Anomaly Radar Explained: Why SAP Can Be 'Normal' Despite a −28 % Chart

An Apparent Contradiction

You open the Annual Cycle for SAP and immediately notice two seemingly contradictory things:

At the top: The Anomaly Radar shows a score of 24 / 100 — Normal. Everything looks fine.

At the bottom: The Seasonal Annual Chart shows that SAP (gold line) in 2026 is roughly 28 percent below the 11-year average. Instead of ~110 points, the price sits at ~72. Visually it looks like a crash.

How can both be true at the same time? The answer is the most important insight about the Anomaly Radar: It does not measure what you see in that chart.

What the Anomaly Radar Actually Measures

The radar answers one very specific question:

"Is the return of the last 10 trading days unusual compared to the historical average for the same calendar period?"

Three keywords are critical:

• Last 10 days — not the whole year, not since January, not the drawdown. Only the last 10 trading days.
• Same calendar period — we compare April 2026 with April 2015, April 2016, April 2017, and so on. Not with January or July.
• Unusual — measured in standard deviations (Z-score), not in percentage points.

The large drawdown visible in the lower chart happened mostly in January–March 2026. That was brutal at the time — but the Anomaly Radar does not ask about back then. It asks: What has happened in the last two weeks? And the answer is: nothing particularly dramatic.

The Calculation in Detail

Let's look at the figures from the screenshot:

SAP is 3.25 percentage points below the historical average for the same calendar period over the last 10 days. That sounds like a lot, but statistically it is not — and this is where the Z-score comes in.

Z-score calculation (simplified):

`

Z = (current return − historical average) / historical standard deviation

`

If the historical spread of 10-day returns in early April for SAP is, for example, ±4 percentage points, then a deviation of 3.25 pp is only about 0.8 standard deviations — not a statistical outlier. The score scales the absolute Z-score by a factor of 30 and caps at 100:

`

Score = min(|Z| × 30, 100) = min(0.8 × 30, 100) = 24

`

24 means: no anomaly. The market is moving within historically normal noise.

Why the Percentile Rank Is Still Interesting

The score says "Normal." The percentile rank says 18th percentile. That is the slider in the screenshot pointing noticeably to the left — in the yellow edge zone, no longer in the green normal range.

What does the 18th percentile mean?

Of all historical 10-day windows measured in early April for SAP, only 18 percent were worse than the current return. 82 percent were better.

This is a complementary perspective to the score:

• The score measures deviation from the mean in standard deviations — "how many σ off?"
• The percentile rank measures position in the distribution — "where do you rank?"

Both can give different signals. In this case:

• Score 24 = few standard deviations off → nothing dramatic
• Percentile 18 = but in the bottom fifth of historical cases → notable, not extreme

This combination is exactly what we want to show: The market is running weaker than normal, but not so weak that it constitutes a statistical outlier. Both statements are true simultaneously.

Percentile slider color coding:

SAP at the 18th percentile falls just inside the gold zone. A value of 5 or 95 would be red — that would indicate something truly striking.

And the Large Drawdown in the Chart?

The annual chart shows a different time horizon and answers a different question:

• Anomaly Radar: Where do we stand over the last 10 days relative to comparable periods?
• Seasonal Annual Chart: How has the entire year developed up to today compared to the multi-year average since January?

SAP had a sharp correction in January and February — clearly visible in the yearly path. In March and April, however, the price has stabilized. It is now making daily moves that fall within the historically normal range. The Anomaly Radar confirms exactly this: "The crash is over; current price action is unremarkable."

That is important information! Anyone looking at the chart and thinking "it will keep falling" is reminded by the radar: The last two weeks were not a crash — they were sideways noise. With a moderate underperformance, but nothing dramatic.

When Should the Score Be High?

The anomaly score jumps to ≥70 (Strongly anomalous) when:

• After a flash crash: The market drops 15 % in 10 days while the same calendar period historically averages +1 %. Difference = −16 pp; with a typical standard deviation of perhaps 3 pp → Z-score ≈ 5 → Score = 100.
• During a parabolic rally: The market rises 12 % in 10 days while historically +2 % is normal. Difference = +10 pp → Score = 80+.
• During unusually stable sideways movement in a historically volatile period — rare, but possible.

The score is not a bullish/bearish signal. It only says: "Something unusual is happening — take a closer look." The direction (up or down) is shown in the other tiles (10-day return).

Score Thresholds at a Glance

How I Use the Radar in Practice

Three practical use cases:

1. Trade filter: Before entering a seasonal trade (e.g., Sell in May, Turn-of-Month), check the score. At ≥70 it is better to wait — the market is currently in an unusual state, and seasonal patterns may be less reliable this week.
2. Post-mortem check: After a sharp market move: Was it really an outlier, or just everyday volatility? The percentile rank and the score give you a sober classification.
3. Context for the chart: When a chart looks visually dramatic, check the radar. As with SAP here — down 28 % looks dramatic, but the last few days are calm. That is a different story.

Conclusion: Two Different Questions, Two Different Answers

The Anomaly Radar and the seasonal annual chart measure different things:

• Annual chart = where do you stand since January relative to the multi-year average?
• Anomaly Radar = is your current 10-day move statistically unusual?

For SAP in the screenshot:

• Annual chart says: The year has been difficult (−28 % vs. average)
• Anomaly Radar says: But right now it is calm (−0.89 % over 10 days, score 24, 18th percentile)

Both are true. Both are useful. The art lies in asking the right question at the right time — and the Anomaly Radar answers a very specific one.

If you want to experiment with it yourself: The radar is active on the Annual Cycle, Monthly Cycle, TDoM Analysis and Overnight vs. Intraday pages. Simply type in the ticker and check the score in the top right. The methodology is explained in the ⓘ badge next to the section title — hover over it to see the formula.