Quant Momentum Strategy

Historically, momentum strategies in Australia have generated roughly 2-4% annual alpha over broad indices like the S&P/ASX 200 - more modest and noisier than US momentum, largely because the index is so concentrated in banks and miners that cross-sectional dispersion is narrower. Returns vary significantly by year - some years +15% alpha, others -5%. Expect higher volatility than index investing. Long-term (5+ years) investors benefit most from momentum's compounding alpha.

25-35 stocks is the sweet spot. Fewer than 20 creates concentration risk and high tracking error. More than 50 dilutes momentum exposure and approaches index returns. 30 stocks provides good diversification while maintaining a meaningful momentum tilt.

Research shows 12-month (excluding the recent month) is optimal. 1-month has too much noise and reversal effects. 6-month works but misses some momentum. 12-month captures the full momentum cycle. The '12-1' formulation (skipping the recent month) further improves results by avoiding short-term reversal.

Monthly is standard - it captures momentum shifts while keeping turnover manageable (~200% annually). Quarterly rebalancing reduces costs but may miss momentum changes. Weekly is too frequent and costly. Monthly with a buffer zone (exit at rank 40, not 30) optimizes the trade-off - and on the ASX the cost of monthly turnover is unusually low because there is no securities transaction tax or stamp duty.

Yes, momentum can and does lose money, especially during 'momentum crashes' when past winners sharply underperform. Crashes of -20% or more happen every few years (2009, 2020, 2022). However, over full market cycles (5+ years), momentum has historically delivered positive alpha. Risk management (volatility scaling, stops) helps manage drawdowns.

Multiple approaches: (1) Volatility scaling - reduce exposure when A-VIX > 22. (2) Portfolio stop-loss at -15% drawdown. (3) Factor monitoring - reduce if the momentum factor underperforms 2+ months. (4) Diversify with other factors (quality, low volatility). (5) Hedge with S&P/ASX 200 / SPI 200 index puts during elevated risk. Combining approaches provides layered protection.

Equal weight is simpler and more diversified - each stock gets the same allocation. Momentum-weighted concentrates on the strongest stocks for potentially higher returns but more risk. Risk-parity (inverse volatility) weighting is another option that balances risk - but on the ASX pair it with a sector cap, since the lowest-vol names are the correlated big-4 banks. For most investors, equal weight with sector constraints works well.

Several approaches: (1) Sequential screening - first filter for momentum, then quality. (2) Composite scoring - weighted sum of momentum, quality, value scores. (3) Separate portfolios - allocate 60% to momentum, 40% to quality, rebalance between them. (4) Intersections - only buy stocks in the top quartile of BOTH factors. The S&P/ASX 200 Quality and Enhanced Value indices are handy references for the other legs. Start simple (sequential) and add complexity as needed.

For liquid ASX 200 stocks: brokerage A$5-10 flat or ~0.05-0.1%, NO securities transaction tax, NO stamp duty (abolished), impact cost 0.05-0.20%. Total round-trip ~0.15-0.35% - roughly half what an equivalent Indian momentum book costs, because the per-trade taxes simply don't exist here. With 200% annual turnover, expect ~0.25-0.35% annual cost. Reduce via: buffer zone (-20% turnover), TWAP execution (-30% impact), and SPI 200 index futures for large beta moves (liquidity, not tax).

Key requirements: (1) 10+ years of data, survivorship-bias free (include delisted stocks). (2) Point-in-time universe (S&P/ASX 200 as of each date, not current constituents). (3) Walk-forward testing (train on 70%, test on 30%, roll forward). (4) Include realistic Australian transaction costs (~0.15-0.35% round-trip - don't pad for transaction taxes that don't exist). (5) Out-of-sample validation - expect 20-30% performance decay from in-sample. Use out-of-sample results for planning, not in-sample optimized results.

Feature engineering: 12-1 return, risk-adjusted momentum, relative strength, earnings momentum, RSI, sector momentum, A-VIX, plus Australian macro features like AUD/USD trend and ASIC short-interest. Target: top-tercile classification or return regression. Model: XGBoost or LightGBM (best for tabular data). Training: time-series split, walk-forward retraining monthly. Integration: ML probability as 40% weight combined with 60% traditional momentum. Expect ~1-2% additional alpha from ML enhancement - more modest than higher-dispersion markets because the ASX cross-section is small and concentrated.

Long leg: buy the top ~40 stocks by momentum. Short leg: this is the hard part in Australia - naked shorts are banned (covered only), single-stock futures are defunct, and only ~15-20 names have liquid options, so you can't simply short 40 bottom-momentum names. Practical routes: short the most liquid losers via CFDs (IG/CMC) or securities lending where borrow is available, and approximate the rest of the short leg with SPI 200 index or sector-ETF shorts. Expect a structural long bias because the short leg can't be fully built, and remember shorts pay the dividend (and lose the franking credit) to the lender. Check ASIC's daily short-position data to avoid crowded shorts. Rebalance monthly; monitor shorts more often (squeeze risk).

Risk parity weights each stock inversely to its volatility so each contributes equal risk. Formula: Weight_i = (1/Vol_i) / Sum(1/Vol_j). Low-vol stocks get more capital, high-vol less. Combined with momentum: select by momentum, weight by inverse vol. Add volatility targeting: scale total exposure to maintain a target portfolio vol (e.g. 15%). On the ASX, cap single names and the Financials sector - otherwise inverse-vol weighting concentrates into the low-vol, highly correlated big-4 banks. Result: momentum exposure with balanced, consistent risk.

Base allocation 100%. Modifiers (A-VIX scale): A-VIX > 22 -> -25%; A-VIX > 28 -> -50%. Momentum factor negative 2 months -> -25%. Bear market regime -> -25% to -50%. Ranging market -> -25%. Calculate weekly, but smooth transitions over 2-4 weeks. Example: A-VIX 24 (-25%), factor weak (-25%) = 50% allocation. When multiple signals are negative, compound reductions but maintain a 25% minimum.

Data pipeline: automated daily price/volume updates, corporate action adjustments, quality checks. Execution: broker API integration (IBKR, CMC, IG offer Australian APIs), order generation from targets, TWAP/VWAP algorithms. Risk: real-time position monitoring, drawdown alerts, factor exposure tracking (watch the Financials cap). Reporting: daily P&L, weekly risk, monthly attribution. Governance: change management process, backtest before modifications. Start simple, add sophistication as AUM grows.