---
title: Does high funding mean price will drop? Hyperliquid data
url: https://comparand.xyz/blog/does-high-funding-predict-price-drop
description: On Hyperliquid, price fell the next day after 53.37% of high-funding days, versus 52.72% of all days. See low funding, 3 days and fees.
updated: '2026-10-21'
category: Data
---

# Does high funding mean price will drop? Hyperliquid data

On Hyperliquid, price fell the next day after 53.37% of high-funding days, versus 52.72% of all days. See low funding, 3 days and fees.

No clear link showed up. On Hyperliquid, price fell the next day after 53.37% of a coin's highest-funding days, with a median move of −0.40%. After all days in the same sample it fell 52.72% of the time, with a median move of −0.28%. The gap between those two is smaller than the noise in the data, and the lowest-funding days were followed by falls even a little more often.

The test covers 18 Hyperliquid coins (not Bitcoin, Ethereum, Solana or HYPE), 5,888 coin-days, from May 2025 to September 2026. These are past cases, not a forecast.

## What does funding actually measure?

A **perp** is a futures contract that never expires. To keep its price near the coin's real price, traders on one side pay the other side a small **funding** payment. On Hyperliquid, funding is paid every hour. When the rate is positive, people holding bets on a rise (longs) pay people holding bets on a fall (shorts).

A high positive rate can mean many traders lean one way. The idea behind the question is that a crowded bet is fragile, so price might drop. That is a fair idea to test. It is not a fact until the numbers show it.

*Diagram: Illustration only, with no rates: longs paying shorts when funding is above zero and shorts paying longs when it is below zero, with an hourly payment arrow.*
## How did we decide what counted as "high"?

A funding rate that is high for one coin may be ordinary for another. So each coin was compared only with its own past.

For each coin and each UTC day, we took the average of its hourly funding rates. Then we ranked that day against the same coin's earlier days only, never later ones. A coin needed at least 60 earlier days before its days were ranked. A day in the top 10% of its own history was "high funding". A day in the bottom 10% was "low funding". The other 80% were "middle" days. Many hours sat at the same small baseline rate, so many days tied in the ranking. That is why the high and low groups below are not the same size.

Using only earlier days matters. If you rank a day against the whole history, you quietly use the future, and the test looks better than it could have been in real time.

## What happened to price the next day?

We measured each coin's change from the close of the funding day to the close of the next day. "Fell" means a strictly lower price. Moves are before fees. The **median** is the middle result when you sort all the moves from lowest to highest.

| Funding day type | Coin-days | Median next-day move | Share that fell |
| --- | --- | --- | --- |
| Top 10% (high) | 386 | −0.40% | 53.37% |
| Middle 80% | 4,722 | −0.23% | 52.16% |
| Bottom 10% (low) | 780 | −0.67% | 55.77% |
| All ranked days | 5,888 | −0.28% | 52.72% |

Two things stand out. First, most of the table is negative, even for ordinary days. In this sample, ordinary days were down a little more often than up, so "price fell the next day" happened about 53 times in 100 with no help from funding. That is the base rate, and any claim has to beat it.

Second, the low-funding days were followed by falls more often than the high-funding days. If high funding warned of a drop, low funding should not have done the same.

The 386 high days fall on 163 different dates, and the 780 low days on 337. Coins often move on the same days, so the number of truly separate cases is smaller than the counts suggest.

## Is the gap bigger than chance?

We also compared average moves, giving each coin equal weight. The average next-day move after high-funding days was −0.23%, and after middle days it was +0.04%. The difference of −0.27 percentage points came with a margin of error (standard error) of 0.47. When a gap is smaller than its margin of error, the test cannot tell it apart from chance.

The low-funding group had an average of −0.30%. Its gap against the middle days was −0.34 points, with a margin of error of 0.40. Again, smaller than the noise.

## What about the next three days?

Over three days, the numbers looked like this.

| Funding day type | Coin-days | Median 3-day move | Share that fell |
| --- | --- | --- | --- |
| Top 10% (high) | 379 | −1.06% | 55.41% |
| Middle 80% | 4,692 | −0.72% | 54.88% |
| Bottom 10% (low) | 785 | −1.61% | 56.69% |
| All ranked days | 5,856 | −0.81% | 55.16% |

Treat these as description only. Three-day windows from neighbouring days overlap, so the margin of error is too small, and the table should not be read as stronger evidence than the one-day table.

## Did it differ by coin or by year?

Yes, coin by coin the picture was mixed. Of the 14 coins with at least 10 high-funding days, the share that fell next day was higher than that coin's middle-day share in 6, equal in 1 and lower in 7. ALT fell after 72.7% of its high days (11 days), while APT fell after 40.0% (30 days). Samples that small can look striking by chance.

By year, the median next-day move after high-funding days was −0.49% in 2025 (205 coin-days) and −0.26% in 2026 (181 coin-days). Middle days had medians of −0.20% and −0.24%. So 2025 had a bigger gap and 2026 almost none, and neither year is large enough to prove anything alone.

## What if you traded on it, with costs?

This is a test of the idea, not advice. We imagined a short at the close of each high-funding day, held for one day. We charged a 0.05% fee on each side. A short on a positive-funding day receives the funding payment, so we added that too.

- 386 short trades: average +0.23% before costs, +0.18% after fees and funding, with a margin of error of 0.47. The share of winning trades was 53.11%.
- 780 trades going long after low-funding days: average −0.17% after costs, with 43.85% winning. Funding was often below zero on those days, so these longs collected 0.24% in funding on average. The price falls cost more than that.

The short result is less than one margin of error from zero. Counting the funding hour a little differently moved it to +0.26%. Neither version is a result you could count on. The long version lost money.

These figures leave out slippage and spread, and they cover mid- and small-size coins, not Bitcoin, Ethereum, Solana or HYPE, whose daily closes were not in the test. Do not read them as covering those coins.

New to funding? Start with [funding rates explained](https://comparand.xyz/blog/crypto-funding-rate-explained). For how often Hyperliquid funding turns negative, see [Hyperliquid funding rate history](https://comparand.xyz/blog/hyperliquid-funding-rates-data). For what holding a long costs in funding, see [what holding a HYPE perp cost](https://comparand.xyz/blog/hype-funding-cost-hyperliquid).

## How we measured this

We used Hyperliquid perps from May 31, 2025 to September 30, 2026. Of 20 coins we requested, 18 had enough daily closes: AI and BADGER did not. We needed 20 or more hourly funding values for a day to count it. A daily close counted only if the day's last recorded minute was 23:59 UTC. Each coin counts equally in the pooled averages. Margins of error are grouped by day, since coins move together. The period is about 18 months, so one bull or bear phase can shape the results.

## Try this in Comparand

With Comparand connected to your AI app, you can repeat parts of this on any Hyperliquid coin:

- "Show hourly funding for one coin for September 2026." hl_funding returns the funding rates for one coin over a window you choose.
- "Show hourly price bars for the same coin and dates." hl_bars returns price bars for a window. The default bar is one minute, and you can ask for a different interval.
- "Show per-minute labels for that coin around a day of high funding." hl_labels returns per-minute labels for one coin over a start and end time in the archive, and you can pick which fields.

On a plan, these data tools are unlimited under fair-use limits. On pay as you go, each of these three costs $0.005 a call. Your AI app can then do the ranking and counting steps for you. For funding on other exchanges, the free venue_funding tool returns settlements for one perp, up to 400 days per call.

You can [try 3 coins free](https://comparand.xyz/try), see the [docs](https://comparand.xyz/docs), or check [pricing](https://comparand.xyz/pricing).

## The short version

- After a coin's highest-funding days on Hyperliquid, price fell the next day 53.37% of the time, against 52.72% for all days.
- Lowest-funding days were followed by falls 55.77% of the time, so funding level did not separate rises from falls.
- The average gap of −0.27 points came with a margin of error of 0.47, which is smaller than the noise.
- A one-day short after high funding averaged +0.18% after fees and funding, still within the noise.
- Past cases, not a forecast, from 18 coins other than Bitcoin, Ethereum, Solana and HYPE, over about 18 months.

## Questions people ask

**Does a high funding rate mean the price will drop?**

Not clearly. On 18 Hyperliquid coins, price fell the next day after 53.37% of each coin's top-10% funding days. After all days it fell 52.72% of the time. The gap is small next to the noise, so the test does not show that high funding leads to a drop.

**What does high funding mean in crypto?**

Funding is a small payment between traders on a perp, a futures contract that never expires. When funding is positive, traders betting on a rise pay traders betting on a fall. High funding means that payment is large compared with the coin's own past.

**What happened after the lowest funding days?**

Price fell the next day after 55.77% of the bottom-10% days, with a median move of -0.67%. That is a few more falls than after the highest-funding days, so the funding level did not split rises from falls in this sample.

**Could you profit by betting against high funding?**

In this sample, a one-day short after top-10% funding days averaged +0.18% per trade after fees and funding, with a margin of error of 0.47. That is smaller than the noise. It is a past test, not a trading idea.