Disclaimer: This article is compiled from public data and historical backtests; it is not investment advice. Historical results do not guarantee the future, and platform fees, market rates, and taxes all change actual net returns. Assess your own situation and consult a professional if needed.
How did Bitfinex lending actually do in 2025? This backtests USD and USDT lending using 2025 full-year data, with a 2023–2025 sample for the longer view, comparing the APEX III engine against a fixed-2-day baseline and rule-based strategies across different market phases.
Scope: 2025 is the primary sample; 2023–2025 provides the longer track record. The full 2019–2026 eight-year backtest is covered separately. All APEX III figures use the same standard: an 8%–16% backtest range across both currencies, averaging about 11% after the 15% platform fee, on the 2023–2025 sample.
1. 2025 market recap: not a straight line
To read the lending market you first read the price action behind it. For BTC, 2025 was a year of “big rally, sharp pullback, divergence” — quite unlike the one-directional moves of the prior two years. That mattered for lending, because funding demand surges exactly when traders pile into leverage, and fades in quiet stretches. The result was a year with both rich and lean periods rather than a single steady rate.
2. The 2025 lending market: USD and USDT highs and lows
Across 2025, USD and USDT funding rates both ranged widely — spiking several-fold during volatile, high-demand windows and falling back toward base levels in calm periods. USD generally sat a few points above USDT, while USDT’s larger swings created more room for active term selection. Neither currency paid a flat rate; the everyday reality was movement.
3. The backtest setup: sample, data, method
The backtest uses hourly snapshots of Bitfinex’s public lending rates as the input, and simulates each strategy under the same capital, the same timeline, and the same fee assumptions so the comparison is apples-to-apples. The 15% platform fee is deducted per fill, so every annualized number quoted is net — what would actually reach a lender’s wallet. The baseline is a mechanical “roll 2-day offers” strategy; the rule-based comparison posts at simple fixed rules. APEX III is the engine under test.
4. APEX III vs baseline vs rules: month-by-month and quarter-by-quarter
Compared month by month and quarter by quarter over the sample, APEX III led the fixed-2-day baseline across the board, and beat the simple rule-based strategies in the large majority of periods. The edge wasn’t uniform — it was wider in volatile, high-rate stretches (where catching highs and avoiding idle time matters most) and narrower in flat, low-rate periods (where there’s simply less to optimize). Importantly, it led consistently rather than winning one big month and lagging the rest.
5. Key 2025 numbers
After the 15% fee, on the 2023–2025 sample, combined USD+USDT annualized averaged about 11%, in a range of roughly 8%–16% (on 180/365-day holding). The fixed-2-day baseline ran about 6%–7% long-run. So the headline isn’t a flashy peak — it’s a steady multiple of the do-nothing baseline. Systemic-event windows briefly showed extreme highs near 80% APR, but those are exceptions and excluded from the headline figure.
6. Bull/bear sub-samples: stability across regimes
Splitting the sample into bear, range, and bull segments, APEX III stayed ahead of the baseline in each. The source of that stability is worth understanding: it doesn’t come from predicting spikes, but from reducing idle time and avoiding the drag of sitting at the bottom of the rate range. In a bear market it can’t conjure high rates that the market isn’t paying — but it loses less to idleness, which compounds over a year.
7. Win/loss analysis: when APEX III led, lagged, or tied
No strategy wins every month. APEX III led in most periods, tied in flat stretches where there’s little to gain, and occasionally lagged a pure fixed-rate bet during a sustained one-directional rate climb (where locking long early happens to win). Across the full sample the expected-value edge was clear and consistent — which is the point of a backtest, versus any single lucky month.
8. Conclusion: what a backtest can and can’t tell you
A backtest can show whether a strategy had a real, repeatable edge over a fair baseline across different regimes — and here it did, after fees. A backtest cannot promise that edge continues: rates could structurally decline, ETFs could permanently change behavior, or liquidity could shift. So read this as evidence that automation reduced idle time and added net yield historically — not as a forecast. Past performance does not guarantee future results, and your principal stays exposed to platform and stablecoin risk throughout.
FAQ
How is the 15% fee applied? Deducted per fill; the ~11% average is already net of it.
Is 2023–2025 a fair sample? It spans bear, range, and bull — more representative than any single year.
Why not headline a peak? Peaks (near 80% in systemic events) aren’t normal; a multi-year average is honest.
Will the return continue? No guarantee — past performance doesn’t predict the future.