Finance YouTubers, measured
Finance YouTubers' predictions, checked: the leaderboard
Every tracked channel with how often its predictions came true, the sample that rests on, and the base rate: how often the called move happened anyway in comparable past windows.
As of Nov 4, 2026, Callsight tracks 4 finance YouTube channels. Correct at one year, on the calls that have a comparable base rate: Graham Stephan 57.1% of 84 (base rate 47.2%); Andrei Jikh 57.1% of 49 (base rate 41.7%); Joseph Carlson 69.1% of 68 (base rate 61.2%); Coin Bureau 46.9% of 160 (base rate 49.2%).
Filters
Each one is when a prediction was judged, not how far back this table looks. Every window covers every prediction old enough for it.
| # | Channel | Hit rate | Base rate | Still correct at 1 year | Return vs benchmarkprice calls with a benchmark | Scored | Score ↓ |
|---|---|---|---|---|---|---|---|
| 1 | Graham Stephan@grahamstephan · 5.1M subscribers | Hit rate60.6%104 callsOn the 84 of 104 calls that have a base rate: 57.1% correct, base rate 47.2% | 47.2%On the 84 of 104 calls that have a base rate: 57.1% correct, base rate 47.2% | Still correct at 1 year9 of 169 of 16 · base rate 71.5% on 16 of 164 of 11 incorrect at 30 days were correct at 1 year | Return vs benchmark+9.9 pts13 price calls−67.8 … +87.5 pts | Scored104 | Score70 |
| 2 | Andrei Jikh@andreijikh · 3.3M subscribers | Hit rate58.8%51 callsOn the 49 of 51 calls that have a base rate: 57.1% correct, base rate 41.7% | 41.7%On the 49 of 51 calls that have a base rate: 57.1% correct, base rate 41.7% | Still correct at 1 year8 of 138 of 13 · base rate 67.6% on 13 of 134 of 9 incorrect at 30 days were correct at 1 year | Return vs benchmark−6.4 pts15 price calls−31.4 … +18.6 pts | Scored51 | Score65 |
| 3 | Joseph Carlson@josephcarlsonshow · 511K subscribers | Hit rate70.0%70 callsOn the 68 of 70 calls that have a base rate: 69.1% correct, base rate 61.2% | 61.2%On the 68 of 70 calls that have a base rate: 69.1% correct, base rate 61.2% | Still correct at 1 year84.0%21 of 25 · base rate 75.4% on 25 of 259 of 22 incorrect at 30 days were correct at 1 year | Return vs benchmark−5.1 pts47 price calls−12.6 … +2.4 pts | Scored70 | Score53 |
| 4 | Coin Bureau@coinbureau · 2.7M subscribers | Hit rate46.3%177 callsOn the 160 of 177 calls that have a base rate: 46.9% correct, base rate 49.2% | 49.2%On the 160 of 177 calls that have a base rate: 46.9% correct, base rate 49.2% | Still correct at 1 year65.5%38 of 58 · base rate 58.7% on 51 of 5815 of 55 incorrect at 30 days were correct at 1 year | Return vs benchmark−5.3 pts114 price calls−20.4 … +9.8 pts | Scored177 | Score39 |
What the columns mean
- Hit rate
- Claims that resolved in the direction called, divided by claims that resolved at all, at the window chosen above. Claims that could not be checked are excluded and counted separately. The sample size counts distinct calls, not repetitions. Saying the same thing about the same instrument in the same direction within a quarter is one forecast however many times it is said, so a channel cannot raise its own sample by repeating itself - and a call that came good is not worth more for having been mentioned in forty videos.
- Base rate
- How often the called move happened anyway in comparable past windows, using only data known on the day; for a fund, a coin or a company, from its own price history. Not every call has one (a crash call has none, nor an instrument with too little history), so each base rate says how many of the checked calls it covers, and the score sets it against the hit rate on those same calls. A hit rate only means something above it.
- Still correct at 1 year
- Of the direction calls that were correct after 30 days, the share whose 1-year result was also correct. Counted on the speaker's own calls, distinct calls only, windows that have closed only. The opposite movement — incorrect at 30 days, correct at 1 year — is counted beside it, because either one alone would be a choice. Price-target calls are left out: a target reached inside 30 days is still reached inside a year. The base rate is how often the same instruments' 30-day moves were still in the same direction at 1 year. Below 20 such calls a count is printed instead of a percentage. The figure is context and enters no score or ranking.
- Return vs benchmark
- Average return in the direction called, minus the benchmark over the same window. Shares, funds, indexes, commodities, Bitcoin and Ethereum are measured against the S&P 500; other coins against Bitcoin. Repeated calls on the same instrument in the same direction within a quarter are averaged into one before this is computed, so saying a thing forty times counts once. In percentage points (pts), not per cent. Only price calls with a benchmark have one, so each average is printed with how many price calls it covers; below 5 of them the count is printed and no average. The bracket is the 95% interval, and it is the important half. Excess return is dominated by its largest values in a way a hit rate is not: one call on an instrument that then rose sevenfold moves this average more than fifty ordinary ones. An interval that spans zero means the sample cannot tell this apart from having no edge at all.
- Score
- A summary of how this channel's checked calls compared with the base rate; it describes the record, not the person. It combines the hit rate above the base rate, on the calls that have one, with the return against a benchmark and calibration, each pulled towards the base rate by how little evidence there is, so fifty means no better than the base rate. A channel whose base rate covers fewer than 50% of its checked calls has no score and is not ranked.
Blog
Do finance YouTubers beat the market? Four channels, checked
How accurate are finance YouTubers? Four channels' calls checked against real prices, base rates and benchmarks, each rate with its sample size.
How accurate are YouTubers' Bitcoin price predictions?
How often Bitcoin price predictions by Coin Bureau, Andrei Jikh and Graham Stephan came true at one year, each beside its base rate and sample size.
Questions people ask
Which finance YouTuber's predictions have been most accurate?
As of Nov 4, 2026, Callsight ranks the one-year records of the 4 channels it tracks by a score that sets each channel's calls against their base rate, not by the plain hit rate. First in that ranking is Graham Stephan's record: 57.1% of the 84 calls that have a comparable base rate were correct, against a base rate of 47.2%. A channel is ranked only when it has enough decided calls and a comparable base rate covers enough of them.
Why is the ranking not by hit rate?
A hit rate on its own rewards predicting what usually happens anyway. So each call is set against its base rate: how often the called move happened in comparable past windows. The score combines the hit rate above that base rate, on the calls that have one, with the return against a benchmark and calibration, each pulled towards no edge where there is little evidence. Every rate on the table carries its sample size and its base rate.
Which channels does Callsight track?
As of Nov 4, 2026, Callsight tracks 4 finance YouTube channels: Graham Stephan, Andrei Jikh, Joseph Carlson, Coin Bureau. Each prediction in the videos read so far is quoted word for word, dated, linked to the second it was said and checked against real prices and published economic data by the same rules.