Trading Signals: What They Are, Plus a Real Test of One Against Luck

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-06 · Updated 2026-11-02 · Reviewed 2026-11-01

Covered on this page: TradingView and LuxAlgo.

A trading signal is a prompt to buy or sell something now. It can come from a fixed rule run on price data, such as a short moving average crossing a long one, or from a person who studies charts and sends out an alert.

Signal services package those prompts for subscribers, usually for a monthly fee, through an app, a chat group or a copy-trading link.

Most pages on this topic list indicator rules and stop there. This one takes one of the most famous free signals in the market, the 50/200-day golden cross, and checks every occurrence on SPY since 1993 against 20,000 sets of random entry days.

Then it turns to the paid version: what a fee does to an account, and how to check a seller using what US regulators tell investors to look for.

How it works

Every signal has the same parts. There is a market, a direction, and a trigger price or time. Better ones add a stop loss, a target and a time limit, so the trade has an exit as well as an entry.

Rule-based signals come from a formula. A moving average crossover fires when a fast average closes above a slow one; an oscillator fires when it crosses a threshold. The same data always gives the same signal, which means it can be tested on history before anyone pays for it.

Discretionary signals come from a person. An analyst reads the chart, the news or the order flow and sends a call. The reasoning stays in their head, so a subscriber sees the call and its result but never the process that produced it.

Delivery decides what you actually get. An alert by text or chat has to be read and acted on by hand. Copy trading and broker bots place the order automatically, which removes the delay but hands timing and size to someone else’s account logic.

A signal and an alert are not quite the same thing. An alert only says a condition was met, such as price touching a level you chose. A signal adds a verdict on what to do about it, and that verdict is the part a service charges for.

A worked example: the 2025 golden cross on SPY

The setup. On 14 Apr 2025 SPY’s 50-day simple moving average closed below its 200-day average, a death cross, with SPY at $539.12. A week later, on 21 Apr 2025, SPY closed at $513.88, the lowest close between the two crosses.

The signal. The 50-day average closed back above the 200-day on 1 Jul 2025, 53 sessions after the death cross. That is a golden cross, and SPY closed at $617.65 that day. The market had already climbed 20.2% from that post-cross low close before the buy signal printed.

The fill. Anyone acting on the alert after the close could buy no earlier than the next open: $617.24 on 2 Jul 2025, 0.07% below the signal price. Small here, but the gap moves both ways, as the next section shows.

What followed. 126 sessions later, on 30 Dec 2025, SPY closed at $687.01, up 11.23% from the signal close. On 30 Oct 2026 it closed at $755.72, up 22.35%, and the 50-day average ($767.49) was still above the 200-day ($728.85), so no death cross had followed yet.

What the example does not show. One cross that went well says nothing about the next one. The original data below puts it next to the other 14 crosses since 1993 and next to random days.

The original data

The test. SPY daily closes from Yahoo Finance, 29 Jan 1993 to 30 Oct 2026. Prices are raw closes, so dividends are not included. A golden cross is the first close where the 50-day simple average is above the 200-day after being at or below it the day before.

The count. With a 200-day average available from 11 Nov 1993, that rule found 15 golden crosses in 8,298 sessions. The first was 14 Sep 1994 and the latest 1 Jul 2025.

The control. For each holding period, we drew 15 random days from every session since 11 Nov 1993 with a full window ahead, took their average return, and repeated that 20,000 times. If the cross carries information, its average should sit beyond almost all of those random averages.

On the average gain, the result is within chance. At every holding period, between 7.8% and 19.3% of the random sets beat it. The middle 95% of random averages at 126 sessions ran from -0.76% to +10.22%, and the golden cross mean of +8.48% sits inside that range.

One number did stand out, and it gets stated plainly. All 15 crosses were higher 126 sessions later. Only 1.1% of the random sets managed 15 of 15. At the other three holding periods the share of crosses that rose was ordinary.

Eight comparisons were run here, so one unusual result is less surprising than it looks. Fifteen events is also a small sample. Each golden cross comes after a rise, and SPY was higher over most six-month windows anyway: 73.8% of all sessions in the test were up 126 sessions later.

The delay between signal and fill is real money. Across the 15 crosses, the next morning’s open differed from the signal close by a median of 0.26% in one direction or the other. The largest gap was 1.25%, on the morning after the 2 Feb 2023 cross. The open was higher in 9 of the 15.

Whipsaw happened once. The cross of 17 Dec 2015 was undone by a death cross 15 sessions later, on 11 Jan 2016, with SPY 6.22% lower. Every other cross lasted at least 203 sessions before the reverse signal.

What a fee does to the arithmetic

A subscription is charged whether or not the signals do anything. Take a $25,000 account that grows 1.5% a month, and remove $199 at each month end. Without the fee the year ends +19.56%. With it, the year ends +9.18%, a drag of 10.38 percentage points from $2,388 in fees.

The smaller the account, the bigger the bite. The fee is fixed in dollars. On a $5,000 account a 1.5% month earns $75, so the same $199 fee is larger than the gain it is meant to improve. A service has to clear that hurdle before it has added anything at all.

Trading costs sit on top. Each alert you act on still pays the bid-ask spread and any commission twice, and fills can suffer slippage when many subscribers chase the same price at once.

How to judge a paid signal service

Start with the regulators’ red flags. A CFTC customer advisory on social-media scammers, read on 1 Nov 2026, says signals, software and courses sold this way often carry hidden costs or do not perform as promised, and warns about profiles that brag of huge “win” percentages.

The SEC’s January 2012 investor alert on social media lists three warning signs: returns that sound too good to be true, promises of a sure return, and pressure to act right now.

Check the seller’s registration. The CFTC advisory points readers to the National Futures Association’s BASIC database and FINRA BrokerCheck for registration and disciplinary history. The SEC alert adds its Investment Adviser Public Disclosure site. When a seller claims to be registered, these databases let you confirm it and see any disciplinary record.

Treat a track record as a claim until you can audit it. The SEC alert warns that dishonest newsletters lie about the payments they receive and about their track records. A record worth reading was posted before each move, keeps every losing call visible, and uses prices a subscriber could actually have got.

Be wary of anything sold as a machine that cannot miss. A January 2024 CFTC advisory on AI trading bots says the technology cannot predict the future. Its case study, Mirror Trading International, took more than $1.7 billion in bitcoin from at least 23,000 people.

Then run the test this page ran. Log every alert in a trading journal for a few months, including the ones you skip. Compare the results with what random entries in the same market would have done before you trust any streak.

When it fails

Sideways markets. Rule-based signals built on averages need a trend to follow. In a flat stretch the lines cross back and forth, as in late 2015, and each cross costs a round trip.

Late entries. A crossover confirms a move after much of it has happened. The 2025 example fired 20.2% above the post-cross low close, and the matching exit signal arrives just as late.

Wrong size. A signal knows nothing about your account. Sizing still has to come from your own risk per trade rule, and the prop firm challenge page shows how much sizing alone changes outcomes.

Curve-fitted rules. A rule tuned until its past results look perfect is often fitted to noise. Backtesting with a random control, as above, is the defense, and overfitting is the failure it guards against.

Partial following. Subscribers who take some alerts and skip others end up with a record unlike the service’s. A tested algo trading rule avoids that only because nobody overrides it.

Questions people ask

Are trading signals reliable?

No signal is reliable on its own. Every SPY golden cross since 1993 was tested against random entry days above, and on average gain the result was within the range luck produces.

Are paid trading signals worth it?

That depends on whether the results clear the fee and trading costs, which you can only judge from a complete, checkable record. On $25,000, a $199 monthly fee cost 10.38 points of a year’s return.

How can I tell if a signal group is a scam?

Between them, the CFTC and SEC flag these warning signs: bragged win rates, promises of sure returns, pressure to decide quickly, and sellers with no registration you can look up.

Are free trading signals any good?

Free rule-based signals, like a moving average cross, can at least be tested on history. Free alerts in chat groups deserve the same checks as paid ones, because their record is just as hard to verify.

Related

Trading mentors teach the reasoning a signal leaves out, at a similar price. A trading journal is how you score any signal honestly over time. And risk per trade is the input no alert can supply for you.

This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.