{"id":522,"date":"2026-06-15T12:00:00","date_gmt":"2026-06-15T12:00:00","guid":{"rendered":"https:\/\/www.tejwin.com\/en\/insights\/factor-strategy-qfii-part-2\/"},"modified":"2026-09-23T18:19:23","modified_gmt":"2026-09-23T18:19:23","slug":"factor-strategy-qfii-part-2","status":"publish","type":"insight","link":"https:\/\/www.tejwin.com\/en\/insights\/factor-strategy-qfii-part-2\/","title":{"rendered":"Factor Strategy \u2013 Integrating Broker Consensus to Enhance Foreign Concentration Strategies \u2013 QFII Part 2"},"content":{"rendered":"
\"\"<\/figure>\n
<\/figure>\n

From Theory to Active Execution \u2014 Translating Foreign Concentration into an Actionable Strategy<\/h2>\n

Following our analysis in\u00a0Part 1:\u00a0\u00a0Factor Research\u00a0\u2013 \u00a0Tracking\u00a0Smart Money Footprints via Foreign Institutional Concentration<\/a><\/strong>,\u00a0we have systematically verified that the foreign institutional trading concentration factor (conc_qfii)\u00a0possesses\u00a0robust and cumulative return predictive power within the large-cap universe. This article takes a practical perspective to transform our empirical findings into fully executable trading strategies. Utilizing an event-driven\u00a0backtesting\u00a0engine, we evaluate real-world feasibility by strictly deducting transaction costs and enforcing realistic trading limitations.<\/p>\n

If you have not yet read our baseline factor analysis\u00a0regarding\u00a0foreign\u00a0institutional inflows and outflows, please refer to Part 1 first<\/a><\/em><\/strong><\/blockquote>\n

Backtesting Framework and Parameter Settings<\/h2>\n

Based on the factor\u2019s unique behavior, we transform it into actionable trading strategies. We embed explicit transaction costs, liquidity filters, and leverage caps into our event-driven engine, while\u00a0simultaneously\u00a0utilizing\u00a0a point-in-time architecture to\u00a0completely eliminate\u00a0look-ahead biase<\/strong>s. The detailed\u00a0backtesting\u00a0parameter configurations are outlined below:<\/p>\n

Backtesting Framework and Parameter Settings\u00a0<\/strong><\/a><\/h3>\n
    \n
  • Stock\u00a0Pool<\/strong>\uff1aAll\u00a0common stocks listed on the Taiwan Stock Exchange\u00a0(TWSE)\u00a0and Taipei Exchange\u00a0(TPEx)<\/li>\n
  • Period<\/strong>: December 2020\u2014May 2026.\u00a0September 30, 2024, is\u00a0designated\u00a0as the strict Out-of-Sample (OOS) cut-off point<\/li>\n
  • Rebalancing<\/strong>:\u00a0\u00a0Quarterly (Every 3 months)\u00a0to\u00a0minimize turnover\u00a0and mitigate\u00a0transaction frictions<\/li>\n
  • Initial capital<\/strong>: NT$10 million<\/li>\n
  • Transaction Costs<\/strong>\uff1a<\/li>\n
  • Buy: 0.1425%\u00a0commission<\/li>\n
  • Sell: 0.1425%\u00a0commission + 0.3% securities transaction tax<\/li>\n
  • Slippage Assumption<\/strong>\uff1a1\u00a0tick per transaction<\/li>\n
  • Leverage Constraint<\/strong>\uff1a0.9\u00a0(portfolio market value cannot exceed 90% of net asset value)<\/li>\n
  • Liquidity Overlays<\/strong>: Stocks that are locked at the limit price for the entire day\u00a0(Limited Whole Day)\u00a0or classified as Disposition Securities are automatically filtered out.<\/li>\n
  • Benchmark Index<\/strong>: Formosa Return Index\u00a0(IR0078).<\/li>\n<\/ul>\n

    Strategy Definitions: Pure Concentration vs. \u201cCross-Broker Consensus\u201d Optimization\u00a0\u00a0\u00a0<\/strong><\/h3>\n

    We construct two strategy variations for empirical comparison. Both strategies follow the exact same core routine (Industry Neutralization\u00a0\u2192\u00a0Market Cap Top 30% Large-Caps\u00a0\u2192\u00a0Select Top 50 Stocks\u00a0\u2192\u00a0Value-Weighted). They differ solely in the mathematical formulation of their final ranking scores.<\/p>\n

    The process of\u00a0industry neutralization<\/strong>\u00a0(subtracting the daily industry mean from an individual stock\u2019s factor score)\u00a0is vital to eliminate structural sector biases.\u00a0Without industry neutralization, concentration rankings would heavily over-concentrate in a handful of industries with naturally higher foreign broker coverage\u00a0(such as semiconductors); neutralization diversifies the portfolio\u2019s sector risk exposures, which ultimately boosts risk-adjusted returns<\/p>\n

      \n
    • Strategy 1\u00a0(Pure Concentration Strategy): Single-Factor Driven\r\n<\/strong><\/mark>Stocks are sorted directly by their industry-neutralized foreign institutional\u00a0concentration\u00a0cross-sectional z-scores, and the top 50 stocks are selected.\u00a0Ranking Score = z\u00a0(conc_qfii)<\/li>\n<\/ul>\n
        \n
      • Strategy 2 (Composite Fusion Strategy): Incorporating Broker Channel\u00a0Disagreement\r\n<\/strong><\/mark>Strategy 2 incorporates the\u00a0Broker Channel\u00a0Disagreement (Disagree)<\/strong>\u00a0factor to perform multi-factor fusion optimization.\r\nThe Broker Channel\u00a0Disagreement factor partitions the entire Taiwan brokerage landscape into three institutional channels:\u00a0Foreign Brokers (F)<\/strong>,\u00a0Private Domestic Brokers (P)<\/strong>, and\u00a0Government-Affiliated Brokers (G)<\/strong>. We first compute the net buy-sell ratio for each channel (daily net buy-sell amount \u00f7 individual stock daily total volume). After\u00a0winsorizing\u00a0the ratios at the 1%\/99% thresholds, we extract their cross-sectional z-scores. The\u00a0Disagree<\/em>\u00a0metric is defined as the cross-sectional standard deviation across these three channels:<\/li>\n<\/ul>\n

        Disagreei,t=13\u2211g\u2208{F,P,G}\u200b(zg,i,t\u2212z\u203ei,t)2,\u2003z\u203ei,t=13\u2211g\u200bzg,i,t<\/p>\n

          \n
        • High\u00a0Disagreement:<\/strong>\u00a0Indicates that the three broker channels are pulling in opposite directions (e.g., foreign brokers are buying aggressively, private domestic brokers are selling heavily, and government banks remain idle). Even if the foreign concentration is high, the directional signal is diluted due to intense confrontation.<\/li>\n<\/ul>\n
            \n
          • Low\u00a0Disagreement:<\/strong>\u00a0Indicates that the net trading directions of all three major broker channels are highly aligned\u00a0(synchronized buying or synchronized selling). This implies that the capital flows captured by foreign institutions are backed by a broader \u201ccross-channel market consensus,\u201d resulting in a cleaner and more reliable chip signal<\/li>\n<\/ul>\n

            Consequently, Strategy 2 defines its composite ranking score as:<\/p>\n

            Ranking Score = z (conc_qfii) \u2013 0.5 x z(Disagree)<\/p>\n

            The negative sign inside the formula penalizes high\u00a0disagreement. Strategy 2 aims to identify stocks that not only exhibit intense foreign concentration but also display low\u00a0disagreement (cross-channel consensus), explicitly steering clear of\u00a0stocks\u00a0prone to institutional tug-of-wars.<\/p>\n

            Table\u00a01: Factor Strategy Configurations<\/p>\n

            \n\n\n\n\n\n\n
            Strategy Name<\/th>\nStrategy Type<\/th>\nRanking Score Formulation<\/th>\n<\/tr>\n<\/thead>\n
            Strategy 1\u00a0(Pure Concentration)<\/td>\nSingle Factor<\/td>\nz(conc_qfii)<\/td>\n<\/tr>\n
            Strategy 2 (Concentration +\u00a0Disagreement Fusion)<\/td>\nMulti-Factor Fusion<\/td>\nz(conc_qfii) \u2212 0.5 \u00d7 z(Disagree)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
            Note: Both strategies implement the identical stock-picking routine and differ only in score metrics;\u00a0z(\uff0d) denotes the cross-sectional z-score after industry neutralization)<\/em>\u00a0<\/figcaption><\/figure>\n

            Backtesting Performance\u00a0<\/strong><\/h2>\n

            To protect our research from historical overfitting traps, we establish a demanding triad of performance hurdle rates: the Full-Sample Sharpe Ratio must exceed 1.256, the Out-of-Sample\u00a0(OOS)\u00a0Sharpe Ratio must exceed 1.668, and the annualized Alpha must be strictly greater than 0. These thresholds are set using the actual performance of the benchmark index\u00a0(IR0078)\u00a0over identical timeframes.<\/p>\n

            Table 2\u00a0and Figure 3 present the net performance metrics after accounting for all real-world transaction taxes, broker commissions, and slippage frictions:<\/p>\n

            Table 2:\u00a0Backtesting\u00a0Performance Metrics for\u00a0conc_qfii\u00a0Top 50 Large-Cap Value-Weighted Portfolio (Net of Costs)<\/strong><\/p>\n

            \n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
            Performance Metric<\/th>\nStrategy 1\r\n(Pure Concentration)<\/th>\nStrategy 2\r\n(Concentration +\u00a0Disagreement)<\/th>\nBenchmark Index\u00a0(IR0078)<\/th>\n<\/tr>\n<\/thead>\n
            Annualized Return<\/td>\n27.65%<\/td>\n30.12%<\/td>\n25.77%<\/td>\n<\/tr>\n
            Cumulative Return<\/td>\n260.67%<\/td>\n298.82%<\/td>\n233.51%<\/td>\n<\/tr>\n
            Annualized Volatility<\/td>\n21.59%<\/td>\n21.33%<\/td>\n19.83%<\/td>\n<\/tr>\n
            Sharpe Ratio (Full Sample)<\/td>\n1.239<\/td>\n1.342<\/strong><\/td>\n1.256<\/td>\n<\/tr>\n
            Sortino Ratio<\/td>\n1.868<\/td>\n2.022<\/td>\n1.799<\/td>\n<\/tr>\n
            Max\u00a0drawdown\u00a0(MDD)<\/td>\n\u221231.03%<\/td>\n\u221226.51%<\/td>\n\u221228.60%<\/td>\n<\/tr>\n
            Annualized Alpha<\/td>\n+1.71%<\/td>\n+3.88%<\/strong><\/td>\n\u2014<\/td>\n<\/tr>\n
            Beta<\/td>\n1.006<\/td>\n0.996<\/td>\n\u2014<\/td>\n<\/tr>\n
            Out-of-Sample Sharpe (OOS Sharpe)<\/td>\n1.969<\/strong><\/td>\n2.000<\/strong><\/td>\n1.668<\/td>\n<\/tr>\n
            Daily Average Turnover<\/td>\n0.74%<\/td>\n0.95%<\/td>\n\u2014<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
            Note: Benchmark is the Formosa Return Index IR0078; full data period: 2020\/12\u20132026\/05; Out-of-Sample\u00a0(OOS)\u00a0cut-off point: 2024-09-30<\/em>\u00a0<\/figcaption><\/figure>\n

            Figure 3: Cumulative Return Equity Curves of\u00a0conc_qfii\u00a0Top 50 Large-Cap Portfolio Variations vs. Benchmark Index (IR0078)<\/em>\u00a0<\/strong><\/p>\n

            \"\"<\/figure>\n

            Dissecting Strategy Performance :<\/p>\n

            Strategy 1 (Pure Concentration) Performance Review<\/h3>\n

            Relying solely on foreign institutional concentration rankings, Strategy 1 successfully satisfies two out of three criteria: first, its risk-adjusted performance during the Out-of-Sample (OOS) phase is outstanding, recording an OOS Sharpe Ratio of 1.969, beating the baseline hurdle of 1.668; second, it delivers a positive annualized Alpha of +1.71%. In terms of raw returns, its annualized return of 27.65% outperforms the benchmark\u2019s 25.77%.<\/p>\n

            Strategy 2\u00a0(Disagreement Fusion)\u00a0Optimization Impact<\/h3>\n

            When we overlay the broker\u00a0disagreement factor onto Strategy 1 to filter for \u201ccross-broker consensus,\u201d Strategy 2 achieves a clean triumph across all three hurdles:<\/p>\n