---
title: What is a chart analog? A guide with real Bitcoin data
url: https://comparand.xyz/blog/what-is-a-crypto-chart-analog
description: A chart analog is a past chart shaped like today's. What bars, windows and shapes mean, how matches are picked, and how a Bitcoin analog rule did.
updated: '2026-10-21'
category: Basics
---

# What is a chart analog? A guide with real Bitcoin data

A chart analog is a past chart shaped like today's. What bars, windows and shapes mean, how matches are picked, and how a Bitcoin analog rule did.

A chart analog is a past stretch of a coin's prices whose shape looks like the chart you are studying, so you can see what the price did after it. It is built from bars (one candle each), a window (how many bars you compare) and a shape (the window put on a common scale so only its form counts). In our Binance Bitcoin (BTCUSDT) test at 300 times, with 48 hourly closes per chart, a rule that followed the ten closest past matches got the next day's direction right 50.67% of the time, the same as always guessing down, and it lost money after fees.

So an analog is a way to look at past cases, not a forecast. Below, we walk through how one is built and what our test found.

For how often the ten matches disagree, see [Similar crypto charts, different outcomes](https://comparand.xyz/blog/when-past-crypto-cases-disagree).

## What does a chart analog compare?

You start with a coin, an exchange and a stretch of its price history. An exchange is a place where people trade. Comparand searches that same coin's past on that same exchange: a Bitcoin search on Binance returns Bitcoin cases from Binance.

Our example uses Binance's BTCUSDT perpetual market. A perpetual, or perp, is a contract tied to a coin's price, with no end date. Here, Bitcoin's price is shown in USDT, a crypto coin built to track the US dollar.

The search does not need a named pattern, such as a triangle. It compares the price changes through the whole chosen stretch. You can find past cases even when you have no name for the chart in front of you.

On Hyperliquid, hl_analogs also checks buying, selling, open interest (the total size of trades still open) and liquidations (forced trade closures when traders run out of money).

## What are a bar, a window and a shape?

A **bar** records the opening, highest, lowest and closing price during a set time. An hourly bar covers an hour. A completed bar has finished, so its closing price is known.

The **window** is the stretch you choose to compare. In our test, each window held 48 completed hourly closing prices. The bar size sets how much time each bar covers; the window length sets how many closing prices you compare.

With shape_analogs, you choose the bar size and window length before comparing past shapes.

To build a **shape**, our test measured the changes between closing prices. It then adjusted those changes for their average and the size of their swings within each window. This put each chart on a common scale, a step called **normalization**.

After that step, charts from different price levels can look alike. The shape no longer tells you the average change or the size of the swings. Similar shapes can come from very different prices and market conditions.

*Diagram: Show a recent Bitcoin window beside a past window, then show their price changes on a common scale with the later outcomes kept separate.*
## How does the search choose its closest matches?

The search gives each past window it can use a **distance** score. This measures how much its shape differs from the chosen chart. A smaller score means a closer shape; it does not tell you the chance that price will rise.

Our Binance Bitcoin test compared the adjusted price changes at matching points in each window. Bigger differences added more to the score. At each of the study's 300 times, it ranked past windows by that score and kept ten separate matches.

Nearby windows share much of the same price history. Counting them all could let one past event fill the list. Our test kept windows that did not share bars, and every match finished before the chart being studied began.

“Closest” still means closest among the available past windows. It does not mean identical. The [methods page](https://comparand.xyz/methods) explains how Comparand builds its searches.

## What happened after the Bitcoin analogs?

After choosing matches by shape, we examined what followed each one. A **return** is the percentage price change over a chosen time. We used the **median** return: the middle value after sorting the returns from lowest to highest, averaging the central pair when needed.

For this test, a median above zero meant choosing up; a median below zero meant choosing down. We checked that choice against the actual later move. Comparand's analog tools show each past case separately; they do not calculate this combined result.

The table shows Binance BTCUSDT, using 48 hourly closes at 300 test times from 1 January 2020 through 29 September 2026. “Always up” shows how often price rose at those same times. The final column shows the average result of an imaginary trade in the median's direction, after fees.

| Time after decision | Analog direction correct | Always up correct | Average test return 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% |

In this same sample, always choosing down also got the next day's direction right 50.67% of the time. The analog rule did no better than that simple comparison. At the other times shown, it either tied or did worse than always choosing up. For context, across every hour in the test period, Bitcoin was higher 24 hours later 52.34% of the time.

It also lost money on average after fees at every time shown. A correct call can bring a small gain, while a wrong call can bring a larger loss. Our [full look-alike test](https://comparand.xyz/blog/do-chart-lookalikes-predict-crypto) explains this, and [Bitcoin after its closest past charts](https://comparand.xyz/blog/bitcoin-after-closest-past-charts) shows real cases.

## What if the past cases disagree?

They often did. In the same Binance Bitcoin test, 65.00% of the 300 searches had four to six rising cases among the ten matches after 24 hours.

That is a close split. Similar shapes frequently led to different next-day outcomes. A single rising example would hide the falling cases beside it.

When reading results, look at each price path and its later move. Ask whether the cases agree, how much the moves differ and whether a single large move changes your impression.

A check you can use in your own research: count how many of the ten cases rose, then compare that with how often the coin rose in any hour. If four to six of ten rose, and the coin rises about half the time anyway, the matches are telling you little beyond the usual up-rate. Our guide to [disagreement among past cases](https://comparand.xyz/blog/when-past-crypto-cases-disagree) explains why the split matters.

## Does a longer chart find the same past cases?

Often it does not. We repeated the Binance BTCUSDT search with 24, 48 and 96 hourly closes at the same 300 test times, from 1 January 2020 through 29 September 2026.

We counted a match as shared when it ended at exactly the same past time in both searches. Different window lengths mean different starting times. The table shows what share of the ten matches had shared end times, averaged across the searches.

| Window lengths compared | Average share with shared end times | Searches with no shared end times |
| --- | ---: | ---: |
| 24 vs 48 hourly closes | 15.57% | 41.33% |
| 24 vs 96 hourly closes | 5.67% | 69.00% |
| 48 vs 96 hourly closes | 19.57% | 35.00% |

A longer window includes more of what happened before the final price. It also changes the spacing needed to keep matches from sharing bars. Both changes affect these results, so the table does not prove that any length is best.

Choose the stretch you want to study before reading the later outcomes. Repeatedly changing the length can hide how much the answer depends on that choice. See [how chart length changes matches](https://comparand.xyz/blog/chart-window-length-matters) for more detail.

*Diagram: Show where the past Bitcoin matches end for short and long windows, highlighting the end times both searches share.*
## How we measured this

We used Binance BTCUSDT perp hourly bars from 8 September 2019 up to, but not including, 1 October 2026. We tested 300 times spread across 1 January 2020 through 29 September 2026, keeping them fixed across window lengths. Each search used completed closes and selected ten separate past charts whose later outcomes were already known by that test time. We compared the direction of their median return with the actual later move. Imaginary trades paid 0.05% to enter and 0.05% to exit, but left out ongoing perp payments, delays and differences between quoted and actual trade prices. These were separate trades, not a test of a growing account, and some past cases shared later price history.

## Try this in Comparand

In an AI app connected to Comparand, ask: “Use shape_analogs to compare Binance BTCUSDT's latest 48 completed hourly bars with its own past.” You get:

- The 10 closest past Bitcoin windows from Binance by default (you can ask for 1 to 20), on the same bar size, one per episode, none overlapping your chart.
- Each case's later moves after 12, 24 and 48 hours, in basis points (one hundredth of a percent).
- Data sources and call cost: $0.01 from prepaid credits.

Ask: “Use hl_analogs for Hyperliquid Bitcoin at 12:00 UTC on 1 September 2026, with a four-hour outcome.” You get:

- The 20 nearest past Bitcoin moments on Hyperliquid by default, matched on buying, selling, open interest and liquidations as well as price.
- Each case's next four-hour outcome; about $0.01 per search from credits.

Neither tool returns an average or hit rate. See the [docs](https://comparand.xyz/docs), or [try 3 coins free](https://comparand.xyz/try).

## The short version

- A chart analog is a similar past stretch of the same coin on the same exchange.
- A closer shape can still lead to a different later move.
- In our Binance Bitcoin test, the rule did no better than always choosing down for the next day. It lost money after fees.
- Changing the window often changed the matches. Pick the question first, then inspect all the cases.

## Questions people ask

**What is a chart analog?**

A chart analog is a past stretch of prices that looks like the chart you are studying. Comparand searches the same coin on the same exchange and shows what happened after each match.

**Does a similar Bitcoin chart predict the next move?**

Not in our test. On Binance BTCUSDT at 300 times from 2020 to 2026, following the ten closest matches got the next day's direction right 50.67% of the time, the same as always choosing down, and lost 0.30% per trade on average after fees.

**Does changing the chart length change the matches?**

Yes. Including more price history changes the shape you compare and can change which past cases look closest. Our Binance Bitcoin test often found different matches when we changed the chart length.

**How do shape_analogs and hl_analogs differ?**

shape_analogs compares price shapes for big coins on Binance, Bybit, OKX and Coinbase, plus Hyperliquid perps. hl_analogs works on Hyperliquid only and also checks buying, selling, open interest and liquidations. Both search only the same coin's past on the same exchange.