Buy cheap. Buy with momentum.
買得便宜，買得有動能。
Not just cheap — already moving.
每一季，我們掃描所有台灣上市股票，用 EBIT/TEV 找出被市場低估的公司——「以便宜的價格買進好的經營能力」。
Every quarter we screen TWSE-listed stocks for operational cheapness via EBIT/TEV, then layer a 12-month momentum filter to ensure the market is already recognising the value.
核心邏輯：找到市場還沒完全定價的公司，但價格已經開始動了。不是純粹等待價值回歸，也不是盲目追漲，而是兩者的交叉點。
The core idea: find companies underpriced relative to operating earnings, but already showing price strength. Not pure value. Not pure momentum. The intersection.
One asks how cheap. The other asks if it's moving.
EBIT（息前稅前盈餘）代表純粹的經營能力，排除融資方式與稅務結構的影響。TEV（企業總值）= 市值 + 負債 − 現金，是「買下整間公司的總代價」。
EBIT/TEV = how much operating profit per dollar of enterprise value. Higher = cheaper.
由 Joel Greenblatt 魔法公式衍生 · Derived from Greenblatt's Magic Formula
過去 12 個月的報酬率，跳過最後 1 個月（避免短期反轉效應）。已經在漲的股票，傾向繼續漲。
Past 12-month return, skipping the last month to avoid short-term reversal. The market is already confirming the thesis.
學術文獻驗證 · Documented since Jegadeesh & Titman 1993
Combined Signal
EBIT/TEV Top 30% ∩ 12-1 Momentum Top 30 = 便宜 × 已經在動 → 入選
Start with all of TWSE. Then filter ruthlessly.
Universe 母體
Taiwan Stock Exchange listed only — better liquidity, more reliable data
EV concept doesn't apply — debt is their business model, not a capital structure choice
Can't meaningfully rank EBIT/TEV on loss-makers
Execution Filters 執行過濾
20-day avg volume ≥ 1,000 lots — ensures we can actually execute at scale
Skip limit-up/down locked stocks on execution day → retry T+2, max 3 days
Conservative 60-day lag on financial data — zero lookahead bias
Two filters. Thirty stocks. One quarter.
Cost Structure 成本結構
Rebalance 換股節奏
每季末（3/6/9/12月）最後交易日 → 訊號日
下季第一個交易日開盤 → 執行日
Signal: Quarter-end close → Execute: T+1 open
Strategy vs 0050 ETF vs TWSE Index · 2014–2024 · Base = 1.0 at start of 2014
NAV Comparison · 淨值對比 (Base = 1.0 at start of 2014)
Year-by-Year · 逐年對比
462筆交易中的兩個極端 · 非對稱性是策略獲利的核心
勝率只有 49.6%（不到一半）——但策略仍然獲利，原因就在這裡：
最大贏家 +940% vs 最大輸家 -51%。贏的幅度是輸的將近 20 倍。
Win rate is under 50%, yet the strategy profits because the biggest winners are ~20× larger than the biggest losers.
最佳十筆 · Top 10 Winners
最差十筆 · Bottom 10 Losers
注意：此策略沒有個股停損設定，出場時間由季末換倉決定（持有到不再入選為止）。因此部分虧損筆數可以持續超過一季。最大單筆虧損 -51% 是因為連續兩季都在選股名單中。 · No individual stop-loss — exits are purely rebalance-driven. A stock stays in until it drops out of the top 30, so losses can compound across quarters.
TWD 1,000,000 → 4,338,608 · 2014–2024 · +334% Total Return
Backtest Architecture · 回測設計
Annual Returns · 年度報酬
Drawdown Profile · 回撤走勢 — how deep, how often
Year-end drawdowns shown. The −32.02% MDD reference line reflects intra-quarter data from the backtest engine — the worst trough within a calendar year exceeds what annual end-points alone capture.
Three honest observations from eleven years of data.