Why a strategy screenshot proves almost nothing
One winning chart is a cherry-pick. Learn why a single screenshot is not evidence, and what a tested sample of trades actually tells you about a strategy.
Part of the Backtesting and Research track on Agenticks. About 9 minutes, written for a advanced reader.
You have seen the post. A clean chart, one perfect entry, a green arrow, and a caption that says this rule prints. It looks like proof. It is not. A screenshot shows one outcome that someone chose to show you. It does not show the trades that did not work, the times the same rule lost, or how many charts the person scrolled through before they found this one. By itself, a single winning chart is closer to a lottery ticket than to evidence. A lottery winner holding a giant cheque is real, but it does not mean buying tickets is a good plan. You are only being shown the winner. The honest way to judge a rule is boring by comparison. You write the rule down so it is precise, you run it over real history, and you read the whole pile of trades it produced. That pile, not the prettiest single frame, is where the truth is.
A single chart is a cherry-pick, not a sample
Almost any rule can produce one beautiful chart. With enough markets, dates, and timeframes to scroll through, the perfect example always exists somewhere. Finding it tells you about the searching, not about the rule.
Think about what a screenshot quietly leaves out: How often the rule actually wins, not just this once. The losers, including the runs of losses in a row that would have hurt. The worst drawdown, the deepest peak-to-trough drop your account would have sat through. The trade count, because two trades and two hundred trades are not the same kind of evidence. None of that fits in one frame. That is the whole problem. The frame is chosen precisely because it looks good, so the things that look bad are exactly the things it cannot contain. A screenshot is a highlight, and highlights are selected. Nobody screenshots the entry that stopped out two minutes later, and nobody posts the week the same rule gave back everything it made. There is a name for showing only the flattering examples: cherry-picking. It is not always dishonest on purpose. Our memories do it for us. We remember the trade that worked and quietly forget the four that did not, so the rule feels better than the record would show.
- A single winning screenshot
- That this rule worked one chosen time
- A backtest over real history
- How the rule behaved across many trades
- Sample size
- How much to trust the numbers
- Maximum drawdown
- The worst losing stretch you would have endured
Evidence lives in the sample, not the snapshot
One trade can be luck. Ten trades can still be luck. Only a large enough sample, read for win rate, expectancy, and drawdown together, starts to separate a real pattern from a coincidence. A backtest gives you that sample; a screenshot gives you a single point.
Here is why one chart can lie even when it is real. Imagine a rule that wins seventy percent of the time. That sounds great, and on any given winning trade the screenshot looks fantastic. But if the three losers out of ten are each five times bigger than the winners, the rule loses money overall. That is the gap between a win and an edge. Expectancy is the average you would expect to win or lose per trade across the whole sample, blending how often you win with how much you win or lose. A screenshot of a winner tells you nothing about expectancy. A test of two hundred trades does. This is also why curve fitting is so tempting: it is easy to keep adjusting a rule until the past looks perfect, then screenshot the result, without ever checking whether the edge survives outside that one cherry-picked stretch.
A rule wins 7 out of 10 trades, but each of the 3 losers is much larger than each winner, and overall it loses money. What does a screenshot of one winning trade prove? Almost nothing about whether the rule makes money Correct. One winner says nothing about the losers or the average result. A high win rate with oversized losses can still have negative expectancy.
cherry-pick sample drawdown expectancy
Put these in order, from weakest evidence to strongest, for believing a rule.
- One winning screenshot
- Ten cherry-picked trades that all look good
- A backtest over a few hundred trades, winners and losers included
- That same result checked on data the rule was never built on
A results screen reads a whole sample at once: trade count, win rate, expectancy, and drawdown, not a single frame.
None of this means screenshots are dishonest by nature. A chart is a fine way to illustrate a single idea. It just is not evidence that the idea works. The next time a perfect entry shows up in your feed, ask the quiet questions the frame cannot answer: out of how many trades, with what losers, and how deep did it ever fall. If those answers are missing, you are looking at a story, not a result.
You can now spot a cherry-pick
You know why one chart proves almost nothing, and why a tested sample read for expectancy, drawdown, and trade count is the real evidence.
Common questions
- Does a winning trade screenshot prove a strategy works?
- No. A single screenshot shows one chosen outcome out of many possible ones. It says nothing about how often the rule wins, how big the losses are, or how many trades stood behind it.
- What actually counts as evidence for a trading strategy?
- A tested sample of trades over real history, read for win rate, expectancy, drawdown, and trade count. The number of trades and the size of the worst losing stretch matter as much as any single result.
- Why is one good chart misleading?
- Because you can almost always find one chart where any rule looked perfect. Showing the best case while hiding the losers is cherry-picking, and it tells you about the chooser, not the strategy.
Terms defined in this lesson
Continue
Sources