 0.14% versus 0.06% during mid-day. Request 'Show this as a chart' and you get an instant visualization highlighting the optimal trading windows.
Basket Comparison for Creation-Redemption Decisions
Authorized participants executing creation-redemption arbitrage need to compare the cost of assembling the basket versus buying the ETF directly. This involves pricing every constituent, calculating total basket cost including transaction fees, and comparing against the ETF's market price plus creation fees. Excel requires complex multi-sheet workbooks with linked pricing tables and cost allocation formulas.
Ask Sourcetable: 'Compare the cost of buying SPY versus assembling the basket with 0.02% transaction costs per stock.' The AI prices all 500+ constituents, calculates total basket cost including per-stock transaction fees, adds the creation fee, and compares against buying 50,000 SPY shares at the ask price. You see immediately: Basket cost $17,512,450, ETF cost $17,525,000, arbitrage profit $12,550. The AI even breaks down which constituents have high transaction costs that erode profitability.
How ETF Arbitrage Analysis Works in Sourcetable
ETF arbitrage exploits temporary price discrepancies between an ETF's market price and the value of its underlying basket. These discrepancies arise from supply-demand imbalances, market microstructure effects, or temporary liquidity constraints. Arbitrageurs profit by buying the cheaper asset and selling the more expensive one, capturing the spread as prices converge.
Step 1: Import ETF and Constituent Data
Start by uploading your data to Sourcetable. You'll need three datasets: ETF pricing data (ticker, current price, bid, ask, volume), basket composition files (constituent tickers, shares per creation unit, weights), and constituent pricing data (current prices for all basket securities). These typically come from market data providers, your broker's API, or ETF issuer websites.
Sourcetable accepts standard formats: CSV, Excel files, JSON feeds, or direct API connections. Upload a file like 'SPY_basket.csv' containing 500+ rows with columns for Ticker, Shares, Weight, and Price. The AI automatically recognizes the structure and links constituent prices to basket calculations. No manual cell referencing, no VLOOKUP formulas, no data validation rules needed.
- Start by uploading your data to Sourcetable.
- Sourcetable accepts standard formats: CSV, Excel files, JSON feeds, or direct AP.
Step 2: Calculate Net Asset Value
With data imported, calculating NAV becomes a simple question. Ask: 'Calculate NAV for SPY based on current constituent prices.' Sourcetable multiplies each constituent's price by its shares in the creation unit, sums all values, adds any cash component, and divides by the creation unit size. For SPY with a 50,000-share creation unit, if the basket value totals $17,500,000 plus $2,500 cash, NAV equals $350.05 per share.
The AI handles edge cases automatically. If a constituent is halted and has no current price, it uses the last traded price with a warning flag. For international ETFs with constituents in different currencies, it applies current exchange rates. When basket compositions include fractional shares or odd lots, calculations remain precise to the penny.
Step 3: Identify Premium-Discount Opportunities
Now screen for arbitrage opportunities. Ask: 'Which ETFs are trading at a premium or discount greater than 0.15%?' Sourcetable calculates the spread for each ETF: Premium/Discount = (Market Price - NAV) / NAV × 100. If SPY trades at $350.60 with NAV of $350.05, the premium is 0.157%. The AI flags this as an opportunity since it exceeds your 0.15% threshold.
Refine your search with additional criteria. Add: 'Only show opportunities with daily volume above 1 million shares and bid-ask spread below 0.05%.' This filters for liquid ETFs where you can execute without excessive market impact. Sourcetable returns a prioritized list ranked by net profit potential, showing you exactly where to focus your trading capital.
- "Which ETFs are trading at a premium or discount greater than 0.15%?"
- Refine your search with additional criteria.
Step 4: Model Transaction Costs and Net Profit
Before executing, model your costs. Tell Sourcetable: 'Calculate net profit for shorting SPY and buying the basket, assuming 0.01% commission, 0.03% bid-ask impact, and 0.04% market impact.' The AI computes: Sell SPY at $350.60 (bid side), pay 0.01% commission ($0.035), pay 0.03% spread ($0.105). Buy basket at $350.05 (NAV), pay 0.01% commission per stock (average $0.035), pay 0.04% market impact ($0.140). Total costs: $0.315. Net profit: $0.55 - $0.315 = $0.235 per share.
For a 50,000-share creation unit, that's $11,750 profit. But Sourcetable doesn't stop there. It calculates return on capital: if you deploy $17.5 million to buy the basket, $11,750 represents 0.067% return. Annualized assuming you can execute this trade 50 times per year: 3.35% return. The AI presents this context so you can evaluate whether the opportunity justifies the capital and risk.
Step 5: Generate Execution Recommendations
Finally, get actionable trade recommendations. Ask: 'Should I execute this arbitrage on SPY?' Sourcetable evaluates multiple factors: spread size (0.157% is moderate), liquidity (SPY averages 80 million shares daily—excellent), volatility (current VIX level affects convergence risk), and historical success rate (based on your past trades if you've logged them).
The AI might respond: 'Recommended. SPY shows 0.157% premium with high liquidity and low execution risk. Historical data shows 87% of premiums above 0.15% converge within 30 minutes. Suggested execution: Short 50,000 SPY at limit $350.58, simultaneously buy basket constituents using VWAP algorithm. Expected net profit $11,750 with 85% confidence.' This level of analysis would take hours in Excel—Sourcetable delivers it in seconds.
Step 6: Monitor and Visualize Performance
After execution, track performance. Upload your trade log with entry prices, exit prices, and actual costs. Ask: 'Show my arbitrage performance over the last month.' Sourcetable calculates win rate, average profit per trade, total return, and Sharpe ratio. Request 'Create a chart showing profit by ETF' and see which ETFs generate the most consistent opportunities.
The AI identifies patterns you'd miss manually. It might notice: 'Your QQQ trades show 12% higher win rate during the first hour of trading' or 'IWM arbitrage opportunities are most profitable on Fridays.' These insights refine your strategy, helping you focus capital on the highest-probability trades. Continuous improvement happens automatically as you feed more data into Sourcetable.
Real-World ETF Arbitrage Use Cases
ETF arbitrage strategies vary based on trader sophistication, capital availability, and market access. From retail traders capturing small intraday discrepancies to authorized participants executing multi-million-dollar creation-redemption trades, Sourcetable adapts to your specific approach. These use cases demonstrate how different market participants apply AI-powered analysis to their arbitrage strategies.
Retail Trader: Intraday Premium-Discount Scalping
Sarah trades with a $100,000 account and focuses on capturing small premiums in highly liquid ETFs. She monitors 50 popular ETFs (SPY, QQQ, IWM, EEM, GLD) looking for premiums exceeding 0.10% that typically converge within minutes. Her challenge: identifying opportunities fast enough to execute before they disappear.
Each morning, Sarah uploads real-time pricing data to Sourcetable—ETF prices, calculated NAVs from her data provider, and bid-ask spreads. She asks: 'Show me ETFs with premium above 0.10% right now.' The AI instantly returns: 'QQQ at 0.14% premium, IWM at 0.11% premium, EEM at 0.18% premium.' She clicks on EEM for details: trading at $44.58 versus NAV of $44.50, volume 8.2 million shares today (well above average), bid-ask spread 0.02%.
Sarah asks: 'What's my net profit shorting 1,000 EEM shares with $7 commission and 0.02% spread?' Sourcetable calculates: Gross profit $80 (0.18% × $44.58 × 1,000), minus commission $7, minus spread cost $8.92, equals net profit $64.08. That's 0.144% return on the $44,580 capital deployed. She executes the trade, and 12 minutes later the premium narrows to 0.04%. She covers for $64 profit after costs—exactly as Sourcetable predicted.
Proprietary Trading Firm: Statistical Arbitrage Across ETF Pairs
A prop trading desk runs statistical arbitrage on correlated ETF pairs—trading one ETF against another when their relative pricing diverges. They monitor 200 ETF pairs like SPY vs. IVV (both track S&P 500), or QQQ vs. XLK (tech sector overlap). When the spread between highly correlated ETFs exceeds historical norms, they execute pairs trades expecting mean reversion.
The desk uploads six months of historical pricing data for their ETF universe into Sourcetable. They ask: 'Calculate the correlation between SPY and IVV, and show current spread versus 30-day average.' The AI returns: Correlation 0.9987 (essentially identical), current spread 0.08% (SPY trading at premium to IVV), 30-day average spread 0.02%, standard deviation 0.03%. The current spread is 2 standard deviations above average—a statistical anomaly.
They ask: 'What's the expected profit from a $1 million pairs trade shorting SPY and buying IVV, assuming spread reverts to mean in 1 day?' Sourcetable models the scenario: Short $500K SPY, long $500K IVV, capture 0.06% spread compression (0.08% current minus 0.02% mean), minus 0.02% total transaction costs, equals 0.04% net profit or $400 on $1M deployed. The AI adds: 'Historical mean reversion occurs within 1 day in 73% of cases when spread exceeds 1.5 standard deviations.' They execute the trade with confidence backed by statistical analysis.
Authorized Participant: Creation-Redemption Arbitrage
An authorized participant (AP) has agreements with ETF issuers to create and redeem ETF shares in exchange for underlying baskets. When an ETF trades at a premium, the AP creates new shares by delivering the basket to the issuer, receiving ETF shares that can be sold at the premium. When at a discount, they redeem ETF shares for the basket, which is worth more than the ETF price.
The AP monitors 500 ETFs for creation-redemption opportunities. They upload basket composition files for all ETFs plus real-time pricing for thousands of constituent securities. They ask Sourcetable: 'Which ETFs have premiums above 0.25% with creation unit sizes under 100,000 shares?' The AI filters and returns: 'XLF at 0.31% premium, 50,000 share creation unit. IJH at 0.28% premium, 50,000 share creation unit. VNQ at 0.34% premium, 25,000 share creation unit.'
They focus on VNQ, a real estate ETF. They ask: 'Calculate net profit creating one VNQ unit, including basket assembly costs of 0.08%, creation fee of $500, and selling VNQ at current bid.' Sourcetable computes: VNQ bid price $78.45, NAV $78.19, premium 0.33%. Buy basket for $1,954,750 (25,000 shares × $78.19), pay 0.08% assembly costs ($1,564), pay $500 creation fee, receive 25,000 VNQ shares, sell at $78.45 for $1,961,250. Net profit: $1,961,250 - $1,954,750 - $1,564 - $500 = $4,436. That's 0.227% return on $1.95M capital for a same-day trade.
Hedge Fund: Cross-Border ETF Arbitrage
A hedge fund specializes in cross-border arbitrage, exploiting price differences between ETFs trading in different markets. For example, iShares MSCI Japan ETF (EWJ) trades in New York while similar ETFs trade in Tokyo. When time zone differences, currency fluctuations, or market sentiment create price discrepancies, the fund profits by trading across both markets.
The fund uploads pricing data from NYSE and Tokyo Stock Exchange, currency exchange rates, and basket compositions for 50 international ETF pairs. They ask Sourcetable: 'Compare EWJ pricing in New York versus equivalent Tokyo ETF adjusted for currency and time differences.' The AI calculates: EWJ in NY trading at $58.20 (as of 4:00 PM ET), Tokyo equivalent trading at ¥8,450 (as of 9:00 AM Tokyo time next day), USD/JPY rate 145.20, implied NY price $58.19. Discrepancy: 0.017% premium in NY.
Too small for a trade. But they ask: 'Alert me when any cross-border ETF pair shows discrepancy above 0.15%.' Sourcetable monitors continuously. Two hours later: 'Alert: EWZ (Brazil ETF) shows 0.19% premium in NY versus Sao Paulo equivalent.' They investigate, model transaction costs including currency conversion fees (0.05%), overnight financing costs (0.03%), and broker commissions (0.02%), and execute a $5 million pairs trade capturing 0.09% net profit—$4,500 for a low-risk arbitrage.
Frequently Asked Questions
If your question is not covered here, you can contact our team.
Contact UsWhat is the ETF arbitrage mechanism and how do authorized participants keep ETF prices near NAV?
How do intraday ETF dislocations occur and what triggers them?
How is ETF intraday value (IIV) calculated and why does it sometimes mislead investors?
What quantitative signals best predict short-term ETF premium/discount reversions?
How do ETF arbitrageurs size positions and manage risk during dislocation events?
What regulatory and infrastructure requirements are needed for professional ETF arbitrage?
How does the cash vs. in-kind creation process affect ETF arbitrage profitability for fixed income ETFs?
