When does the rank predictor become stable?

Question

Hi, I’ve been wondering, the sample size this year has been growing very slowly. I’ve heard in years 2018 and 2019 there have been over 30k entries. However, last year there were only 13k entires(what I’ve heard from a senior). This year it’s been over 2 weeks and only 5.2k entries.

Also, they are showing very good ranks for relatively bad marks.

Would like to know everyone’s thoughts.

Thanks,

Comments

Plenty of useless predictors that can't even get the marks calculation right have popped up, leading to distribution of datapoints over various platforms, I'm guessing. Equally confused about the marks vs rank part.

---

The other rank predictors have been in existence for a long time. Don't get it. Very unnerving.

---

You mean to say marks to rank predictor is inaccurate and scores predicted are inflated on GO predictor?

---

@Ankitkjs

It may be, slightly.

Answer 1

@aspirant22

Think they have not used uniform distribution but some sort of exponential distribution ....

What it means is that whenever people lower than your rank increase that thing is already factored in the distribution so that is expected and not changes things much ....

 

1. https://gateoverflow.in/205581/Rank-predictor 

 

2. https://gateoverflow.in/205154/Rank-predictor 

 

3. https://gateoverflow.in/gate-cse-rank-predictor 

 

4. https://www.youtube.com/watch?v=pSIJSg9MOOg 

 

5. https://gateoverflow.in/mymarks/VisualizeMarks.php 

 

6. https://gateoverflow.in/372419/Post-gate

 

Comments

Yes, that is how they do it every year. This year, very few people are submitting their response sheets. I mean, really, very few. Previously, there would be 30k up submissions. But this year, more than 2 weeks have passed since the predictor was launched and only 5.2K submissions.

As to how they calculate the rank, I think they first calculate the top 0.1%, then estimate a cut off and put it into the score formula. Then they map that score to a rank, as score v rank is usually stable throughout the years.

---

@aspirant22 You can check on github what the rank predictor code is doing. If not at least do not post rubbish things about it.