---
title: 'Crypto base rates: how often does Bitcoin go up?'
url: https://comparand.xyz/blog/past-crypto-returns-not-forecasts
description: Binance Bitcoin rose in 51.88% of 24-hour periods, and 10 big coins ranged from 48.39% to 53.84%. Why a base rate comes before any pattern claim.
updated: '2026-10-21'
category: Basics
---

# Crypto base rates: how often does Bitcoin go up?

Binance Bitcoin rose in 51.88% of 24-hour periods, and 10 big coins ranged from 48.39% to 53.84%. Why a base rate comes before any pattern claim.

A base rate is how often price rose anyway, before you pick any pattern. On Binance BTCUSDT, Bitcoin ended higher 24 hours later in 51.88% of 62,335 hourly starts from September 2019 to October 2026, and after 12 and 48 hours in 51.30% and 52.23%. Across ten big Binance coins the 24-hour rate ran from 48.39% (ADA) to 53.84% (TRX), and Bitcoin's own rate swung from 58.03% in 2020 to 46.88% in 2022.

So any pattern claim has to beat a number like this, not 50%. These are past cases, not a forecast.

For whether chart look-alikes beat these base rates, see [Do chart patterns predict crypto prices?](https://comparand.xyz/blog/do-chart-lookalikes-predict-crypto).

## What does a crypto base rate tell you?

A **base rate** is how often something happened before you picked a special pattern. Here, it means how often the price ended higher after a set wait. A **return** is the percent change from the start price to the end price.

Say you find past charts that mostly ended higher. First ask how often that market rose anyway, without picking those charts. Otherwise, you may credit the pattern for something the market did most of the time.

We measured base rates on Binance with hourly bars. A **bar** records prices during a set time. These markets are **perpetual futures** (perps): contracts tied to a coin's price that never expire. They are priced in USDT, a token designed to track the dollar.

Each cell shows the share of periods that ended higher, then how many complete periods we counted. Each period starts at an hourly close. Unchanged prices do not count as rises. All histories end on October 20, 2026, at 00:00 UTC.

| Binance market and history start, UTC | Up after 12 hours | Up after 24 hours | Up after 48 hours |
| --- | --- | --- | --- |
| Bitcoin, BTCUSDT: September 8, 2019, 17:00 | 51.30% (62,347) | 51.88% (62,335) | 52.23% (62,311) |
| Ethereum, ETHUSDT: November 27, 2019, 07:00 | 51.47% (60,437) | 52.03% (60,425) | 52.54% (60,401) |
| Solana, SOLUSDT: September 14, 2020, 07:00 | 49.42% (53,429) | 49.66% (53,417) | 49.63% (53,393) |

All nine numbers sit within about 3 points of a coin flip. Bitcoin and Ethereum rose slightly more often; Solana slightly less. Their histories start on different dates, so this table cannot tell how much of the difference comes from the coin and how much from the dates.

The same held for all ten big coins we measured on Binance. Their 24-hour up-rates ranged from 48.39% (ADA) to 53.84% (TRX).

## How large were the price moves?

For Bitcoin's 62,335 complete 24-hour periods on Binance BTCUSDT, the median move was +0.07%. The **median** is the middle result after sorting the moves from lowest to highest.

The cutoff for the lowest tenth of moves was −3.07%. The cutoff for the highest tenth was +3.49%. Some moves went further than both.

So an up-rate near 50% does not mean prices stayed still. It counts a tiny rise and a big rise the same way, and it says nothing about how large a loss could be.

*Diagram: Show the Binance Bitcoin daily move cutoffs and median from this section, with arrows showing that some moves went beyond both cutoffs.*
## Can fees erase a small price gain?

Yes. In those same 62,335 Bitcoin 24-hour periods, the share with a positive result fell from 51.88% to 49.30% after fees. The median result went from +0.07% to −0.03%.

This assumes a fee of 0.05% to enter and 0.05% to exit, at the start and end closes. These are test assumptions, not the fee you would pay. A price can end higher while the result after fees ends lower.

## Does the usual up-rate stay the same each year?

No, not even in one market. On Binance BTCUSDT, the 24-hour up-rate was 58.03% across 8,784 starting hours in 2020. It was 46.88% across 8,760 starting hours in 2022.

The full-history rate blends years like these. It does not tell you which year, if any, looks like today. Picking a good year after seeing its result does not test whether that choice would have helped at the time. Our [test of 1,000 random trading rules](https://comparand.xyz/blog/overfitting-1000-random-trading-rules) shows how choosing after the fact fools you.

## Did similar charts beat a plain guess?

A **chart analog** is a past chart with a similar price shape. At each of 300 test times, our rule took the last 48 completed hourly closes of Binance BTCUSDT and found the ten closest earlier Bitcoin charts on Binance that did not overlap each other. Their later moves were already known at the test time. If the median of those later moves was positive, the rule guessed up; if negative, down.

Here is how that rule did at each wait, at the same 300 times:

| Wait after test time | Rule picked the right direction | Price rose (always-up guess) | Average result per imagined trade, after fees |
| --- | ---: | ---: | ---: |
| 12 hours | 51.33% | 52.67% | −0.15% |
| 24 hours | 50.67% | 49.33% | −0.30% |
| 48 hours | 52.33% | 52.33% | −0.07% |

At 12 hours, the rule did worse than always guessing up. At 48 hours, it tied. At 24 hours, it beat always-up, but only matched always guessing down (50.67%). All three average results were negative after fees.

The cases also disagreed often. In 65.00% of the 12-hour tests, only four to six of the ten matches had risen. Similar shapes often had mixed later results.

The wider test gave the same answer. Across 12 markets and 3,600 test times, the rule picked the 24-hour direction right 50.22% of the time. Always guessing up was right 50.92% of the time. Our [broader chart look-alike test](https://comparand.xyz/blog/do-chart-lookalikes-predict-crypto) shows all 12 markets.

## Why aren't all these periods separate pieces of evidence?

Nearby periods share price history. A period starting at one hourly close and one starting an hour later can both include the same big fall. Counting both describes each start, but it counts that event twice.

The chart tests can reuse the same earlier matches, too. We also only used test times with complete later prices; in live use, you would not know that ahead of time. And these studies cover coins that are still trading, so failed coins are left out. Large counts alone do not prove a rate will hold.

## How can you check a pattern claim yourself?

Use one simple habit: put the base rate next to every pattern result.

1. Write down the coin, exchange, bar size, dates and wait.
2. Find how often price rose over that wait, for that same market and dates, with no pattern.
3. Compare the pattern's hit rate with both always-up and always-down. In our Binance BTCUSDT 12-hour test (300 times, 2020 to 2026), the chart rule was right 51.33% of the time, but always guessing up was right 52.67%, so the rule added nothing.
4. Then apply fees, and look at the size of the losing cases, not only the share of winners.

## How we measured this

We used completed hourly prices for the Binance perps and dates named above. For base rates, we compared each starting close with the close 12, 24 or 48 hours later, keeping only periods with every hourly price present. For the chart rule, we spread 300 test times across Bitcoin's eligible history and compared price-change shapes only with earlier windows whose later results were already known. The fee examples assume 0.05% to enter and to exit; they leave out funding payments between perp traders, borrowing, bid-ask gaps and slippage. Each imagined trade was scored on its own, so this is not a test of a real account over time.

## Try this in Comparand

Ask your connected AI assistant: "Find ten charts from Bitcoin's own Binance history that look like its latest 48 completed hourly bars. Show each case's path and later moves, including the ones that fell."

You get:

- **Chart cases (shape_analogs):** matches only from Bitcoin on Binance, with the same hourly bar size.
- **Later moves:** each case's path and price change after 12, 24 and 48 hours, in basis points (100 basis points = 1%).
- **No summary score:** these are past cases, not a forecast. The search returns no average, median or hit rate, so you do the counting and compare it with the base rate.
- **Cost:** $0.01 per search, paid from credits.

For Hyperliquid coins, you can also ask: "Show Bitcoin's minute-by-minute labels on Hyperliquid for the first hour of a complete archive day." **hl_labels** returns per-minute labels that describe the past, for one coin over the time range you choose. Each answer also lists the data behind it and what the call cost.

You can [try 3 coins free](https://comparand.xyz/try). The [documentation](https://comparand.xyz/docs) explains these requests. Our [guide to Bitcoin's similar charts](https://comparand.xyz/blog/bitcoin-after-closest-past-charts) shows why both rising and falling cases matter.

## The short version

- Bitcoin on Binance rose in 51.30% to 52.23% of 12- to 48-hour periods from 2019 to 2026: close to a coin flip.
- Our closest-chart rule never beat the better of always-up or always-down, and lost money on average after fees.
- Fees, move sizes and year-to-year swings matter as much as the up-rate.
- Always compare a pattern's result with the base rate for the same market and dates.

## Questions people ask

**Can past price predict crypto?**

Not reliably in our tests. On Binance BTCUSDT hourly data, a rule that followed the closest past charts picked the next 12-hour direction right 51.33% of the time, while always guessing up was right 52.67% of the time at the same 300 test times.

**What is a crypto base rate?**

It is how often something happened across a chosen history, before you pick a special pattern. For Binance BTCUSDT from 2019 to 2026, price was higher 24 hours later in 51.88% of hourly starting points.

**Does being right more often mean a trading rule makes money?**

No. Gains and losses can be different sizes, and fees can turn a small rise into a loss. With 0.05% fees each way, the share of positive Bitcoin 24-hour results fell from 51.88% to 49.30%.

**Does Comparand return a forecast or a win rate?**

No. Comparand returns individual past cases and what followed each one. Its chart search does not calculate an average, a middle result or a share of correct calls.