Puzzle Rush Experiment: Results

'Youth is typically proud of its deep, accurate, and rapid calculation of variations. At that time, Tal was already famous for this quality, and I had no wish to be left behind.' - GM Lev Polugaevsky

A couple of months ago, I ran an experiment known as the #PuzzleGrind, where several dozen players tracked their results to test my theories on the benefits of puzzle rush. Here is a summary of my previous claims.

• Puzzle rush can improve pattern recognition and visualization speed.
• Snowball effect: the more you do, the more valuable future training will be.
• You need to maintain your visualization speed boost with consistent training.
• Everyone is capable of intuitive improvement, although youth tends to improve faster.
• Intuitive improvement sets a higher baseline for slow calculation and raises your ceiling.

My 3-minute puzzle rush peaked from 39 to 43 within five months despite only being consistent for around 4 of those months. When I put the most consistent high-volume training in, I peaked by two within two weeks - though admittedly, at least one of those peaks can be attributed to delayed integration of previous training. With two hours of puzzle rush per day, I would expect to continue peaking at least once per 1-1.5 months, and metrics suggest that I can hope for more.

My slow calculation performance on Aagaard and Dvoretsky puzzles has significantly improved, as fast training is not guaranteed to harm slow calculation habits. I am still working to improve my thought process, and puzzle rush has been improving what I can quickly see without hurting my calculation discipline.

It is clear that improving at puzzle rush is based on neuroplasticity, as the kids and teenagers are improving significantly faster than adults with the same volume of training. That said, the results I have had as a young adult feel so extreme that I've revised my views on the limitations of adult improvement. Patient hard work can overcome lazy talent!

Results of those who participated for at least three weeks:

Generally, those who did not peak in speed chess during this experiment were not active in fast controls. Those who got more results than usual for their age category did so because they trained more, not because they had more talent. I rounded blitz and bullet peaks for simplicity. When the old peak is unknown, the highest score in the first week was used as the baseline.

Age: younger than 18

• Tanitoluwa Adewumi (age 9). 3m PR: peak by 7 (47 to 54).
• Roger Shi (age 11). 3m PR: peak by 5 (27 to 32), blitz: +300.
• Zaeem Alam (age 12). 3m PR: peak by 6 (30 to 36).
• Kevin Fong (age 12). 3m PR: peak by 2 (30 to 32).
• Josh Satter (age 16). 5m PR: peak by 5 (19 to 24).
• Gus Logozar (age 17). 5m PR: peak by 3, tactics trainer: +250, blitz: +200, bullet: +200.

Kevin's results may look unusually low, but he didn't train anywhere near as much as the others.

Age: 18-24 years old

• Elijah Logozar (age 19). 3m PR: peak by 2 (41 to 43), +200 in puzzle battle (2550 to 2750).
• Dennis Norman (age 21). +100 in puzzle battle (2400 to 2500), peaks in blitz/bullet.
• Randall Warners (age 24). 3m PR: peak by 6 (28-34), minor bullet peak (inactive).

All participants in this age category averaged at least 1-2 hrs/day for 40+ days.

Age: 25+ years old

a. Standard results: already active in puzzle rush.

• Michael Marmorstein (age 26): 5m peak by 1 (37 to 38)
• Michael Porcelli (age 27). 3m PR: peak by 1 (28 to 29), blitz: +75.
• IM Kostya Kauvitsky (age 28). 3m PR: peak by 1 (37 to 38).
• Skyler Ashby (age 31). 5m PR: peak by 1 (25 to 26).
• Joel Rockey (age 35). 5m PR: peak by 1 (30 to 31).

It is easier to measure pattern recognition improvement in those who are already active.

b. Non-standard results: previously inactive in puzzle rush or low initial score.

• Ezra Etzel (age 25): 3m peak by 4 (28 to 32). Peak by 2 occurred during week 1.
• Hunter Brookshire (age 30). 5m PR: peak by 3 (20 to 23). Hard work and low initial score.
• Denis Bojanic (age 31):  5m PR: peak by 2 (33 to 35). Peak occurred on day 2.
• Alejandro Bonillo (age 33). 5m PR: peak by 2 (34-36). The first peak was on day 2.
• Jeffery Ward (age 34). 5m PR: peak by 5 (14 to 19). Peak by 3 occurred on day 11.
• Jonathan Ingellis (age 42). 5m PR: peak by 2 (35 to 37). First peak occurred on day 8.
• Drew Tuck (age 43). 5m PR: peak by 2 (27 to 29). Hard work and low initial score.
• Firas Sawaf (age 45). 5m PR: peak by 4 (19 to 23). Peak by 2 occurred in week 1.
• Olaf Müller-Michaels (age 53). 5m PR: peak by 4 (18 to 22). All peaks in the first 2 weeks.

Ezra's results seem exceptional, but he was already 1950 USCF and probably underrated in PR. Slow specialists tend to get faster results when doing puzzle rush because they already have a lot of patterns partially automated.

Those who were previously inactive at puzzle rush tended to peak near the beginning of training because they got a visualization speed boost that improves performance. Most visualization speed improvement tends to occur within the first two weeks of training.

Those who are already doing stable puzzle rush training were maintaining this gain rather than further improving visualization speed. There tends to be a ceiling that can only be raised by an increased volume of consistent training. Therefore, most peaks of active adults tend to take at least a few weeks.

Fast peaks in adults are only plausible with a visualization speed boost or inactive stats unless your initial score is relatively low. The only alternative way to get a fast peak is delayed integration of previous study. Improving your pattern recognition takes time and effort!

In my opinion, the fact that high-volume speed tactics can improve visualization speed in adults is a game-changer that allows older players to perform - with their current intuition - as if they were younger.

Additional conclusions and advice:

• Music helps you remember more of what you study provided that you get used to it.
• Serious results require hard work! 2 hrs/day seems to be a tipping point.
• Those who want to make sure that fast training doesn't interfere with slow results should learn how habits work and deliberately respond differently in fast and slow settings.
• Measure your success by your effort and celebrate any improvement! Habits work based on positive reinforcement - punishment is ineffective for positive change.
• A growth mindset helps avoid tilting. This allows you to keep your attention on solving and reviewing missed problems. Your mistakes don't define you!