ChessTempo Mixed-Mode Performance Analysis
You become what you think about all day long.
-- Ralph Waldo Emerson
My chess friends know that I have developed a health obsession with ChessTempo Mixed-Mode since I started using it 18 months ago. The truth is that I do a lot of different types of tactics problems. The vast majority are via saltmines, aka woodpecker-style repetition sets, using very easy problems. But I've also attempted 935 Mixed-Mode problems where I take as much time as necessary to solve moderately difficult problems (served up based on my rating using ChessTempo's algorithm).
One of the nice features of ChessTempo (at least for premium members) is that one can download one's tactics data in spreadsheet form and do analysis. I did that with my 18 months of Mixed-Mode this week and here is what I found.
This first chart shows a breakdown of the three types of Mixed-Mode problems: Mate in N, Play to Win, and Play and Don't Lose (IE defensive problems). I clearly do better with Mate in N problems and that's not terribly surprising. What is a bit surprising is that my Play to Win and Play and Don't Lose performance is about equal.
This is useful feedback. I need to focus on specialized tactics sets with Play to Win and Play and Don't Lose motifs for a bit, as that is were my tactical weaknesses reside.
The next chart shows the breakdown between solving problems in less than five minutes vs solving in five minutes or more. This chart shocks; it tells me that I have better performance solving Mixed-Mode problems when I take less than five minutes rather than more!
How can this be? I looked at the average rating of the problems in both groups and they were essentially the same (1592 for less than 5 minutes vs 1634 for more). A 40 point rating different couldn't explain this gap in performance.
The only thing that makes sense is that when I quickly recognize the patterns in a problem, I can then quickly calculate that my pattern recognition is correct and solve the problem successfully.
But, when I don't quickly recognize the patterns, I both take much longer to solve the problem--I am groping for any clues as to how to proceed--and I do so much less successfully.
This data-driven evidence convincing points to the importance of doing pattern recognition-style repetitive tactics sets in order to build our bank of quickly recognizable patterns--especially if one is an adult improver who didn't pick up these patterns as a youth.