Index Arbitrage Strategy

Theoretically yes, but practically there are risks: execution risk (both legs may not fill at expected prices), tracking error (the basket may not perfectly match the XJO), franking/dividend risk (the value of imputation credits embedded in the basis is uncertain), SOQ settlement timing on expiry day, and margin risk (you can face margin calls despite being hedged). With proper execution and a correct franking model, risks are manageable but not zero.

The contract size drives it. One SPI 200 contract is ~A$220,000 notional at index 8,800; the Mini SPI 200 (A$5/point) is ~A$44,000 and is the natural small-capital vehicle. For ETF-vs-future arbitrage: ~A$50-100k can work using Minis. For a meaningful optimized basket: A$500k+. For a full 200-stock basket: institutional scale (A$5m+). Calendar spreads can be done with relatively little since both legs are futures.

In heavy franked-dividend windows (Feb-Mar, Aug-Sep), the cash dividends plus the value of attached franking credits subtracted from carry can exceed the carry itself, so fair value - and the actual future - trades BELOW spot. This negative basis is normal in Australia and reflects dividend imputation; it does not by itself signal an arbitrage.

After costs, expect modest annualized returns in normal markets - lower than India's because the carry (BBSW ~4.4%) and therefore the basis are smaller, and the large-cap market is efficient with HFT present. The lower cost base (no STT) helps, but clean mispricings are less common. The richest, more reliable edges come from franked-dividend season and the quarterly expiry/rebalance window rather than everyday basis.

Yes, through put-call parity on XJO/SPI 200 options: Synthetic Future = Call - Put + Spot should equal the actual future. If they differ beyond costs, arbitrage exists - and the franking value also shows up in the option-implied forward. Options add complexity (Greeks, liquidity) and are typically used by sophisticated traders.

Track ex-dates, cash amounts and franking percentages for the top ~15 names (banks + big miners drive ~60% of the impact). Adjust fair value daily as ex-dates approach, separating the cash dividend (which the index drops by) from the franking (valued at gamma). The future should already reflect the expected dividend and the market's franking value; your edge depends on whether YOUR franking value differs from the market's gamma.

Cash-futures involves a spot basket vs the SPI 200 future - it requires stock execution, has franking exposure, and zero tracking error if done in full. Calendar spread involves the near quarter vs far quarter future - simpler (both futures), lower cost, no tracking error, but each spread spans a full quarter of carry and the dividend/franking estimate dominates. Calendar is more accessible for smaller capital.

Select the top 15-20 stocks by index weight and liquidity. The XJO is top-heavy, so the top 15 (CBA, BHP, CSL, the big-4 banks, Macquarie, Wesfarmers, Goodman, etc.) cover ~55-60% of the index. Capture the remaining weight via slight overweighting and beta adjustment. Target tracking error < 0.5% and rebalance if it drifts - especially around the quarterly reconstitution.

Around: franked-dividend season (Feb-Mar and Aug-Sep, when fair value gets complex and franking is mispriced), the quarterly expiry+rebalance window (SPI 200 third-Thursday expiry plus the index rebalance the next day cluster the flows), overnight gaps (the SPI 200 trades a long session tracking US markets), and any stress that dislocates the basis. HFT captures everyday intraday mispricing; focus on these structural windows.

Use basket orders for simultaneous stock execution, monitor both legs in real time, and have contingency plans for partial fills. Execute during high-liquidity periods and use realistic limit prices on the futures. Critically, on expiry day unwind the basket into the open to converge against the SOQ rather than waiting until the close. Accept some slippage as a cost of business.

Use a BBSW/OIS (AONIA) term structure matched to expiry; model each dividend with precise timing and present-value discounting, SEPARATING cash from franking; estimate gamma (the market franking value) by backing it out from the SPI 200 basis around franked ex-dates; and generate a confidence interval, not a point estimate. Backtest against historical convergence to the SOQ and recalibrate gamma each franked-dividend season.

Sector relative value (Financials vs Resources via ETFs or stock baskets - statistical, no guaranteed convergence), residual arbitrage (stocks with significant residuals vs the XJO), quarterly index-rebalancing arbitrage (front-run passive flows - four windows a year, compounded by expiry-day proximity), and franking-clientele transfers (moving franking from low-value to high-value holders). These offer higher returns but lack the guaranteed convergence of SPI 200-vs-cash.

Do not compete on speed. HFT owns the SPI 200/cash lead-lag and ETF-vs-future microsecond loops from co-location at the ALC. Focus on what HFT structurally avoids: franked-dividend-season mispricing (multi-day gamma/dividend model risk), the quarter-end expiry+rebalance flow window, overnight gaps, and longer carry holds. Your edge is a correct franking model and patience, not latency.

Minimum: real-time data for the 200 constituents, an XJO calculation engine, a FRANKING-AWARE fair value calculator (with a gamma estimate), broker API integration (IBKR is the primary Australian SPI 200 + API route), and a position/margin dashboard. Advanced: low-latency execution, automated basket generation, SOQ-aware expiry unwind, and risk/alerting systems. Build incrementally - get the franking model right first, then add speed.

Test: correlation breakdown (basket vs XJO diverges 2%+); franking gamma wrong by +/-0.2; a major bank cuts or changes its franking percentage; execution failure (one leg only); SOQ leakage (basket not unwound at the open); covered-borrow unavailable on the reverse leg; and an ASX Clear (Futures) margin spike of 50%. For each, quantify the impact and define a contingency. Run tests quarterly (aligned to the expiry/rebalance cadence) and after any system change.