Bitcoin Funding Rates Are Predictable. Are They Profitable?

Predicting Bitcoin’s (BTC ) next price move is difficult. Predicting the cost of holding a leveraged Bitcoin position may be more achievable. That distinction matters because a trade can correctly anticipate market direction and still lose money through recurring funding payments and transaction fees.

A new study, “Predictability of Funding Rates,” by Emre Inan examines whether statistical models can forecast those payments on Binance and Bybit.¹ Published in the International Review of Financial Analysis, the research finds evidence of predictability, but its most useful lesson concerns execution: better forecasts become valuable only when they improve decisions after costs.

For investors, the broader implication extends beyond a particular trading strategy. As cryptocurrency markets develop, understanding their financial plumbing can matter as much as understanding the assets themselves.

How Bitcoin Perpetual Futures Funding Rates Work

Perpetual futures allow traders to take positions on Bitcoin without a conventional expiration date. Unlike buying Bitcoin directly, holding these contracts involves an additional mechanism designed to keep their prices aligned with the underlying spot market.

That mechanism is funding. When the funding rate is positive, traders holding long positions pay traders holding short positions. When it is negative, payments move in the opposite direction. These transfers create incentives that help pull the contract price toward spot prices.

The payment depends on the position’s notional value, meaning its total market exposure. For illustration, a 0.01% funding rate on a $100,000 position produces a $10 payment for that interval. Repeated payments can materially change a trade’s economics, particularly when leverage makes exposure much larger than the trader’s deposited capital.

Funding therefore acts as a variable holding cost or receipt. A positive rate does not guarantee that Bitcoin will rise, and receiving funding does not guarantee a profitable position. Price movements can outweigh the payments.

What the Bitcoin Funding Rate Study Found

Inan analyzes BTC/USDT perpetual funding rates from January 11, 2024, through March 17, 2025. Observations arrive every eight hours. The final 384 observations, covering November 10, 2024, onward, form the out-of-sample forecasting period.

The research uses double autoregressive models, or DAR models. In accessible terms, these models account for both persistence in funding rates and changes in their volatility. They are statistical forecasting tools, rather than evidence that an artificial intelligence system can predict cryptocurrency markets.

The comparison is a simple no-change forecast: assume the next funding rate equals the latest observed rate. Depending on the model’s estimation method, directional accuracy ranges from 48.70% to 60.16% on Binance and from 49.48% to 62.5% on Bybit.

There is an important qualification. Raw forecast errors do not consistently favor the DAR models. The paper’s stronger statistical conclusions emerge after applying an adjustment for parameter estimation noise. Readers should distinguish that statistical evidence from an unqualified claim that every model produces smaller practical errors.

Predictability also changes over time. The analysis finds greater variation in stability on Bybit, suggesting that a forecasting relationship can weaken precisely when market conditions become unsettled.

Why Trading Fees Reverse the Backtest Results

The study tests a cross-exchange strategy that holds opposing Bitcoin perpetual positions. It shorts the contract with the higher expected funding rate and buys the contract on the other exchange, seeking to capture the difference while offsetting much of the directional exposure.

The simulation starts with $100 million and applies five times leverage. One scenario assumes no trading fees. Another applies a 0.0090% fee on Binance and zero fees on Bybit. These are specific modeled conditions, rather than a representative fee schedule for every investor.

Without fees, the simple benchmark wins convincingly. With modeled fees, the DAR strategy produces the stronger return while executing fewer trades. Its maximum drawdown is also smaller in the fee-inclusive scenario.

These figures are annualized backtest metrics from a roughly four-month evaluation period. They are not returns observed over a full year, and they do not establish what a live strategy would earn today.

The Real Advantage May Be Knowing When to Wait

The most useful interpretation is that forecasting can improve selectivity. A trader does not need to exploit every expected funding difference. The difference must be large enough, and sufficiently persistent, to justify opening or changing positions.

This introduces a practical decision threshold. A modest expected advantage may disappear after entry fees, exit fees, bid-ask spreads, and execution slippage. A larger advantage can still be unattractive if the forecast is unusually uncertain.

Three questions help separate a useful signal from an expensive distraction:

• Does the expected funding difference exceed the full cost of execution?
• How sensitive is the expected return to forecast errors?
• Can both exchange accounts withstand adverse price movements?

The paper does not independently isolate the contribution of every avoided trade. Nevertheless, its results support a broader insight: evaluating forecasts solely by accuracy misses how they interact with trading rules. The best model for predicting a number may differ from the best model for deciding whether to commit capital.

Market-Neutral Bitcoin Trading Still Requires Risk Capital

Opposing positions reduce directional exposure, but they do not eliminate operational or financing risk. Each exchange maintains its own account balances and margin requirements. A gain on one venue may not be immediately available to support a losing position elsewhere.

Consequently, a combined portfolio can appear balanced while one account approaches liquidation. Transfer delays, collateral requirements, and temporary price differences can turn a theoretically offsetting trade into a liquidity problem.

The paper improves on an earlier approach by including exchange-specific price changes between consecutive funding periods. Even so, a period-by-period backtest cannot establish that every intraperiod margin demand would have been manageable in live trading.

Counterparty assessment also remains essential. Securities.io’s recent explanation of why proof of reserves is not an audit provides relevant background: evidence of assets alone does not establish an exchange’s complete financial condition. A funding advantage must compensate for more than market exposure.

Why Perpetual Futures Research Matters Beyond Crypto

Funding mechanisms are becoming relevant to a wider range of markets. Securities.io recently covered regulatory relief for true perpetual index futures, illustrating how the contract structure is extending beyond its cryptocurrency origins.

That does not mean Bitcoin forecasting results transfer automatically to equity or index products. Trading hours, reference prices, participant behavior, and funding rules can differ substantially. Each market requires separate testing.

It does suggest a growing need for tools that estimate total holding costs across venues. For a professional investor maintaining a hedge, improved funding forecasts could help determine where to place exposure or how long to retain it, even without operating a dedicated arbitrage strategy.

Coinbase Offers Exposure to Derivatives Infrastructure

For investors interested in the infrastructure supporting these markets, Coinbase provides a relevant publicly traded company to examine. Its connection is the operation and development of trading services, rather than ownership of the forecasting technique studied here.

In March 2026, Coinbase announced stock perpetual futures for eligible international customers, extending its derivatives offering beyond cryptocurrency.

More recently, Coinbase reported that its International Exchange and Deribit integration was completed on October 1, 2026. The consolidation brings together infrastructure supporting perpetuals and other products. Product availability remains subject to eligibility and jurisdiction.

Investors assessing this opportunity should examine trading activity, fee economics, liquidity, competition, and risk controls. Funding transfers between traders should not be confused with exchange revenue. Likewise, expanding a product lineup does not guarantee profitable growth.

Latest Coinbase (COIN) News and Developments

Better Funding Forecasts Need Better Execution

The study provides evidence that Bitcoin funding rates contain useful forecasting information. Its strongest practical message is that information becomes valuable through disciplined execution, realistic costs, and adequate capital.

That makes the research more useful as a framework for evaluating trading decisions than as a ready-made yield promise. In increasingly competitive derivatives markets, an advantage may come from identifying the trades whose expected returns survive real-world constraints.

References:

¹ Inan, E. (2026). Predictability of funding rates. International Review of Financial Analysis, 119, Article 105464. https://doi.org/10.1016/j.irfa.2026.105464