Last Updated: 2026-10-26
A signal is a proposed action, not a verified result. The useful question is whether a complete, timestamped record can be reproduced under the prices, delays, costs and limits available to a subscriber. A screenshot of selected winners cannot answer it.
The earlier provider ranking relied on personal-test claims and performance figures without the records needed to verify them. Those endorsements are withdrawn; this page does not recommend a paid group. It provides a prospective paper-review method. No money, account access or API key is required. Trading can lose the entire committed amount, and subscribing does not establish an edge.
Request a record that can fail
Scroll sideways to see every column. Keyboard: focus the table region and use the arrow keys.
| Field | What a usable record contains |
|---|---|
| Identity and timing | Stable signal ID, original publication time and time zone, edit/deletion history |
| Instrument | Venue, exact spot or derivative contract, direction and settlement currency |
| Entry | Price condition, validity window and what happens if the entry is missed |
| Exit | Original stop/target rules, later changes and partial-close instructions |
| Result | Actual quantities and fills, open positions, costs and unresolved signals |
| Method | Which messages count, who decides exclusions and how costs are allocated |
Save the rule before observing the outcome. A target touched on a chart is not proof a subscriber filled there. For a limit order, the market can trade at the same price without filling your place in the queue. An entry posted after a price move cannot be evaluated as though it arrived earlier.
Use three distinct statuses: eligible paper observation, rejected under a prewritten rule, unresolved. Keep all three. Recording every signal does not require taking every trade. A stale quote, unavailable product, missing exit rule or already-used exposure allowance should remain a rejection with its reason, not become a convenient deletion or invented fill.
A small worked comparison
Synthetic example: three hypothetical signals each concern one unit of an asset. Assume $1 total commission for each completed round trip; entry-price differences below already represent the execution difference, so they are not charged again as a separate spread.
Scroll sideways to see every column. Keyboard: focus the table region and use the arrow keys.
| Signal | Published entry → exit | Subscriber paper entry → exit | Paper net after $1 commission |
|---|---|---|---|
| A | 100 → 104 | 101 → 104 | +$2 |
| B | 100 → 97 | 100 → 97 | −$4 |
| C | 100 → 102 | No valid entry before expiry | No trade; rejected |
The selected published outcomes sum to +$3 before costs. The two eligible paper trades sum to −$2 after costs. This tiny invented sample demonstrates accounting differences, not the quality of any provider. The missed signal stays in the coverage report. Do not assign it a profitable fill just to match the headline record.
Suppose a separate monthly study produces $45 after execution costs but before an assumed $30 subscription. Net is $15. If $500 was the fixed capital throughout and there were no transfers, that is 3% for that month, not an expected monthly return. A $30 subscription alone is a 6% monthly cost burden on $500, or 0.6% on $5,000. That arithmetic does not justify depositing more or promise that trading will cover the bill.
Accuracy alone cannot establish profitability
With 55% wins, an average winning trade of $1 and an average loss of $2, pre-cost expectancy is 0.55 × $1 − 0.45 × $2 = −$0.35 per trade. Change the payoff or costs and the answer changes. Use realized average wins and losses from the same defined sample, not advertised targets combined with a different historical win rate.
For drawdown, use a chronological account-equity series including open positions and cash flows. The worst losing trade or worst individual leader is not automatically portfolio drawdown. Several groups recommending the same asset can increase concentration despite having different names.
There is no magic number of trades or fixed trial duration that proves future profitability. Preserve an untouched later period, report exclusions and missing records, and compare with a simple preselected alternative under the same costs. Regime changes and selection bias remain possible even after a clean calculation.
Check identity, incentives and permission boundaries
Confirm a provider's identity through independently located official channels, not a handle copied into a direct message. Ask how it earns money: subscriptions, referrals, execution volume or positions in recommended assets. An integration logo does not constitute an independent audit. Edited messages, missing losses and unverifiable partnerships are reasons to withhold a conclusion, not by themselves proof of a specific crime.
The CFTC's AI trading advisory warns about high or guaranteed return claims attached to bots and signal services. It does not endorse a group listed on this site. We checked that source on 26 October 2026.
Automation adds transport and execution steps; it does not remove latency. Trading permission without withdrawal permission can still cause losses through orders. A paper review needs neither permission. Keep secrets and recovery phrases out of messages, and do not pay a purported unlock fee to recover a claimed balance.
End the review with a specific decision: missing records, unsuitable mechanism, costs greater than the observed result, or sufficient evidence for further paper observation. None of those decisions authorizes live trading. See the bot requirements sheet for operational questions if automation is part of the proposal.
Disclosure: AITR may earn commissions from marked affiliate links elsewhere. No ranked provider, funded test, subscription purchase or affiliate recommendation is presented here. Examples are educational and do not establish investment suitability.