""" QQQM vs QQQ: fee, tracking, liquidity and what switching costs (FinLab research, 2026-11). QQQM listed on 2020-10-13. Both funds track the Nasdaq-100. Test (M1): buy QQQM instead of QQQ and hold. Nine settings, one for each pair of start 2020-10-30 / 2021-04-30 / 2021-10-29 and end 2025-10-31 / 2026-04-30 / 2026-10-30. The original is the longest window, 2020-10-30 to 2026-10-30. QQQM passes only if 1. at least 7 of the 9 windows have a higher daily Sharpe (no risk-free rate) than QQQ, 2. it beats QQQ in both halves of the original window, 3. a QQQ holder who pays 0.1% per side to switch to QQQM (0.2% in all) still beats QQQ: buy QQQ at the first close after the start, switch at the close after that (2020-11-03). The switch is repeated on the next six closes to show how much the fill date matters. Run: python research.py writes metrics.json and data.csv next to this file """ import json from datetime import date from pathlib import Path import finlab import numpy as np import pandas as pd from finlab import data from finlab.backtest import sim STARTS = ("2020-10-30", "2021-04-30", "2021-10-29") ENDS = ("2025-10-31", "2026-04-30", "2026-10-30") START, END = STARTS[0], ENDS[-1] LISTED = "2020-10-13" CONVERSION = "2025-12-22" # QQQ became an open-end ETF and its fee fell from 0.20% to 0.18% COST_PER_SIDE = 0.001 FEE_GAP_NOW = 0.0003 # 0.18% - 0.15% WINDOWS = {"1y": 252, "3y": 756, "5y": 1260} TRADE_SIZES = (10_000, 100_000, 1_000_000, 10_000_000) OUT = Path(__file__).resolve().parent FUNDS = ["QQQ", "QQQM"] data.truncate_end = END data.set_market("us") def table(key): frame = data.get(f"us_fund_price:{key}")[FUNDS] return pd.DataFrame(frame.loc[frame["QQQ"].notna()].loc[LISTED:]) PRICES = table("adj_close") DAYS = PRICES.index def daily_sharpe(returns): return returns.mean() / returns.std() * 252 ** 0.5 def cagr(curve): years = (curve.index[-1] - curve.index[0]).days / 365.25 return (curve.iloc[-1] / curve.iloc[0]) ** (1 / years) - 1 def compare(start, end): window = PRICES.loc[start:end] returns = window.pct_change().dropna() sharpe = {t: daily_sharpe(returns[t]) for t in FUNDS} growth = {t: cagr(window[t]) for t in FUNDS} return { "start": str(window.index[0].date()), "end": str(window.index[-1].date()), "qqqm_sharpe": round(sharpe["QQQM"], 6), "qqq_sharpe": round(sharpe["QQQ"], 6), "qqqm_cagr": round(growth["QQQM"], 6), "qqq_cagr": round(growth["QQQ"], 6), "cagr_gap": round(growth["QQQM"] - growth["QQQ"], 6), "qqqm_total": round(window["QQQM"].iloc[-1] / window["QQQM"].iloc[0] - 1, 6), "qqq_total": round(window["QQQ"].iloc[-1] / window["QQQ"].iloc[0] - 1, 6), "qqqm_mdd": round((window["QQQM"] / window["QQQM"].cummax() - 1).min(), 6), "qqq_mdd": round((window["QQQ"] / window["QQQ"].cummax() - 1).min(), 6), "beats_qqq": bool(sharpe["QQQM"] > sharpe["QQQ"]), } def run(position): curve = sim(position, trade_at_price="close", fee_ratio=COST_PER_SIDE, market="us_fund", upload=False).creturn moved = (curve != curve.iloc[0]).to_numpy().nonzero()[0] return curve.iloc[max(moved[0] - 1, 0):] def switching(fill): """Buy QQQ at the start, switch to QQQM at the close of days[fill], against staying in QQQ.""" days = DAYS[DAYS >= START] stay = pd.DataFrame({"QQQ": 1.0}, index=days) move = pd.DataFrame({"QQQ": 1.0, "QQQM": 0.0}, index=days) move.iloc[fill - 1:] = [0.0, 1.0] # a signal fills at the next close stay_curve, move_curve = run(stay), run(move) common = stay_curve.index.intersection(move_curve.index) stay_curve, move_curve = stay_curve.loc[common], move_curve.loc[common] sharpe = {k: daily_sharpe(c.pct_change().dropna()) for k, c in (("stay", stay_curve), ("move", move_curve))} stay_multiple, move_multiple = stay_curve.iloc[-1] / stay_curve.iloc[0], move_curve.iloc[-1] / move_curve.iloc[0] return {"filled": str(days[fill].date()), "stay_sharpe": round(sharpe["stay"], 6), "move_sharpe": round(sharpe["move"], 6), "stay_multiple": round(stay_multiple, 6), "move_multiple": round(move_multiple, 6), "lead": round(move_multiple / stay_multiple - 1, 6), "beats_qqq": bool(sharpe["move"] > sharpe["stay"])} def evaluate(): grid = [compare(s, e) for s in STARTS for e in ENDS] base = grid[2] days = PRICES.loc[START:END].index middle = str(days[len(days) // 2].date()) halves = [compare(START, middle), compare(middle, END)] switch = switching(2) wins = sum(g["beats_qqq"] for g in grid) checks = {"grid": wins >= 7, "halves": all(h["beats_qqq"] for h in halves), "cost": switch["beats_qqq"]} return {"grid": grid, "grid_wins": wins, "original": base, "halves": halves, "switching": switch, "switching_later_fills": [switching(fill) for fill in range(3, 9)], "checks": checks, "passes": all(checks.values())} def tracking(): window = PRICES.loc[START:END] returns = window.pct_change().dropna() gap = returns["QQQM"] - returns["QQQ"] curve = window / window.iloc[0] ends = window.resample("YE").last() yearly = ends / ends.shift(1).fillna(window.iloc[0]) - 1 monthly = window.resample("ME").last().pct_change().dropna() lead = curve["QQQM"] / curve["QQQ"] before, after = compare(START, CONVERSION), compare(CONVERSION, END) return { "before_conversion": before, "after_conversion": after, "yearly": {str(d.year): {"qqq": round(row["QQQ"], 6), "qqqm": round(row["QQQM"], 6), "gap": round(row["QQQM"] - row["QQQ"], 6)} for d, row in yearly.iterrows()}, "share_of_months_qqqm_ahead": round(((monthly["QQQM"] - monthly["QQQ"]) > 0).mean(), 4), "share_of_days_qqqm_ahead": round((gap > 0).mean(), 4), "correlation": round(returns.corr().iloc[0, 1], 6), "tracking_error": round(gap.std() * 252 ** 0.5, 6), "largest_daily_gap": round(gap.abs().max(), 6), "largest_daily_gap_date": str(gap.abs().idxmax().date()), "median_abs_daily_gap": round(gap.abs().median(), 6), "end_lead": round(lead.iloc[-1] - 1, 6), "lead_since_2021_04_30": round(lead.iloc[-1] / lead.loc["2021-04-30"] - 1, 6), } def holding_windows(): prices = PRICES.loc[START:END] out = {} for label, days in WINDOWS.items(): returns = (prices / prices.shift(days) - 1).dropna() gap = (returns["QQQM"] - returns["QQQ"]).set_axis(prices.index[:-days]) worst = gap.nsmallest(3) out[label] = {"windows": len(gap), "qqqm_beats": round((gap > 0).mean(), 4), "median_gap": round(gap.median(), 6), "worst_gap": round(gap.min(), 6), "worst_starts": {str(d.date()): round(v, 6) for d, v in worst.items()}} return out def liquidity(): close, volume = table("close").loc[START:END], table("volume").loc[START:END] high, low = table("high").loc[START:END], table("low").loc[START:END] value = close * volume latest = value.iloc[-252:] returns = PRICES.loc[START:END].pct_change().iloc[-252:] median = latest.median() yearly = value.groupby(value.index.year).median() return { "yearly_median_dollar_volume": {str(y): {k: round(v) for k, v in row.items()} for y, row in yearly.iterrows()}, "yearly_ratio": {str(y): round(row["QQQ"] / row["QQQM"], 2) for y, row in yearly.iterrows()}, "last_year": { "median_dollar_volume": {k: round(v) for k, v in median.items()}, "median_shares": {k: round(v) for k, v in volume.iloc[-252:].median().items()}, "lowest_dollar_volume": {k: round(v) for k, v in latest.min().items()}, "price_move_bp_per_million": {k: round(v, 5) for k, v in ((returns.abs() / latest).median() * 1e6 * 1e4).items()}, "median_high_low_range": {k: round(v, 6) for k, v in ((high - low) / close).iloc[-252:].median().items()}, }, "price": {k: round(v, 2) for k, v in close.iloc[-1].items()}, "one_cent_in_bp": {k: round(0.01 / v * 1e4, 3) for k, v in close.iloc[-1].items()}, "trade_as_share_of_median_day": {str(size): {k: round(size / v, 8) for k, v in median.items()} for size in TRADE_SIZES}, } def distributions(): rows = pd.DataFrame(data.get("us_fund_corporate_actions")) rows = rows[rows["stock_id"].isin(FUNDS) & (rows["date"] >= START) & (rows["date"] <= END)].copy() rows["lag"] = (pd.to_datetime(rows["payment_date"]) - pd.to_datetime(rows["date"])).dt.days rows["era"] = np.where(rows["date"] <= CONVERSION, "before", "after") close = table("close").loc[START:END] last_year = rows[rows["date"] > pd.Timestamp(END) - pd.DateOffset(years=1)] return { "days_ex_date_to_payment": {f"{k[0]}_{k[1]}": round(v, 1) for k, v in rows.groupby(["stock_id", "era"])["lag"].median().items()}, "trailing_yield": {t: round(last_year.loc[last_year["stock_id"] == t, "dividend"].sum() / close[t].iloc[-1], 6) for t in FUNDS}, "same_ex_dates": set(rows.loc[rows["stock_id"] == "QQQ", "date"].dt.date.astype(str)) == set(rows.loc[rows["stock_id"] == "QQQM", "date"].dt.date.astype(str)), } def break_even(): switch_cost = COST_PER_SIDE * 2 return {"switch_cost": switch_cost, "fee_gap_now": FEE_GAP_NOW, "years_at_fee_gap": round(switch_cost / FEE_GAP_NOW, 1)} def main(): out = {"window": {"listed": LISTED, "start": START, "end": END, "conversion": CONVERSION}, "run_date": str(date.today()), "finlab_version": finlab.__version__, "pandas_version": pd.__version__, "test": evaluate(), "tracking": tracking(), "windows": holding_windows(), "liquidity": liquidity(), "distributions": distributions()} out["break_even"] = break_even() print("QQQM", "PASS" if out["test"]["passes"] else "FAIL", out["test"]["grid_wins"], out["test"]["checks"]) window = PRICES.loc[START:END] curve = window / window.iloc[0] frame = curve.rename(columns={"QQQ": "QQQ total return", "QQQM": "QQQM total return"}) frame["QQQM lead over QQQ"] = curve["QQQM"] / curve["QQQ"] - 1 value = (table("close") * table("volume")).loc[START:END] frame["QQQ dollar volume"] = value["QQQ"].round() frame["QQQM dollar volume"] = value["QQQM"].round() frame.round(6).to_csv(OUT / "data.csv", index_label="date") (OUT / "metrics.json").write_text(json.dumps(out, indent=1, ensure_ascii=False)) if __name__ == "__main__": main()