chart-forensics

Score Breakdown

Your Submission

{
  "issues": [
    {
      "chart_id": "chart-1749934918-1",
      "issue_type": "wrong_height",
      "description": "Millet point at cy=63. Scale: y=270=0, y=40=10, so 23px/unit. (270-63)/23=9.0, but table value is 7. ~28% overstatement."
    },
    {
      "chart_id": "chart-1749934918-2",
      "issue_type": "missing_data",
      "description": "Seattle bar rect starts at x=405, width=50 — entirely outside the SVG viewport (chart width=400px). The bar and label are not visible, making Seattle effectively absent from the chart even though it exists in the data table (rainfall_mm=37)."
    },
    {
      "chart_id": "chart-1749934918-3",
      "issue_type": "misleading_scale",
      "description": "Y-axis starts at 34 (not 0). Top label=98 at y=40, bottom label=34 at y=270. This compresses the range to 64 units over the full axis, visually exaggerating differences between subjects. History (57) and Art (98) appear far apart, but the gap looks much larger than 41 points would merit on a 0-100 scale."
    },
    {
      "chart_id": "chart-1749934918-4",
      "issue_type": "swapped_label",
      "description": "South and West labels are swapped. SVG position 3 is labeled South (cy=130.644), but value=(270-130.644)/230*916=555=West. SVG position 6 is labeled West (cy=64.858), but value=(270-64.858)/230*916=817=South. The data points are correct but the labels are exchanged."
    },
    {
      "chart_id": "chart-1749934918-5",
      "issue_type": "inverted_order",
      "description": "Data values are plotted in reversed order. Jan bar height=88.89, value=88.89*740/230=286=May table value. Apr bar=250=Jun table value. Jun bar=447=Apr table value. May bar=740=Jan table value. Feb (middle position) correctly shows 331. The entire series is reversed around the midpoint."
    }
  ],
  "methodology": "For each chart I extracted SVG geometry: rect heights for bar charts, circle cy for line charts. I determined the y-axis scale from axis label text positions (subtracting 4px text-baseline offset) to compute: value = y_min_val + (y_bottom - cy) / (y_bottom - y_top) * (y_max_val - y_min_val). I then compared computed values against the JSON data tables. For bar charts I used rect height directly: value = height * max_val / total_px. Chart 2: identified that Seattle rect x=405 exceeds SVG width=400 (off-screen). Chart 3: noted axis min=34 not 0 (misleading_scale). Chart 4: confirmed South/West swap by cross-referencing cy-derived values against table values by name. Chart 5: confirmed inverted_order by observing chart position 1 (Jan) had Mays value, position 2 (Apr) had Juns value, etc. in perfect reversal."
}