TIL: 5 years ago, a 1000 rapid rating was 49th percentile. Today it's the 81st!

...and 8 years ago, a 1000 rapid rating was the 39th percentile!

Inspired by a semi viral forum post on another site, which I thought would be interesting to continue it over here, the conclusion was essentially "...so don't worry if your rating is low - there has been a more significant rating deflation than most people realise".

My gut instinct was actually not that the ratings have significantly deflated over the years, but rather, the pandemic brought in so many casual players that were new to chess, that the bell curve has just simply shifted lower, but ratings beyond that curve probably haven't been impacted too much.

To test this, I sampled 80 rapid games across a stretch in 2019 and across a second stretch in 2024 with random 1000 rated players on chess.com.
If there truly has been significant deflation - we should see a difference in average accuracy between the two samples.

• 2019 - 1000 rapid rating is 49th percentile
  • 40 games reviewed- average rating of opponents was 1012
  • All players must have had a rating of + or - 50 points of 1000 in 15/10 rapid.
  • Games shorter than 15 moves were not counted.
  • Average move count was 35.
  • Ave chess.com accuracy was 68% (70.3% ave for all winners, 64.3% ave across all losers).
  • Ave centipawn loss of all players was 82 
  • Half of the sample games were mine since I was a 1000 in 2019, the other half were random 1000 rated players against each other.

• 2024 - 1000 is now 81st percentile
  • 40 games reviewed - average of opponents was 1011
  • Same criteria (eg. +/- 50 rating points, games under 15 moves not counted, etc)
  • Average moves was 41.
  • Ave accuracy was 70% (74% was the ave of all winners, 66% ave of all losers).
  • Centipawn loss TBC - I ran out of analysis credits on the site that shows centipawn data
  • None of these games were mine since I'm no longer 1000

Conclusion, the 1000s of 2024 are, on average, only 2% more accurate than the 1000 rated players of 5 years ago.

With a larger sample size, it might be that there's almost no meaningful difference at all in playing strength at all between the two ranges of time... but it is interesting to see the significant shift in demographics / percentiles on the site in recent years!

It's also good to see that more new people are coming to the game.
If your rating has remained relatively flat, have you perceived any difference in percentiles or relative strength of your opponents over recent years?

Yeah, people use the words inflation and deflation incorrectly all the time in chess. Makes it impossible to google for information on the topic.

Neat that you checked the accuracy. I agree it was due to a horde of casual players joining.

I didn't realise the mobile app butchered how the main body was presented. View in web browser if you want to see the data. Can do this via logging into chess.com via your phones browser instead of the app.