3+0 vs 3+2 the difference?

3+0 is a lot more unforgiving. The interesting thing is that this impacts not just how players manage their time (as you mostly looked at here), but also the decisions they make during the game.
For instance, say white is winning due to having an attack going against the black king, but has little time on the clock. Their opponent gives them a chance to liquidate into a winning-technical endgame. In 3+2, they will most likely choose to liquidate because the position will be more straightforward to convert, but in 3+0 the same player would probably keep the attack going in hopes of shortening the game.
I also think that players are more willing to play unorthodox/dubious openings in 3+0 because it might provoke their opponent to spend more time if they want to punish it better.

@oortcloud_o said in #9:
> I keep my analysis in Jupyter and Markdown. I don't have plans on sharing the analysis yet.

Well, first good choce.

You should share or some of us will lose interest. This will become a showroom. Such nice work under secrets. This is not like preparing for a tournament, or using your credibility as chess player for us to take you at your word. We need to be able to check your figures. That is part of data analysis. I think. But you might want to stick to chess old standards of communication. Some things, like the art of playing maybe be hard to give the source of. But population data analysis, come on!

Do you plan to make a repository on github for example for the notebook? Later about your blogs?
If not, then your blogs should have enough details, and they already have more than many other chess blurbs showing data result I have seen, legend, ticks, units and appropriate axis labels, that is great, so that people could interpret your results from the source of data you pointed to.

I understand the need to go to the gist of the story and not distract a first pass reading from your well chose figures. That having to write down the other details might be a chore, or appear optional to communication. But some of us are not just listening to what you want to say, but whether it holds and how it holds. They want to reason with you. The jupiter trace being shared might be enough to go by. Or I am eager to see your analyses notebook on jupiter (I do not care as much about the markdown, or not at all, that is what the blog needs to show use your presentations).

That is my 2.5 cents.

Edit: well the markdown, would go smoothly through github readme files and would be nice with the data processing python notebook (assuming python because of python chess and jupyter).

Excellent work, is very interesting this topic an the analysis of the topic are true

Hey, cool post.
What about analyzing the differences between 5+0 and 5+3 next?

Very high rated players mostly go for 3+0, when it comes to blitz. When they click on the 3+2 button, they often have to wait very long and might also get paired against someone suspect more often than when going for 3+0. (I personally know some strong players who prefer a time control 3+2, yet mostly go for 3+0.)

@Willem_G said in #10:
> eresting to read! It would also be nice to see if there's a difference in rating gain / loss for 3+0 vs 3+2. For example, if the pool of players playing 3+0 is stronger than the 3+2 pool, and there's a player who plays both 50%, you would expect him to lose points on 3+0 and wi

I think the exact opposite is the case. You often see 1700 Blitz in 3+2 who are 2000-2100 Rapid...
I play 3+0 and 10+5, 30+20... The middle ground of 5+0 and 3+2 is impossible for me because I find that those who play "slow blitz" are much stronger than those who play bullet blitz... I guess it comes down to strengths and weaknesses..
But if I play on calculation I will always lose on time, if I go by intuition I can beat people on 3+0.. but to really use my brain I need at least 10 minutes with inc.. or more..

The easier way is to just play 3+1

What a provoking article! This got me thinking about the differences in these two time controls with variants as well, regarding the outcome and performance of a game. This was wonderfully written, thank you for sharing your research and results!

are the graphs for move types (normal/dubious/bunders) incorrectly labeled? it actually shows that with more time spent per move, the proportion of mistakes is higher...

Yeah, makes sense!