The percentile conundrum

 

There is no such thing as 100 percentile. The definition of a percentile is:

The value below which a given percentage of observations fall.

For the 100th percentile, it would mean:100% of observations fall below this value. But that is not achievable. The maximum value in a dataset cannot have 100% of values below it, because it itself is in the dataset. So strictly speaking, no value can be at the 100th percentile. So what is the maximum value?

The maximum score is technically at the 99th percentile, or 99 plus as many decimals as the total set increases in size (or more precisely, approaches but never reaches the 100th). It simply means:

Scored higher than 99%(or 99.9%, or 99.99%, or, 99.999%...) of everyone else. That is the ceiling, not because the score is 99, but because the ranking system tops out at 99, or more correctly, anywhere between 99 and 100, 99 inclusive.

So, when I saw this report in Republic TV, I was a little confused. 

Asked help from Claude AI to clarify. I asked:

If the number count of the set taking test approaches infinity, will the first person's percentile approach 100? For example, this Republic TV report says one test taker scored 100 percentile in 4 subjects. Is this possible?

Claude:

Part 1: Does percentile approach 100 as N approaches infinity?

Yes, mathematically:

> Percentile of Rank 1 = ((N − 1) / N) × 100

N (total candidates) : Rank 1 percentile

10      : 90.00

100      : 99.00 

1,000      : 99.90

11,00,000 (11 lakhs) : 99.9999% 

| ∞ | approaches 100, never reaches it

So it is an asymptote — forever approaching 100, mathematically never arriving.

Part 2: Is the Republic TV claim valid?

Almost certainly not mathematically precise — but here is what is likely happening. The rounding explanation

With 11 lakh (1.1 million) candidates, Rank 1's actual percentile is:

> ((11,00,000 − 1) / 11,00,000) × 100 = 99.9999%

When rounded to two decimal places → 100.00

So CUET likely rounds to two decimal places, and 99.9999% rounds up to 100.00 — which they then report as "100 percentile."

The Republic TV report is a rounding artifact, not a mathematical truth. With 11 lakh candidates, the top scorer's percentile is so close to 100 that after rounding it displays as 100 — but the strict answer from our earlier discussion still holds: 100th percentile is unreachable in principle. CUET is essentially trading precision for readability.

Me:

Thank you.  So the second and third positions would also get 100 percentile with rounding off? Republic TV reports, 'The high level of competition was evident in the results, with Saksham Goyal securing second place with 1230.82 marks, narrowly followed by Himachal Pradesh’s Udit Chaturvedi in third with 1207.21 marks.'

Claude:

Calculations

> Percentile = ((N − rank) / N) × 100

| Rank | Calculation | Actual Percentile | Rounded to 2 decimal places |

|1 (Devina)| (1099999 / 11,00,000) × 100 | 99.999909% | 100.00% |

|2(Saksham)| (1099998 / 11,00,000) × 100 | 99.999818% | 100.00% |

|3 (Udit)| (1099997 / 11,00,000) × 100 | 99.999727% | 100.00% |

So yes — all three round to 100.00%

With 11 lakh candidates, each rank step changes the percentile by only:

> 1 / 11,00,000 × 100 = 0.000091%

That difference is so tiny it is completely invisible after rounding to 2 decimal places.Yes, with a population as large as 11 lakhs, rounding to 2 decimal places causes the top several ranks to all display as 100 percentile — even when their actual scores differ meaningfully. Percentile is a blunt instrument at the extremes.

Complete picture

|Rank| Name |Marks|Actual Percentile|Rounded Percentile|

| 1 | Devina | 1232.19 | 99.999909% | 100.00% |

| 2 | Saksham Goyal | 1230.82 | 99.999818% | 100.00% |

| 3 | Udit Chaturvedi | 1207.21 | 99.999727% | 100.00% |

Score gaps

| Comparison | Marks difference |

| Rank 1 vs Rank 2 | 1.37 marks |

| Rank 2 vs Rank 3 | 23.61 marks |

| Rank 1 vs Rank 3 | 24.98 marks |

Two interesting observations

1. The reporter called Rank 3 "narrowly behind" Rank 2 — but the gap between them (23.61 marks) is 17 times larger than the gap between Ranks 1 and 2 (1.37 marks). If anything, Ranks 1 and 2 are the narrow contest, not 2 and 3.

2. All three show 100.00% percentile — yet their raw scores tell a far more differentiated story, with nearly 25 marks separating first from third.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx