Research methods
The report separates observed fund history from simulations, compares products only within compatible cohorts, and preserves known data breaks rather than smoothing them away.
Actual results show observed security history, while modeled results are simulations with explicit assumptions.
Actual results use that security's observed history. Modeled results apply explicit daily leverage, financing, fee, rebalance, and return-floor assumptions. Modeled series remain separate from live fund history and are excluded from live-history rankings.
Adjusted close incorporates reported splits and cash distributions, but it does not capture investor-level trading costs.
Returns use vendor-adjusted close at the daily session. They omit intraday execution, bid-ask spread, slippage, premium/discount, investor taxes, and account-level cash flows; null sessions are dropped, not filled forward.
A 3× daily objective does not promise three times the benchmark's return over a longer holding period.
Multi-day returns compound from the sequence of daily moves. Persistent trends can help; volatile chop can hurt. The route, financing, and rebalance all matter, so 'decay' is neither constant nor guaranteed.
The headline expense ratio or investor fee may represent only part of the product's economic cost.
Displayed fees come from current disclosures. Derivative financing, ETN financing formulas, futures roll and basis, spreads, transaction costs, and taxes may sit outside that number or flow through NAV.
Performance rankings are most informative when products share an objective family, wrapper, and common date window.
Rankings require the same objective family and a common date window. Daily-reset index ETFs, ETNs, single-stock and crypto products, inverse funds, commodities, and stacked portfolios stay in separate cohorts.
Keeping closed products and security-identity changes visible reduces survivorship bias in the reported results.
Closed, called, or replaced products remain visible as history. CUSIP changes, ticker reuse, leverage or benchmark changes, missing sessions, and short records are flagged; broken series are excluded from rankings.