How to Read a Winning-Trades Screenshot: We Reverse-Engineered One, Pixel by Pixel

What the picture claimed

The image is a plain bar chart of an index future — black background, roughly 227 bars, a clean V that sells off about a hundred points and recovers all of it. Drawn on top: seven small green triangles pointing right, seven small red triangles pointing left, and six white dashed lines, each running up and to the right from a green triangle to a red one.

There is not a single number on it. No entry prices, no P/L, no win rate, no instrument, no timeframe, no date. Six rising dashed lines is the entire argument — and it works, because six lines that all slope up read as six winners before you've consciously counted anything.

So we measured it instead of trusting it.

How we read a chart that has no numbers on it

A screenshot is data if you treat the pixels as data. We colour-segmented the markers to get each triangle's centroid, fitted the dashed lines with RANSAC to find which marker each one connects, and calibrated the price axis off the spacing of its own label rows — 0.1746 index points per pixel, checked against the chart's own last-price tag, which landed within half a pixel. That's about ±0.09 points of precision on every marker, well inside a single tick.

The triangles turn out not to be signals at all: they're arrowheads on the ends of the lines. Each green one points right, into the line that starts at it; each red one points left, back down the line that ends at it. Reading them that way, the whole overlay resolves into pairs, and the pairs are strictly first-in-first-out in time. That is a trade visualiser — entry, exit, and a connector.

We also reconstructed the underlying bars, so the price series below is the screenshot's own price action, re-measured rather than redrawn.

The trade log we recovered

Seven entries, seven exits, matched in order. Six of them have a dashed line on the original. One does not.

Six wins, one loss, 85.7% win rate, +205.57 index points net. Those numbers are real — we're not disputing the wins. The interesting part is what had to happen to produce them.

The tell: the loser has no line

There are seven entry markers and seven exit markers, but only six connectors. The pair with no line drawn is the only pair where the exit is below the entry — the one trade that lost money. Every winner gets a line; the loser gets nothing.

That single omission is worth more than any of the numbers, because it tells you the picture is not a neutral record of what the system did. It's a record of what the system did when it worked. Whether that was a deliberate filter or just how the plotting was set up, the effect on the viewer is identical.

Where an 85.7% win rate actually comes from

Six of the seven entries land inside a 26-bar window, all on the way down — 30,065.93, then 30,051.40, 30,063.66, 30,073.42, 30,054.23, and 30,039.73. That is not six independent signals. That is one position being averaged into a falling market, then unwound first-in-first-out as price recovers. A seventh entry comes much later, at bar 120.

Scaling into weakness and scaling out into strength manufactures a high win rate almost mechanically, because every lot gets to wait for its own green exit. What it does not do is remove the risk — it concentrates it in the middle, where nobody screenshots.

The lower panel of the chart above is that middle. At bar 65 the account was holding five longs at once, together 137.9 points underwater — nearly twice the largest single win on the whole chart, and 4.6× the only loss it ever actually booked. Trade 3 finished +16.96 after having been 44.51 points in the red, a hole 2.6× deeper than the profit it eventually printed.

None of that appears in the picture. The six dashed lines connect the two moments that look best and skip every bar in between.

What the dashed lines are not

It's tempting to read six neat rising lines as an indicator — a trendline, a rising stop, a target ladder — and to go looking for the script that draws them. We checked, and they're none of those things.

If a line were a support trendline or a trailing stop, price would respect it while the trade is open. It doesn't: on every one of the six, price trades below the line repeatedly, by 26 to 53 points at the worst point. The lines have no predictive content whatsoever. Their slope is just (exit − entry) ÷ (bars held), computed after the fact.

This matters because the most common way a chart like this converts is that someone asks for the indicator. There isn't one. The picture is a drawing of outcomes, not a method.

One more detail worth noticing: two of the exits print on the same bar at different prices (30,092.13 and 30,090.38). Exits at the close of a bar would be identical, so those fills sit at trade-specific levels — consistent with resting limit orders rather than a signal firing.

Six checks for the next results screenshot you see

• Count the markers, not the lines. Seven entries and six connectors is the whole story here. Unequal counts mean something was filtered.
• Look for overlapping entries. Several entries before the first exit isn't a run of winning trades — it's one position being averaged, and its risk is the sum, not the average.
• Ask for the worst open loss, not the win rate. A win rate is a property of the exit rule; the drawdown is the property that decides whether you'd still be in the trade.
• Check whether the drawn lines are respected. If price cuts through them, they're annotations, not a system.
• Notice what's cropped. A chart that starts at a bottom and ends at a top has chosen its window. The same rules run across a V that doesn't come back look nothing like this.
• Ask what happens when the recovery doesn't arrive. Every leg here was rescued by the same bounce. That's one event, not six pieces of evidence.

None of this proves the underlying strategy is bad. Averaging into weakness is a real technique with real practitioners, and it wins most of the time by design. It's the picture that's the problem: it shows the 85.7% and hides the exposure that bought it.

What we can't tell you

We're being explicit about the limits, because the point of this piece is how to read evidence honestly. The screenshot doesn't name its instrument, timeframe, or date. Prices quoted in 0.25 increments around 30,100 fit an index future, but we're not going to pretend to know which one. We can't see position size, so the point figures above are per-contract and the dollar risk could be anything.

We also can't tell whether this is a live account, a simulation, or a backtest, and we're not identifying or accusing whoever posted it — plenty of people post charts like this without meaning to mislead anyone. Everything above is measured from one image, and it describes that image only.

What the measurement does establish is narrow and solid: seven trades, six connectors, one undrawn loser, and 137.9 points of open loss that the picture never shows. That's enough to know the screenshot isn't evidence of an edge — and it's the same test you can run on the next one.