How to Judge a Signal Service: Win Rate, Expectancy and Drawdown
Nathan Njuguna
To judge a signal service, look past its win rate and calculate its expectancy: the average gain or loss per trade once winners and losers are weighted by size. Win rate in trading only tells you how often a service is right; expectancy tells you whether following it makes money.
A signal service sends buy and sell alerts on stocks or other assets. Its track record is a list of trades, and three numbers from that list do most of the work: win rate, expectancy and maximum drawdown, the largest fall from a peak.
This guide is for self-directed investors deciding whether to pay for, or act on, someone else's signals. You will learn how to calculate each number from a published trade list, what good values look like, and a six-step test, using 1,520 real closed trades as the worked data.
Key takeaways
Expectancy = (win rate × average win) − (loss rate × average loss); a positive number means the average trade made money, whatever the win rate.
In TruthSayer AI's published stock trades as of November 4, 2026, Earnings AI© had the highest win rate (73.3%) but the lowest expectancy (+0.94% per trade), because its average loss was more than twice its average win.
Dark Pools AI© won only 65.9% of its 279 closed stock trades yet earned +3.99% per trade, about four times Earnings AI©'s expectancy.
A win rate from 85 trades can be off by about ±9 percentage points by chance alone; you need roughly 320 trades to pin a 70% win rate down to ±5 points.
Maximum drawdown shows the pain you would have sat through: from launch to November 2026, the three TruthSayer indexes fell 17.5% to 28.1% from a peak, against 18.9% for the S&P 500.
What is win rate in trading, and why is it not enough?
Win rate in trading is the share of closed trades that made money: winning trades divided by all closed trades. A service with 73 winners out of 100 closed trades has a 73% win rate.
Win rate is not enough because it ignores size. A trade that gains 2% and a trade that gains 40% each count as one win. A service can win most of the time and still lose money if its few losses are large.
Three other numbers fill the gap:
Average win and average loss. The mean percentage gain on winning trades and the mean percentage loss on losing trades.
Payoff ratio. Average win divided by average loss. A payoff ratio of 2 means the typical winner is twice the size of the typical loser.
Profit factor. The total of all gains divided by the total of all losses. Above 1 means gains outweigh losses.
A closed trade has both an entry and an exit. Open trades are not results yet, so win rate and expectancy should be calculated on closed trades only.
What is expectancy in trading, and how do you calculate it?
Expectancy in trading is the average result per trade, after weighting wins and losses by how often they happen. It is the single number that answers "did following these signals make money on average?"
The formula is:
Expectancy = (win rate × average win) − (loss rate × average loss)
Loss rate is 1 minus the win rate, and average loss is written as a positive number. A shortcut gives the same answer: add up the result of every closed trade and divide by the number of trades.
What win rate do you need to break even?
The break-even win rate is the win rate at which expectancy is exactly zero. It equals average loss ÷ (average win + average loss), or 1 ÷ (1 + payoff ratio). A service whose average win equals its average loss needs to win more than 50% of the time. A service whose average loss is twice its average win needs to win more than 66.7%.
Comparing a service's real win rate with its break-even win rate is the fastest test there is. The gap between the two is the service's cushion against bad luck and costs.
How do you calculate expectancy from a real trade list?
You calculate expectancy from a trade list by splitting closed trades into winners and losers, averaging each group, and combining them with the formula above. Here it is for TruthSayer AI's Dark Pools AI© model, using every closed stock trade on its public trade list as of November 4, 2026.
- Count closed trades. The list shows 279 closed stock trades: 184 marked profit and 95 marked loss.
- Find the win rate. 184 ÷ 279 = 65.95%, so the loss rate is 34.05%.
- Average the winners. The 184 winning trades returned an average of +10.87% each.
- Average the losers. The 95 losing trades returned an average of −9.33% each.
- Apply the formula. (0.6595 × 10.87%) − (0.3405 × 9.33%) = 7.17% − 3.18% = +3.99% per trade.
- Check the break-even win rate. 9.33 ÷ (10.87 + 9.33) = 46.19%. The actual win rate of 65.95% clears it by about 20 percentage points.
- Translate it to your account. TruthSayer sizes each position at 5% of the account. At that size, +3.99% per trade adds about 5% × 3.99% = 0.20% of the account per trade on average, before taxes.
The last step matters most. A per-trade expectancy describes one position, not your whole portfolio. Position size decides how much of it reaches your account balance.
Paste a service's trade list into a spreadsheet with one column for each trade's percentage result. The average of that column is the expectancy. If the service won't give you a list you can paste, that tells you something too.
What do win rate and expectancy look like in 1,520 live trades?
In 1,520 live-traded stock signals, the model with the highest win rate had the lowest expectancy. We calculated both numbers for each of TruthSayer AI's three models from the trade lists published on truthsayer.com.
Method. We downloaded each model's full public trade list on November 4, 2026 and kept closed stock trades only, leaving out options trades and open positions. We used each trade's published return, which is net of commissions, exactly as listed, including the largest losses. Our win rates match the headline win rates shown on each product page.
Win rate trading data: average win vs average loss per closed stock trade
Average winAverage loss
The chart shows why win rate misleads. Earnings AI© wins most often, but its average loss of 15.83% is more than twice its average win of 7.05%. Valuations AI© and Dark Pools AI© win a little less often but keep their average losses closer to their average wins.
Expectancy per closed stock trade by model
| Model (closed stock trades) | Win rate | Payoff ratio | Break-even win rate | Expectancy per trade | Profit factor | Longest losing streak |
|---|---|---|---|---|---|---|
| Valuations AI© (85) | 72.9% | 1.14 | 46.6% | +8.82% | 3.09 | 5 |
| Earnings AI© (1,156) | 73.3% | 0.45 | 69.2% | +0.94% | 1.22 | 10 |
| Dark Pools AI© (279) | 65.9% | 1.17 | 46.2% | +3.99% | 2.26 | 7 |
Profit factor here is the sum of winning trades' percentage returns divided by the sum of losing trades' percentage returns. Losing streaks are counted in order of exit date.
Valuations AI© has the highest expectancy, but it also has the smallest sample. The next section shows why 85 trades is a thin base for any conclusion.
How many trades do you need before a win rate means anything?
You need a few hundred closed trades before a win rate is precise to within a few percentage points. With fewer trades, luck alone can move the number a long way.
The standard error of a win rate is √(p × (1 − p) ÷ n), where p is the win rate and n is the number of trades. Doubling it gives a rough 95% range. Applied to the three models:
Valuations AI©, 85 trades: 72.9% ± 9.4 points, so anywhere from about 63% to 82%.
Dark Pools AI©, 279 trades: 65.9% ± 5.6 points.
Earnings AI©, 1,156 trades: 73.3% ± 2.6 points.
Rearranging the formula tells you the sample you need. To measure a 70% win rate to within ±10 points takes about 81 trades; to within ±5 points takes about 323.
The same logic applies to expectancy. Dividing each model's expectancy by its standard error gives a t-statistic, a measure of how far the average sits from zero relative to the noise. Dark Pools AI© scores 3.11 and Valuations AI© 2.98, both above the usual threshold of 2. Earnings AI© scores 1.99, right on the line: positive, but not yet clearly separated from zero.
A service that shows "90% wins" from 20 trades has proved very little. At that sample size, the 95% range runs from about 77% to 100%, and one bad month can reverse the picture.
What is maximum drawdown, and why does it matter for a signal service?
Maximum drawdown is the largest percentage fall from a peak in account value to a later low, before a new peak is reached. It measures the worst stretch an investor following the service would have lived through.
Drawdown matters because most investors quit during one. A strategy with strong expectancy is useless to you if you abandon it halfway through a 28% decline. We calculated each model's maximum drawdown from the daily index values published on its product page, and the S&P 500's from the Federal Reserve's FRED database.
Maximum drawdown since launch: TruthSayer indexes vs S&P 500
Two of the three TruthSayer indexes had deeper drawdowns than the S&P 500 over the same period. Dark Pools AI© fell 28.1% between February 14 and April 8, 2025, the same spring sell-off that took the S&P 500 down 18.9%. Valuations AI©'s 21.4% drawdown ran longer, from August 13, 2025 to February 23, 2026.
A higher expectancy does not mean a smoother ride. Ask every service for its maximum drawdown, the dates, and how long recovery took.
What is a good win rate, expectancy and drawdown?
A good win rate is any win rate comfortably above the service's own break-even win rate; there is no universal good number. The guide below sets out what to look for on each metric.
| Metric | Good sign | Average | Warning sign |
|---|---|---|---|
| Win rate vs break-even | 10+ points above break-even | 3–10 points above | Within 3 points, or no break-even can be calculated |
| Expectancy per trade, after costs | Positive, with a t-statistic above 2 | Positive, t-statistic below 2 | Zero or negative, or not disclosed |
| Payoff ratio | Above 1 with a win rate above 50% | 0.7–1 | Below 0.5 with a high win rate (small wins, big losses) |
| Profit factor | Above 2 | 1.2–2 | Below 1.2 (little room for costs) |
| Sample size | 300+ closed trades | 100–300 | Under 100, or only "highlights" |
| Maximum drawdown | Disclosed, with dates, and one you could sit through | Disclosed without dates | Not disclosed at all |
These bands are rules of thumb, not standards. What counts as tolerable drawdown depends on your own account and nerves, which the "Best for" section below covers.
How does a signal service compare with the alternatives?
A signal service is one of four common ways to pick stocks, and each is easier or harder to check. The table compares them on how well you can verify the record, not on returns.
| Option | Can you see every trade? | Who decides? | Best for | Weak spot |
|---|---|---|---|---|
| Do it yourself | Yes (your own journal) | You | Investors with time and a written process | Hard to stay consistent; most people never measure expectancy |
| Stock-picking newsletter | Varies; often picks only, not exits | You, from their picks | Long-term investors who want ideas | Exit prices and losers are often missing, so expectancy can't be calculated |
| Live-traded signal service with a full trade list | Yes, if every entry and exit is published | You, alert by alert | Self-directed investors who want rules-based ideas they can check | You still have to act on each alert and manage size |
| Fund or ETF | No (holdings, not trades) | The manager | Hands-off investors | Fees, and no say over individual positions |
No single option is best for everyone. An index fund suits someone who never wants to look at a trade. A signal service only makes sense if you will actually read the trade list and follow the rules.
Which kind of signal service is best for you?
The right signal service depends on how long you hold, how much time you have and how much drawdown you can stand. Match the service's numbers to your situation, not to its marketing.
Best for long-term value investors
Look for a high expectancy per trade and longer holds. In TruthSayer's data, Valuations AI© stock trades had a median hold of 41 days and the highest expectancy, +8.82%, but a 21.4% drawdown on the index.
Best for active traders
Look for a large sample and a positive expectancy after costs, because small edges are eaten quickly. Earnings AI© stock trades had a median hold of 7 days over 1,156 trades, with a thin +0.94% expectancy.
Best for beginners
Start with a service that publishes every trade, including losers, and paper-follow it for a month before using money. Compare what you recorded with what the service shows later.
Who a signal service is not right for
Investors who need the money within a year, since drawdowns of 17% to 28% happened within two years in this data.
Anyone who will skip losing signals or double up after wins; that changes the expectancy you actually get.
Day traders hoping to replace a salary. Among Brazilian individuals who day-traded futures for more than 300 days, 97% lost money.
What does following a signal service cost, and how much time does it take?
Following a signal service costs the subscription, trading commissions and spreads, and taxes on gains, and every one of those comes out of expectancy. Costs matter most for services with small average wins.
A real example shows the size of the effect. TruthSayer's Valuations AI© bought 10 shares of Procter & Gamble at $145.52 on April 23, 2026 and sold at $149.18 on June 9, 2026. The gross gain was $36.60 on $1,455.20, or 2.52%. A $9.98 commission cut that to $26.62, or 1.83%, as published.
Subscription. Prices vary widely. TruthSayer AI's signals are free; its full app costs $10.99 a month or $71.88 a year after a 7-day trial.
Trading costs. A flat commission weighs more on small positions. Use our stock profit calculator to see the effect on your own size.
Taxes. In the US, gains on positions held one year or less are short-term capital gains, taxed as ordinary income rather than at long-term rates.
Time. Checking a service properly takes an afternoon with a spreadsheet; following it takes a few minutes per alert, plus the discipline to act on every one.
After you act on a signal, the job becomes record-keeping: log entry, exit and size, then recompute your own expectancy every 50 trades and compare it with the service's.
What are the common mistakes when judging a signal service?
The most common mistake is judging a service by its win rate alone. These five errors account for most bad choices:
Reading win rate without the payoff ratio. A 73% win rate with a 0.45 payoff ratio leaves only a 4-point cushion over break-even.
Trusting a small sample. Under 100 trades, a win rate can be off by 10 points by chance.
Ignoring costs. A $9.98 commission on a $1,455 position is 0.69% of the trade, which can wipe out a thin edge.
Mixing backtests with live results. Simulated trades skip real fills and real nerves. Our guide to backtested vs live-traded signals shows how much results shrink once real money is used.
Skipping drawdown. Expectancy says nothing about the path. Ask for maximum drawdown and its dates.
Expectancy has honest limits too. It is an average of past trades, it can be carried by a few outliers, and it says nothing about whether market conditions will repeat.
Is a high win rate a good sign? The contrarian view
A high win rate is often a neutral or even a warning sign, not a selling point. Strategies that take small profits quickly and let losses run tend to post high win rates and thin or negative expectancy.
TruthSayer's own data shows it. Earnings AI© posts the highest win rate of the three models, 73.3%, and the lowest expectancy, +0.94%. Dark Pools AI© posts the lowest win rate, 65.9%, and earns about four times as much per trade.
US regulators make a related point about promotion. The CFTC, which oversees futures trading, tells investors to ask a direct question of anyone selling a trading system:
"Will the system promoter provide you with independent verification of the claimed trading results?"
— US Commodity Futures Trading Commission, "Commodity Trading Systems Sold on the Internet" advisory
The SEC's investor education office gives similar advice about tips found online:
"Investors should never make investment decisions based solely on information from social media platforms or apps."
— SEC Office of Investor Education and Assistance, Social Media and Stock Tip Scams investor alert
How do you judge a signal service? The TruthSayer Trade-List Test
The TruthSayer Trade-List Test is a six-step check you can run on any signal service in an afternoon. It needs only the service's public trade list and a spreadsheet.
- Get every closed trade. Entry date and price, exit date and price, and the result. No list, no test.
- Calculate win rate and break-even win rate. The gap between them is the cushion.
- Calculate expectancy after costs. Average every closed trade's net result; it must be positive.
- Check the sample. Count trades and apply the ± range; under 100 trades, treat every number as provisional.
- Find the maximum drawdown and its dates. Decide in advance whether you could hold through it.
- Confirm it was live and checked. Real orders in a funded account, losers included, ideally with independent verification of how performance is reported.
A service that fails step 1 or step 6 fails the test, however good its other numbers look.
How does TruthSayer AI publish these numbers?
TruthSayer AI publishes every alert from its three models as a funded-account trade, with entry, exit and result, so anyone can run the Trade-List Test on it. The figures in this article come from those public lists on the Valuations AI©, Earnings AI© and Dark Pools AI© pages.
As of November 4, 2026, the product pages show Valuations AI© at a 78.44% return since inception against 32.27% for the S&P 500 over the same dates, Dark Pools AI© at 33.69% against 29.76%, and Earnings AI© at 31.85% against 35.09%. Earnings AI© trails the index, and the trade list shows it.
TruthSayer's trading rules limit the damage from any single loss: a 5% position size, a trailing stop set at 3× the average true range, and averaging down only when a stock is 2–3 standard deviations below fair value. TruthSayer AI claims compliance with GIPS®, with firm-wide policies independently verified for September 2024 to June 2026; see the GIPS page.
We used closed stock trades only because options returns are measured on a different base and would distort averages. We kept every listed result, including outliers, rather than removing trades that looked unusual, and calculated drawdowns from the published daily index values.
What should you do next?
Pick one signal service you use or are considering, download its trade list, and run the six steps above before you put money behind its next alert. If it has no trade list, that is your answer.
Learn how signals are made and the red flags regulators list in AI stock signals: what they are and how to judge them.
Check whether a track record is real or simulated with backtested vs live-traded signals.
Understand what an independent performance check covers in what GIPS compliance means for investors.
Compare apps side by side in our guide to the best AI stock apps.
Look up any unfamiliar term in the TruthSayer glossary, or open a live trade list on the Dark Pools AI© page and run the test yourself.
The bottom line: Win rate tells you how often a signal service is right; expectancy tells you whether it pays, and drawdown tells you whether you could stick with it. In 1,520 live TruthSayer trades, the highest win rate came with the lowest expectancy. Run the TruthSayer Trade-List Test on any service before you follow it, and walk away from any service that won't show every closed trade.
Frequently asked questions
What is a good win rate in trading?
A good win rate is one well above the strategy's break-even win rate, which equals average loss divided by average win plus average loss. A 46% break-even makes 66% strong; a 69% break-even makes 73% thin. In TruthSayer's November 2026 data, those two cases were Dark Pools AI© and Earnings AI©.
What is a good expectancy in trading?
Any expectancy that stays positive after costs, across at least 100 closed trades, is a good start. Strength depends on the noise: divide expectancy by its standard error, and a result above 2 suggests the edge is not luck. Dark Pools AI© scored 3.11 on 279 trades as of November 4, 2026.
Can you be profitable with a 40% win rate?
Yes, if your average win is large enough. With a 40% win rate you break even when the average win is 1.5 times the average loss, because 1 ÷ (1 + 1.5) = 40%. Any payoff ratio above 1.5 at that win rate gives positive expectancy before costs.
How often should I recheck a signal service after I start following it?
Recheck every 50 closed trades or every quarter, whichever comes first. Recalculate win rate, expectancy and drawdown from the service's list and from your own fills, then compare. A gap of more than about 1 percentage point in expectancy per trade usually means costs or skipped signals are eating your results.
What should I do if a service I follow hits a large drawdown?
Compare the drawdown with the service's historical maximum before reacting. If it is within the past range, such as Dark Pools AI©'s 28.1% in 2025, the strategy may be behaving normally. If it is far deeper, or the trade list has stopped updating, cut position size or stop following it.
Is profit factor better than expectancy?
Neither is better; they use the same inputs. Profit factor divides total gains by total losses, while expectancy subtracts and averages them per trade. Earnings AI©'s profit factor of 1.22 and expectancy of +0.94% both describe the same thin edge as of November 2026.
Sources
- TruthSayer AI, Valuations AI© trade list and index values (accessed November 4, 2026)
- TruthSayer AI, Earnings AI© trade list and index values (accessed November 4, 2026)
- TruthSayer AI, Dark Pools AI© trade list and index values (accessed November 4, 2026)
- Federal Reserve Bank of St. Louis, FRED series SP500 (daily closes through November 2, 2026)
- Chague, De-Losso & Giovannetti, "Day Trading for a Living?", SSRN (2019, revised June 2020)
- CFTC, Commodity Trading Systems Sold on the Internet (accessed November 4, 2026)
- SEC Office of Investor Education and Assistance, Social Media and Stock Tip Scams (updated February 2026)
- TruthSayer AI, GIPS® compliance and verification (verification period September 2024 to June 2026)
This article is for educational purposes only and is not investment advice. Past performance does not guarantee future results.
Nathan Njuguna
Nathan Njuguna is Lead Data and Machine Learning Engineer at TruthSayer AI, specializing in quantitative research, machine learning, and data engineering. He develops and validates systematic trading strategies, builds predictive models, and designs backtesting frameworks so trading signals are statistically validated, reproducible, and ready for production.
