cfalas's blog

IOI vs CF performance: 2019 edition
By cfalas, history, 5 years ago, In English

In response to blogs such as this and this, I have made a similar graph for IOI 2019. I was interested in finding how I could expect to do if I qualified, and the only metric I have at the time is CF rating (since there isn't anywhere to submit the problems as of now)

Spoiler
Spoiler 2

So, here it is:

Here are some things to take into consideration:

  • A lot of the participants didn't have their CF account on stats.ioinformatics.org. These participants were dismissed when making the graph
  • The ratings of the participants were taken at present time

As you can see, as in previous IOI demonstrated by the blogs above, even though there is a clear trend, there's too much variation in the data to get something out of. However, here are some interesting facts:

  • Out of the people that got gold, the one with the lowest rating, has a lower rating than the person that has the lowest rating out of the silver medals.
  • Only 30.58% of the participants have a linked CF account(!)
  • Mean ratings for each medal: (only with the ratings of the participants that had their account listed on stats.ioinformatics.org)
    • Gold: 2465.5625
    • Silver: 2248.24
    • Bronze: 2026.125
    • No Medal: 1799.2
  • Does this mean that people with lower ratings can't get a medal?
    • They can, but it's not as easy. The lowest ratings for each medal are:
      • Gold: 1883
      • Silver: 1932
      • Bronze: 1597
      • No Medal: 1223

Sources:

If there are any issues with the statistics please let me know

  • Vote: I like it
  • +118
  • Vote: I do not like it

»
5 years ago, # |
  Vote: I like it +41 Vote: I do not like it

I didn’t had time to put every CF handle from participants thread on CF onto IOI Stats this year, sorry. I’ll try to do it once IOI is over, at which point you may get more data.

→ Reply
»
5 years ago, # |
  Vote: I like it +6 Vote: I do not like it

Rip me

→ Reply
»
5 years ago, # |
  Vote: I like it +32 Vote: I do not like it

Unfortunately CF doesn't give an estimate for rating uncertainty, but if you could somehow incorporate it into the statistics that'd be interesting. I.e. check and mark people who have only done very few contests or whose last participation was long ago. This way it should be possible to differentiate between noise due to an inaccurate CF rating and noise due to actual differences between IOI and CF contests.

→ Reply