No single number tells you whether a strategy is good. This guide explains every metric in TradingPal's track-record panel, in plain English: what it measures, what can fool you, and which other number to read beside it.
A track record can feel like alphabet soup at first. It gets much easier when you stop asking which number is “best” and ask which question each number answers:
Read them as a group, not as a leaderboard. A high return with a severe drawdown may be too difficult to hold. A high win rate with tiny wins can still lose money. A beautiful Sharpe ratio from a short sample can be luck. The strongest record is the one whose growth, risk, per-trade edge, and amount of evidence all tell a consistent story.
Win rate is the percentage of trades that made money. Count the trades that ended in profit, divide by the total number of trades, and you have it. Nothing more mysterious than that.
A worked example: say you took 20 trades last month. Eleven of them closed with a profit — even a small one counts — and nine closed with a loss. Your win rate is 11 ÷ 20, or 55%.
It's the most natural number in the world to care about, because it feels like a grade on a test: more wins, better trader. Beginners often see 50% and think “coin flip,” “random,” or “no profit.” That instinct is exactly what this article is here to correct. Trades are not equal-sized guesses, so 50 wins and 50 losses do not automatically cancel out.
Profit factor is gross gains divided by the absolute value of gross losses. The unit depends on the record: setup cards add results in R, while Autopilot and live-account reports add dollars. A profit factor of 1.50 means the recorded gains totaled 1.5 times the recorded losses. Above 1 is historically positive before any costs that record does not model; below 1 is historically negative.
Expected R per trade — also called expectancy or average R — measures what one average trade earned in risk units. One R is the amount the trade planned to lose if its safety exit was hit. If you risked $100, a +2R win made $200 and a −1R loss lost $100. An expected R of +0.30 means the strategy earned about 30% of planned risk per trade on average. If R is new to you, start with the plain-English R-multiple guide.
The two numbers describe the same edge from different angles. Profit factor compares the total pile of wins with the total pile of losses. Expected R tells you what one more trade was worth on average. Positive expected R and profit factor above 1 should agree; if either rests on very few trades, treat it as early evidence rather than a dependable edge.
Average R and profit factor are the headline edge numbers, but the supporting metrics tell you what kind of edge produced them. A record built from many ordinary wins is different from one rescued by two giant outliers.
Here's the honest answer most articles won't give you: a win rate has no verdict on its own. It counts how often you were right — not what being right paid, or what being wrong cost.
If ten wins average $150 and ten losses average $100, a 50% win rate makes $500 before costs. If the average win is only $60 against a $100 average loss, the same win rate loses $400. The payoff and the hit rate have to work together. Fifty percent is not automatically random, bad, profitable, or break-even.
Watch the trap in action. Imagine a trader who wins 9 trades out of 10, making $100 each time — $900 in the bank. On the tenth trade she has no exit plan, holds on hoping, and loses $1,500. Add it up: $900 won, $1,500 lost. She's down $600, with a 90% win rate on her scorecard.
Now flip it. Another trader wins only 4 trades out of 10, but each win makes $300 and every loss is cut quickly at $100. Four wins earn $1,200; six losses cost $600. He pockets $600 while being wrong most of the time.
A 40% win rate just beat a 90% win rate, comfortably. A “good” win rate only means something next to the average size of the wins and losses.
Cumulative return is the account's total gain across the full test after compounding. A +300% cumulative return means the ending account was four times the starting account: the original 100%, plus 300% in gains. Because it keeps accumulating, a longer test can show a much larger cumulative return even when its yearly pace is ordinary.
CAGR (compound annual growth rate) turns that full journey into one smoothed yearly rate. It asks: what constant annual rate would have grown the starting account into the ending account over this exact span? CAGR makes records of different lengths easier to compare, but it is not what the account earned every calendar year. Real returns arrive unevenly.
TradingPal's pattern panel simulates a capital-constrained account: roughly 1% planned risk per trade, at most 16 open positions, no more than 20% of equity in one name, and no more than 95% gross exposure. It reports the representative middle path from 25 allocation runs rather than the luckiest path. Card records also preserve the p25 and p75 return paths; Autopilot preserves the p10, median, and p90 CAGR paths. That spread shows how much scarce-slot ordering changed the result.
The full methodology explains why those constraints and runs matter. Return without the risk and capacity required to earn it is half a result.
A drawdown is the fall from an account peak to the low point that follows it. If an account reached $100,000, fell to $80,000, and later recovered, that stretch was a 20% drawdown. The realized max drawdown is the deepest such fall in the one historical order that actually occurred.
But streaks matter. TradingPal therefore runs a block bootstrap: it samples contiguous blocks of about 15 observed trades with replacement until it builds a synthetic history of the same length. Some blocks repeat and some are omitted, which creates alternate hot-and-cold paths without pretending the trades were independent. Cards build between 200 and 2,000 paths depending on sample size; Autopilot builds 2,000.
If the p95 number is 28%, it means 95 out of 100 sampled histories had a worst drawdown of 28% or less, while about 5 out of 100 were worse. It is a stress-planning threshold, not a promise, not the chance that next year's drawdown will be 28%, and not a conventional 95% confidence interval. Use it to ask whether the position sizing is survivable before real money is involved.
Average hold is the mean time a position stayed open. It tells you whether the edge behaves like a day trade, a swing trade, or a longer position trade. It is an activity description, not a quality score: shorter is not automatically better, and longer is not automatically safer.
Trade count is the sample behind the statistics. Ten trades can produce almost any win rate or Sharpe ratio by luck; hundreds or thousands give the estimates more weight. The symbol count adds breadth: a result seen across many stocks is less dependent on one unusually friendly chart.
The backtest start and end dates show which market history was included. “Ran” or “last updated” tells you when TradingPal most recently recomputed the record. A long span is useful only if the rules used information available at the time, included losing trades, and applied realistic portfolio limits. Read how TradingPal backtests for those safeguards, the cost scenarios, and the limits that still remain.
A strategy can show an edge and still be difficult to deploy. These metrics explain how often it trades, how long capital stays tied up, and whether portfolio limits forced the account to pass on otherwise eligible signals.
A smooth headline can hide one lucky year or one dominant pattern family. Consistency metrics show where the result came from, so you can see whether the historical edge was broad or concentrated.
TradingPal doesn't borrow a win rate from a textbook. The latest completed nightly result contains the results of our own rules across 500+ stocks and tens of thousands of simulated pattern trades. The table below reads that stored result: each pattern's win rate, average R, profit factor where available, and sample size.
Don't crown the row with the highest win rate. Look at it beside average R and profit factor, then use the full panel for return, bootstrap drawdown, Sharpe, test dates, and where the data came from. Trust large, broad samples more than a short lucky streak.
One example of what this kind of counting buys: in an entry study across roughly 54,000 backtested trades, a confirmed-level first-poke entry and a breakout-close entry both won about 53% of the time. The earlier eligible fill produced roughly 28% more total R. The win rate alone would have hidden that difference.
| Pattern | Usual break | Win rate | Avg return | Backtested trades | Fresh (45d) |
|---|---|---|---|---|---|
| Bullish Pennant | Up | 55.2% | +0.58R | 13,034 | 74 |
| Bearish Pennant | Down | 44.1% | +0.15R | 11,004 | 51 |
| Ascending Triangle | Up | 55.2% | +0.58R | 13,034 | 14 |
| Descending Triangle | Down | 44.1% | +0.15R | 11,004 | 4 |
| Symmetrical Triangle | Either way | 55.2% | +0.58R | 13,034 | 95 |
| Falling Wedge | Up | 53.9% | +0.38R | 13,048 | 83 |
| Rising Wedge | Down | 49.3% | +0.13R | 14,154 | 56 |
Historical results of a simulated strategy, refreshed nightly. Triangle rows show their usual break direction's family; each guide breaks out both directions.
These statistics aren't buried in a report — they're the furniture of the app. On the screener, the track-record panel shows the current pattern family's win rate, cumulative return and CAGR, expected R, bootstrap 95% max drawdown, Sharpe ratio and its confidence check, profit factor, average hold, sample size, and test dates. That's the exact panel below, populated from the latest completed stored result.
You'll meet the same numbers in two more places: a setup tile can carry that stock's own win-rate badge for the pattern it is showing, and the copilot can quote the relevant historical follow-through rate in its plain-English read. A 62% badge means that stock's recorded history with that pattern—not a universal promise.
Not a screenshot — this is the real panel from the screener, populated from the latest completed stored nightly result.
Before experienced traders ask “will this trade win?”, they ask a better question: “if it wins, how much could it make — and if it loses, how much is planned at risk?” That comparison is the risk/reward ratio.
Anyone can say a chart pattern “works.” Backtesting is how you check: write the rule precisely, replay it over years of prices, and keep every simulated trade—winners and losers alike.
If you've ever looked at a smooth equity curve and thought, “Okay, but what did you leave out?”—good.
A bull pennant is a short breather after a stock runs higher.
Educational content, not investment advice. Backtest statistics are historical results of a simulated strategy. Publication timing varies by record; the numbers describe the past, not the next trade.