The fine print,
in plain type

LAST UPDATED · 31 JULY 2026

Disclosures Methodology Kill criteria Privacy Terms

Disclosures

What Signal is

Signal Media Co. ("Signal"), the publishing arm of the Finance Wisdom research lab, publishes general and impersonal market commentary of regular circulation, every trading day at 9:00 a.m. ET. Every subscriber receives the identical edition. There are no tiers inside it, no early access, and no better version held back for anyone. The paid tier, when it opens, will sell access to the research workbench and the ledger; it will never sell a different opinion, a faster answer, or a private one. Signal is a publisher. It is not an investment adviser, broker-dealer, or fiduciary, and nothing it publishes creates an adviser-client relationship.

What Signal does not do

No personalised investment advice, portfolio review, or individually tailored recommendations ever. We do not know your holdings, finances, or risk tolerance, and we are structured so that we never will.
No one-on-one communication about securities. Our channels are broadcast-only by design. Staff do not answer individual "should I buy X" questions, in any channel, for any subscriber.
No trade execution, auto-trading, or brokerage integration of any kind.

Conflicts

No issuer compensation. Signal accepts no compensation of any kind, whether cash, securities, advertising purchases, or in-kind from any issuer, underwriter, dealer, or investor-relations firm, for coverage or otherwise. This prohibition binds every principal, employee, and contributor.
Personal trading blackout. Staff may not trade a covered security from the moment it enters the research pipeline until 72 hours after publication. Firm or staff positions, where they exist, are disclosed alongside the published card.
No trading against published work without prompt disclosure of the reversal to all subscribers.

Risk

Trading involves substantial risk of loss. On a long position that loss is capped at what you put in. Signal also publishes short ideas, and a short is different in kind: it requires a margin account, the loss has no ceiling because a price can rise without limit, and your broker can close the position or demand more collateral at a time of its choosing, not yours. Short-term trading in particular produces losses frequently. Short-term trading in particular produces losses frequently. Roughly 45% of our published risk cards are expected to close at a loss even when the system performs as designed. Past results, real or simulated, do not predict future results. Never trade with money you cannot afford to lose.

Availability

Signal is not directed at, and is not available to, residents of the United Kingdom, the EU/EEA, Australia, or Canada. Subscriptions from these jurisdictions are declined or cancelled with a refund.

Methodology

How results are calculated

Entries are assumed at the next session's opening price after publication, a price a subscriber could actually obtain rather than the signal-time price.
Exits occur at the published stop, target, or time stop, whichever is touched first.
Results are reported net of modelled slippage and commissions.
Every published card is counted. None are removed, restated, or "annotated" after the fact. The complete record, losers included, is on the record page.
The published-vs-rejected spread compares the average 60-minute return of everything we published against everything the scanner fired on but our filters rejected, over a rolling 90-session window, recomputed nightly from the journal. The horizon is sixty minutes because that is what the scan log measures: the ledger column is named ret_5d for historical compatibility but is computed from the price sixty minutes after the scan. This page said “5-session” until 2026-08-25, which was wrong, and the emails said it too. The same computation feeds the website and every email, so the two cannot disagree.

Current status: public validation window

Figures on this site are generated from our own research journal rather than entered by hand, and they are records of what the system actually did. They are not a performance claim. Our own replay of the product as shipped is negative, and we have said so above rather than waiting for a validation window to end and hoping. Where a figure cannot be honestly computed, this site shows nothing in its place rather than a zero or a placeholder.

Hypothetical and simulated results

HYPOTHETICAL OR SIMULATED PERFORMANCE RESULTS HAVE CERTAIN LIMITATIONS. UNLIKE AN ACTUAL PERFORMANCE RECORD, SIMULATED RESULTS DO NOT REPRESENT ACTUAL TRADING. ALSO, SINCE THE TRADES HAVE NOT BEEN EXECUTED, THE RESULTS MAY HAVE UNDER-OR-OVER COMPENSATED FOR THE IMPACT, IF ANY, OF CERTAIN MARKET FACTORS, SUCH AS LACK OF LIQUIDITY. SIMULATED TRADING PROGRAMS IN GENERAL ARE ALSO SUBJECT TO THE FACT THAT THEY ARE DESIGNED WITH THE BENEFIT OF HINDSIGHT. NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL OR IS LIKELY TO ACHIEVE PROFITS OR LOSSES SIMILAR TO THOSE SHOWN.

Kill criteria

Where we stand against our own kill criteria

21 August 2026

We published four conditions under which we would tell you this is not working, and we published them before we had a single paying subscriber. Two of them have now tripped. Here is the whole thing, including the part where we got a number wrong against ourselves and had to correct it.

Criterion: profit factor below 1.2. TRIPPED.
Profit factor is 0.571 against our published threshold of 1.2, measured over 307 gated picks, net of modelled slippage and commissions. Profit factor is gross winnings divided by gross losses; below 1.0 means the losers outweigh the winners. It is equal-weighted on returns rather than weighted by position size, because position size is not recorded on our published cards, so a size-weighted figure cannot be computed and we will not imply one. Under recomputation (31 August 2026). Computed from paper fills using stop levels the desk never intended to trade. The trip stands until a corrected figure exists.

Until today nothing in our code computed profit factor at all. We had confidence intervals and expected value, but not the quantity our own criterion names. Announcing a trip on a number that measures something adjacent would have been its own kind of dishonesty, so we built it first.

Criterion: win rate below 45% over 200 consecutive cards. TRIPPED.
28.1% over 307 cards. Under recomputation (31 August 2026). Computed from paper fills using stop levels the desk never intended to trade. The trip stands until a corrected figure exists.

Criterion: published minus rejected spread at or below zero for two consecutive quarters. CANNOT BE MEASURED.
This one is worse than a trip. The published half of that comparison had zero resolved outcomes behind it, so the spread was never actually measured; the number our site was prepared to display was the rejected half with a placeholder standing in for the other side. We found it, we fixed it at source, and the tile now shows nothing at all unless both halves have real observations behind them. But it means a condition we asked you to hold us to has never once been evaluated. We are leaving it in place and we will report it the first quarter it can honestly be computed.

Criterion: underperforming the S&P 500 on return and drawdown for four quarters. NOT YET MEASURED.
We have not run it. We are not claiming it passes.

The number underneath all of this

Replaying our own published picks, with every quality gate applied, over 307 picks spread across 218 distinct publication days and about a year: expected value of negative 0.250% per pick, 95% confidence interval [negative 0.358%, negative 0.143%]. We give the number of days as well as the number of picks because picks published on the same morning share that morning's market move, so counting picks alone would overstate how independent the sample is. The interval does not include zero. The system as shipped loses money.

What we got wrong, and corrected

An earlier internal figure put that loss at roughly negative 0.55%, more than twice as bad. It was wrong, and it was wrong because of a convention rather than a measurement. When a single day's price range touches both the stop and the target, a daily bar cannot say which came first, and our code priced every one of those days as a loss. That is a sensible caution when you are trying not to fool yourself into a good result. When the result is already bad, it manufactures the answer.

There were 387 such sessions. We built a resolver that reads five-minute bars and settles 372 of them properly: 113 hit the target first, 259 hit the stop first. The remaining 15 are still priced at the stop. That is where negative 0.250% comes from, and it is the figure we stand behind. We are publishing the correction of our own overstatement because a shop that only corrects errors in one direction is not correcting errors.

What this does and does not say

It says the product as we ship it loses money. It does not say the underlying signal is worthless. Run the same replay with trading friction set to zero and the result is negative 0.074% with an interval of [negative 0.182%, positive 0.033%], which contains zero. So what is established is that this geometry, a two to one target and stop that resolves against us about seven times in ten, does not survive the cost of trading it. What is not established is that the detectors predict nothing. That distinction matters, and we are not going to blur it in either direction.

One further limitation we found while checking this: our replay assumes every published pick could be entered at the next open with the stop as printed. At least one could not, because the broker rejected the order. We do not yet know how often that happens, and we are going to find out.

What happens now

Nobody has ever paid us anything. There is no payment processor connected to this site and there never has been, so no refunds are owed to anyone. The refund promise attached to these criteria stands for whoever comes later.

What we are not going to do is tune our way out of this. Adjusting the gates, changing the volatility windows, improving the assumed fills: we have measured all three and they are dead ends, and the one adjustment we already made measurably made things worse. Publishing picks as the product is not the business. The workbench is: the rig, the ledger, and the discipline of writing down the kill condition before the test runs. That is what we are building, and this notice is an example of it working.

Every figure above is reproducible. The archived result, including the exact commands, is in our research repository under kill_criteria_profit_factor_2026-08-21.

We will report all four criteria on this page every quarter from now on, whether they trip or not. If we only speak up when something breaks, then our silence becomes a claim, and we have just learned what happens when a number nobody checks sits on a page waiting to be believed.

The criteria themselves, published before we accepted our first subscriber. If any of the following trips, we say so prominently on this site and refund the current quarter to paying subscribers:

The published-vs-rejected spread falls to zero or below across two consecutive rolling quarters.
Win rate falls below 45% measured over 200 consecutive published risk cards.
Profit factor falls below 1.2 net of modelled slippage and commissions over a rolling year.
Published results underperform the S&P 500 on both return and maximum drawdown over four consecutive quarters.

These thresholds may be tightened but will never be loosened, and any change is announced with its date on this page. An edge you can't measure in public is a story, not an edge.

Privacy

We collect the minimum required to deliver a newsletter:

Your email address, when you subscribe, held with our email service provider (Resend, Inc.) and used solely to deliver Signal publications and service messages. We never sell, rent, or share subscriber lists.
Payment details, for paid tiers, are processed by Stripe and never touch our servers.
Basic delivery events (whether an email was accepted or bounced) used to maintain deliverability. Open and click tracking are switched off on our sending domain and always have been, so we do not know whether you opened anything, and we would rather say that than publish an open rate we know is unreliable.
How you found us, recorded as a short tag when you subscribe (for example, the site you clicked from), so we can tell whether anything we do is working. It is never used to address you and never appears in an email.

We deliberately do not collect information about your finances, holdings, net worth, or risk tolerance anywhere, including "for analytics." Unsubscribing is one click and takes effect immediately for marketing and within one publication cycle for editorial sends. To request deletion of your data, email privacy@financewisdom.io.

CALIFORNIA RESIDENTS: we do not sell personal information as defined by the CCPA.

Terms of service

The service. A subscription grants personal, non-transferable access to Signal's published commentary for the paid period. Content may not be redistributed, resold, or republished.
No advice. All content is general and impersonal commentary. You are solely responsible for your investment decisions. See Disclosures above, which are incorporated here.
Billing and cancellation. Subscriptions renew automatically at the disclosed price until cancelled. Cancellation is self-serve, takes two clicks, and stops the next renewal; access continues through the paid period. Annual founding-price locks persist for as long as the subscription remains continuously active.
Refunds. First-time subscribers may request a full refund within 14 days of first payment. Kill-criteria refunds are described above.
No warranty. The service is provided "as is." To the maximum extent permitted by law, Signal disclaims all warranties and limits its total liability to fees paid in the twelve months preceding a claim.
Governing law and disputes. [Pending counsel review: governing law, venue, and any arbitration provision will be stated here before paid subscriptions open.]
Changes. Material changes to these terms are announced by email at least 14 days before taking effect.

The fine print,in plain type