Graph of Games Popularity on lichess

I couldn't find this data so I made a chart myself. Based on all games played on lichess in May 2023. The popularity of different time controls based as a function of player's rating.

The games popularity is weighted by time control (so 1 game of 10+0 is worth as much as 10 games of 1+0)

Here is the image.

i.gyazo.com/9b617222179183a7a8a316b6433b0be3.png

Clear winners are 10+0 below 1900, 3+0 1900-1500. and 1+0 for 2500+. rankings above 2800 have too few games and it's just noise over there so I didn't include it.

Known issue: the ratings between e.g. bullet and rapid aren't comparable so one has to adjust for that.

Never would have guessed 15/10 to be so unpopular.

@kbbllee very cool project! I always LOVE when people do such statistics!

I would suggest that you also take other parameters into account aswell for the final graph, because two or three other things that come to my mind are important aswell:

1. The amount of players. There are very few 2400+ players, while the majority of players statistically is below 2000. So it would make sense to somehow also consider the amount of players. When you display a player of 2400 the same way like a player with low rating, making them equal, you are missing the fact that there are thousands of low rated players for every master. In other words, much more weak players play bullet than strong players. Your graph leaves the impression, bullet would be a master thing, but it is not. Only for the individual, but not overall.

2. People don't use the full time. So comparing a rapid game of 10 minutes to 10 bullet games of 1 minute is false and inaccurate. I have played many rapid games that ended after 3 minutes. Even more extreme is the difference between bullet and classical. While in bullet you use up the full time, in classical you often use only a fraction of your time. Especially low rated chess players play classical like a bullet game - so they might have 1 hour time, but lose or win the game after only 5 minutes. This is why you need the real game times, not the time format times. The way to solve this would be to at least assume certain average numbers, but for sure 10 minutes for rapid is wrong.

3. Keep in mind what time format gives people the most joy. Some people may play a time format out of tilt, addiction or desperation and not because they enjoy it. So comparing 10 bullet games to 1 rapid game doesn't really do it for me for this reason aswell. However point 3 is just a sidenote. Of course you can't measure happiness.

With a renewed version of this, I would be glad to read a blog article from you about it.

@CheerUpChess-Youtube said in #3:
> @kbbllee very cool project! I always LOVE when people do such statistics!

Thank you :)

>
> I would suggest that you also take other parameters into account
>
> 1. The amount of players.

I'm fully aware of that. But I wanted to know which games are popular among which level of players in particular to know at which time control there are the most players of my level and where is the easiest to find a game.

Of course nobody should read from the graph that most bullet games are played by top players.

>
> 2. People don't use the full time

You are right, but there is no data of time used by both players in the games dump. I still think some weighting is important to track which games are most popular and taking full time control is better than nothing.

>
> 3. Keep in mind what time format gives people the most joy.
Of course but you can neither scrap it from a PGN nor present on the graph.

>
> With a renewed version of this, I would be glad to read a blog article from you about it.

I don't have any deeper insight on the data :) my main motivation was to answer a question "if I keep my progress - for how long will I be able to find opponents on 5+3" and I have satisfied my curiosity here

@kbbllee said in #1:
> The games popularity is weighted by time control (so 1 game of 10+0 is worth as much as 10 games of 1+0)

Good idea to measure actual time played instead of number of games. How did you weigh the increment games considering the length is dependent on number of moves?

@i-bex said in #5:
> Good idea to measure actual time played instead of number of games. How did you weigh the increment games considering the length is dependent on number of moves?
I used the 60-move rule.

@kbbllee said in #6:
> I used the 60-move rule.
There's 50 move rule (results in a draw) and google says 40 is average number of moves per game. What is 60 move rule?

@i-bex said in #7:
>
In fide when you decide to which category a time control belongs it uses a basic time plus 60 x increment formula

@kbbllee said in #8:
> In fide when you decide to which category a time control belongs it uses a basic time plus 60 x increment formula

Thanks, never knew how it's decided.

I tend to think your criteria for popularity is correct. It's not a matter of how long the people used or what their rating is, the question is whether they decided to play the time control or not. If I decide to play a 10-0, it doesn't matter if I only use one minute, my intent was to play 10-0. Therefore it should get one vote for popularity (or two since 2 people chose to play).