Blog

Finance YouTubers by accuracy: 4 channels, every call sourced

By Callsight Research · Editor: Mạnh Hoàng · Method: rule v4 · Published Nov 12, 2026 · Updated Nov 12, 2026

Figures as of Nov 18, 2026. They change as new windows settle.

As of Nov 18, 2026, Callsight tracks 4 finance channels and 1,044 of their own calls. 404 have been judged Correct or Incorrect at one year. The table below gives each channel’s hit rate beside its sample size and its base rate, how often the called move happened anyway, because a hit rate alone cannot be compared. Graham Stephan 57.1% on 84 against a base rate of 47.2%; Andrei Jikh 57.1% on 49 against a base rate of 41.7%; Joseph Carlson 69.6% on 69 against a base rate of 60.5%; Coin Bureau 46.9% on 160 against a base rate of 49.2%.

How accurate are the finance YouTubers Callsight tracks?

One row per channel, at the one-year window. The hit rate is measured on the decided calls that have a base rate, and the base rate on the same calls, so the two can be read against each other. The rows follow the order of the record page, which allows for both the base rate and the size of the sample. That order is a measure of the calls. It is not a judgement of any person, and a different window can give a different order.

Each channel's own calls at one year, with sample size and base rate
ChannelOwn callsCorrect, 1 yearIncorrect, 1 yearPendingHit rate (calls with a base rate)Base rate
Graham Stephan22663412057.1% (84)47.2%
Andrei Jikh1033021757.1% (49)41.7%
Joseph Carlson1295021469.6% (69)60.5%
Coin Bureau58683953646.9% (160)49.2%
Own calls leave out the 448 forecasts the channels read out from banks or other people, which are shown with their source and never counted. Correct and Incorrect count a call repeated on one subject, in one direction, within 90 days once. A channel with fewer than 20 such calls is shown and not ranked. Own calls also include calls the record lists but does not score, and repeats that count once; Partly correct and Within band results are shown on each call and counted in neither column.

Read the hit rate and the base rate together. A channel whose calls are mostly that a market will rise can post a high hit rate in a rising market, and the base rate shows how much of that the market gave anyone.

Does the window change the result?

Yes, often. A call can be Incorrect at 30 days and Correct at a year, so every call is checked at the same 6 windows and all of them are shown. The counts below are Correct and Incorrect at each window; a window still open for a call is left out until it closes.

Correct and Incorrect calls at every window, by channel
Channel30 days90 days180 days1 year3 years5 years
Graham Stephan49 Correct, 48 Incorrect55 Correct, 51 Incorrect61 Correct, 43 Incorrect63 Correct, 41 Incorrect64 Correct, 25 Incorrect40 Correct, 15 Incorrect
Andrei Jikh22 Correct, 27 Incorrect24 Correct, 27 Incorrect27 Correct, 23 Incorrect30 Correct, 21 Incorrect20 Correct, 14 Incorrect0 Correct, 2 Incorrect
Joseph Carlson29 Correct, 42 Incorrect40 Correct, 32 Incorrect45 Correct, 29 Incorrect50 Correct, 21 Incorrect31 Correct, 10 Incorrect2 Correct, 3 Incorrect
Coin Bureau76 Correct, 103 Incorrect77 Correct, 103 Incorrect83 Correct, 97 Incorrect83 Correct, 95 Incorrect52 Correct, 41 Incorrect—
Counts only: each cell is a sample of its own, and a rate on it would need its own base rate.

How is this different from other finfluencer trackers?

A tracker that counts a call as a hit when the price moved the called way after the video, from a price on the day it went out, can reach a different result from Callsight’s method in several ways.

Every call here is quoted word for word from the video and linked to the second it was said, so a reader can check what was claimed. It is measured from the first close after the video went public; a price from earlier the same day can include a move no viewer could have acted on. Each hit rate sits beside how often the called move happened anyway over a window that long, since a channel that mostly says “up” in a rising market looks accurate by default. Every call is checked at the same windows, from 30 days to 5 years, and all of them are shown, so no window can be picked after the fact. And a bank’s target read out on air is shown with its source and never counted for or against the channel.

The rules are the same for every channel and are set out on the methodology page. Code writes the verdicts. A settled verdict changes only through an upheld correction or a new rule version, and both keep the earlier verdict on record.

Why is there no “best finance YouTuber” here?

The record measures statements. Each row counts calls and how they turned out, and says nothing about anyone’s own portfolio or what they will say next. The four channels have samples of different sizes on different subjects, so the windows and base rates are shown in full instead of a single score.

Where can I read each channel’s record?

Each channel’s page lists every call behind its row: Graham Stephan, Andrei Jikh, Joseph Carlson and Coin Bureau. Returns against the S&P 500 are compared in do finance YouTubers beat the market. Crypto calls alone are in how accurate are crypto YouTubers. To check a call yourself, see how to fact-check a financial prediction. A channel or reader who thinks a call was misread can report it on the corrections page; an upheld correction changes the record and keeps the earlier verdict.

Figures are computed by code from the same record as the rest of the site. This article was drafted with the help of AI tools and reviewed by a person before publication. It is not investment advice. Found an error? Report it.