I paid $6,700 to find out whether paid crypto trading groups actually make you money.

TL;DR

* I subscribed to 9 well-known paid crypto signal Telegram channels for 12 months. $6,700 in subs. 3,505 trades logged.
* Under a uniform mechanical -5%/+10% SL/TP at 10x, 4 of 9 channels had positive per-trade EV. Under a fixed $1k/mo subscriber budget net of fees, only 3 of 9 ended the year net-positive.
* The free public traders we already track on the same platform beat every paid channel in the cohort.

**Disclosure first:** I work at Centaur, a platform that ingests crypto trader messages from Telegram and X and normalizes them into structured trade events. We already track around 100 free public traders. Earlier this year Algod tweeted that "successful traders don't run paid groups" and I wanted to find out if that was actually true.

So we paid for 9 of the most-marketed paid signal channels on crypto Telegram and ran them through the same pipeline as everything else we track.

**The Trader Lineup**

|Channel|Annual Cost|Posting Style|
|:-|:-|:-|
|Bitcoin Habebe|$0 ($100 WEEX deposit, retained)|Copy-trade ongoing|
|The Giver|$1,080 ($90/mo)|Idea-sharing|
|ClearEdge|$2,000 lifetime|Idea-sharing|
|Coin Signals|$588 ($49/mo)|Copy-trade ongoing|
|Crypto Knight|$948 ($79/mo)|Copy-trade ongoing|
|Flying Dutchman|$900 ($75/mo)|Idea-sharing|
|Mizer|$1,200 ($99.98/mo)|Copy-trade challenge|
|Wolf of Trading|$0 ($100 WEEX deposit, retained)|Copy-trade ongoing|
|Delbert Trades|$0 ($89 FUNDEX deposit, retained)|Copy-trade challenge|
|Total|\~$6,716||

Deposit-model fees show as $0 because the deposit is retained as your own trading capital, not collected as a fee.

**Methodology** Every channel message goes through an extraction model that parses each post into trade events (open, close, increase, decrease) with asset, direction, entry, and stated exit when present. Stated prices are matched against actual exchange k-lines at the announced timestamps to compute realized PnL.

Three evaluation frameworks, applied per the trader's posting style:

1. **TP/SL & Real Exits at RR2 (primary).** Use the trader's explicit close if posted; otherwise apply a mechanical -5% SL / +10% TP. Most complete view.
2. **Pure SL/TP at RR2.** Same mechanical rule on every position, ignoring trader closes. Isolates entry quality.
3. **Action-based.** Only positions the trader explicitly closed in-channel. Most flattering; only meaningful if close rate is above 50%.

All returns shown at 10x leverage, which sits at the upper-mid end of what most retail crypto exchanges allow and reflects how an aggressive subscriber would actually run these signals.

**Per-trade EV leaderboard** 1-year mean per-trade return at 10x, primary framework, all 3,505 trades:

|Trader|Mean Return at 10X|n (number of trades in period)|
|:-|:-|:-|
|The Giver|\+26.3%|111|
|Bitcoin Habebe|\+4.6%|1,357|
|Delbert Trades|\+3.7%|37|
|Coin Signals|\+3.1%|498|
|Flying Dutchman|\-3.1%|139|
|Mizer|\-6.5%|164|
|ClearEdge|\-8.2%|337|
|Wolf of Trading|\-11.3%|42|
|Crypto Knight|\-20.3%|129|

4 positive, 5 negative. Cohort splits roughly into thirds: clear winners, marginal/flat, clear losers.

**What a subscriber actually nets** Per-trade EV is abstract. Below is the dollar P&L for a fixed $1k/mo budget split equally across each channel's signals at 10x leverage, net of fees, over 12 months:

|Trader|Gross/mo|Fee/mo|Net/mo|
|:-|:-|:-|:-|
|The Giver|\+$263|$90|\+$173|
|Bitcoin Habebe|\+$46|$0|\+$46|
|Delbert Trades|\+$37|$0|\+$37|
|Coin Signals|\+$31|$49|\-$18|
|Flying Dutchman|\-$31|$75|\-$106|
|Wolf of Trading|\-$113|$0|\-$113|
|Mizer|\-$65|$100|\-$165|
|Crypto Knight|\-$203|$79|\-$282|
|ClearEdge|\-$82|$229|\-$311|

Only 3 of 9 channels stay net-positive after fees on a real subscriber budget.

**The cherry-picking signature** Some channels look profitable because they only confirm closes on winners. Losers quietly disappear.

Close-confirmation rate (positions where the trader posted an explicit close):

* Mizer: 9% (17/197)
* Crypto Knight: 11% (18/159)
* Bitcoin Habebe: 27% (424/1,596)
* Wolf of Trading: 33% (15/45)

If a channel only confirms 1 in 10 trades closed, ask which

1. The "trader's stated wins" view skews wildly more flattering than what a strict copy-trader would actually experience.

**The long-bias trap** The window covers both the late-2025 rally and the 2026 bear leg. The three worst performers in the cohort were also the three most directionally long-biased:

* Crypto Knight: 85.5% long
* ClearEdge: 83.1% long
* Mizer: 82.7% long

One-direction traders are one-regime traders.

**Free public traders vs paid (the surprise)** The best paid channel on the action-based view is The Giver at +34.1%/trade at 10x. Five free public traders we already track on Centaur, filtered to ≥100 closed trades over the year, beat that bar:

|Free Public Trader|n (closed trades in period)|Mean per-trade at 10X |
|:-|:-|:-|
|Dawn|108|\+176.6%|
|Champ|182|\+118.5%|
|Wolf of Trading (Public)|114|\+70.4%|
|Coin Signals (Public) |179|\+55.1%|
|CryptoFreemium|309|\+48.8%|

Wolf of Trading and Coin Signals both run free public channels as well as paid groups. Interestingly, their closed public trades outperform their paid ones.

**Asset class edge** Most channels had edge in exactly one asset class.

* The Giver is the only channel with broad edge across crypto classes: positive on BTC/ETH (+9.5%), large alts (+24.5%), and mid/small alts (+61%).
* Bitcoin Habebe's headline is entirely a mid/small alts sleeve (+5.6% on n=1,247). Take that class out and his edge collapses.
* Bad channels can still have a good sleeve. Crypto Knight is bottom of cohort overall but +35.7% on memes (n=7). Coin Signals is +3.1% blended but +52.8% on memes (n=13). Filtering by asset class typically beats copy-trading the whole book.

**Key Takeaways**

1. The paid market splits into thirds: clear winners, marginal/flat, clear losers. Marketing doesn't tell you which is which.
2. Per-trade EV matters more than trade volume. The fixed budget view kills channels that win on cadence rather than edge.
3. Close-confirmation rate is the single most useful tell for whether a channel's stated stats reflect a real subscriber's experience.
4. Long-bias plus a regime shift kills more channels than bad entries.
5. The best free signals we track outperform the best paid channels in this cohort. Sample size caveats apply, but the gap is large.

Data and full report Happy to answer questions. Suggest improvements for my methodology. Tell me I'm wrong. That's how this gets better. Posted a version with more charts on X with instructions of how to access full report in last post: [https://x.com/centaur\_io/status/2061503357366259717?s=20](https://x.com/centaur_io/status/2061503357366259717?s=20)

**Notes**

* *All returns shown at 10x.*
* *Five channels launched mid-window and don't have a full 12 months (ClearEdge \~9mo, Flying Dutchman \~6mo, Mizer \~6mo, Wolf of Trading \~2.5mo, Delbert \~1.5mo). Stats computed on available history.*
* *Uniform mechanical -5%/+10% SL/TP framework. Some channels publish per-trade SL/TP we couldn't apply at scale, so we benchmarked every position against the same rule. Fairer for cohort comparison but less generous to channels with strong custom SL discipline.*
* *Public-trader sample filtered to ≥100 closed trades. n=108 is the lower bound, n=309 the upper. Bigger samples would tighten the comparison.*
* *AI extraction isn't 100%. We spot-checked against manual reads of each channel.*
* *Returns above 500% absolute are capped to drop microcap pump artifacts.*