Pi Matrix New World Record


this last Saturday I set a new world record in the Pi Matrix discipline. The Pi Matrix is usually referred to as the “Everest challenge” of memory. The first 10,000 digits of Pi are divided into 2000 different blocks. The judges picks a random block and the competitor has to recall the 5 digit before and after. This is repeated 50 times. No errors are allowed. When all 50 blocks are recalled correctly the clock stops. The previous record was set by Kevin Horsley in 2013, 16 min 38 seconds. My new record is 14 min 33 seconds.

I started this challenge exactly 3 months before the day of the event. The reason that I started it was to develop a 3 digit system and I thought doing something competitive and productive while learning the system made it more fun and engaging than just repeating different images over and over. I watched Brad Zupps and Nelson Dellis attempts several times and took notes from what they learnt. Brad tried to develop a new system like me but ultimately had to little time to complete the challenge, he had about a month to his disposal I think. So I decided that it needed to be spread out over a longer period of time,. Nelson had different walks where every 1000 numbers became one “walk” with 200 locis. I decided to use that strategy. I talked to Lars Christiansen at the WC in Vienna, 2018 when he decided to try the challenge (he did, 16,50 in march 2019 - missing the world record by 12 seconds!) and he had decided to code two 3 digit images with the middle digit being overlapped and coded twice. I thought that was smart as the challenge is about getting everything right and this was an extra precaution to make sure you recall the images in the correct order. He also wrote on FB about focus for the challenge and how he kind of wasted 2 efforts on the day of his event. I decided to train so that even if I failed on a few attempts - I would still be able to deliver and keep a positive mindset.

Ultimately most of the people who has done this challenge seemed to store a lot of digits at once, kind of resembling the maraton events at normal memory competitions. But Nelson talked about how he struggled to know all of the images by heart (in this challenge 98% is not good ennough) so I decided to take it slow. I did 100-200 digits per day in the beginning. After one month of practice I had 3800 numbers done. After 2 months I had the full 10000. The last month was all about reviewing. I practiced on the event - using a randomizing software and did 1000s of drills. About 1,5 hours a day over the last month. I also went back and forth through the different walks. By the last week, I knew everything and made extremely few mistakes in training (I think one over the entire last week of practice). Kevin Horsely spoke in a youtube clip about pride and doubt being the enemies of focus and concentration. I found out that the beginning of this project had a lot of doubt involved. The last weeks I really struggled with pride (“I already know this, I don’t need to practice anymore”) and I really had to force myself to continue on the schedule I had decided for myself.

On the day of the test the judges are gathered with me in a room at the public library in Gothenburg (Thank you to Idriz Zogaj, Sylvain Arvideu, Jonas Von Essen and Florian Minges). Before we start I decided to warm up with Florian who read some random blocks out loud. The first one he read… I felt like I had never heard before. After a quick reminder I was back on track. On the first real attempt I had another problem. Everything was running smoothly and I felt really strong. The clock was about 15 minutes into tha challenge, 1,5 minutes below the world record, when it was time for the last block - which I failed. At that point I thought that I would have to go home being incredibly close to beating the world record. The next attempt was better. I completed all 50 blocks. But the time was 16.51. A tie with Mark Aaroe Nissen for 3rd best of all time (behind Kevin Horsely and Lars Christiansen). The final attempt started. And this time I just felt confident and decided to push for it but also to be certain before I called out “next”. On the 49th block I hit a stumble and asked 3 times to hear the block again. Then I realized that it was the same block that I had made a mistake during training eralier that week - my only mistake that week! And I could push through. After all 50 blocks were done I set a new world record, 14 minutes and 33 seconds.

Here’s a video of the event (it’s in swedish but it’s subtitles so you can follow along). Also, you basically need to learn 14 swedish words to keep up:
innan = before
efter = after
Korrekt = correct
Nästa = next
and the numbers
noll ett två tre fyra fem sex sju åtta nio

That’s it.


On the first succesful attempt, the 49th block happened to be a collision, that is a case where the 5-digit block that’s called out corresponds to two different places and for which Martin needed to tell us about both places, meaning 20 digits instead of 10. Would it not have been collision in this particular random set, he would probably have broken the WR already then … but then he might not have ended the session with his 14-minute stunner!

It was a pleasure to be there and witness this feat!


I’ve aways wondered about that… do the judges (who know what’ll be called out) adjust the blocks down to just 49 in this case, because there is already 50 answer?

Or is it simply bad luck for the competitor if they end up with one or, even though statistically unlikely, even two or three collisions?

1 Like

It’s just bad luck, so no adjustment is done. In total there are 24 possible collisions (ie 48 number blocks out of 2000), so the chances of not getting a duplicate is actually only around 30% ((1952 / 2000) ^ 50).


Sorry, I worded that badly… just meant that more and more collisions become less and less likely… be really weird if you had all 24 collisions in your 50 blocks.

So just like getting and OLL or PLL skip in speed cubing (or rather the opposite.) Well, fair enough… I was just curious, because unlike cubing the judges know the questions in advance. And to make it even playing grounds, the case could be made that the number of answers rather than the number of questions should be the determinant.

Anyways, not trying to change the format or anything… was just a question that I was always wondering about. Thanks for clarifying.

@mrburt congrats and thanks for the nice write up of how this record came to be. I first heard about the matrix on Nelson’s podcast, when he and Brad were prepping for it.

1 Like

Thanks! It’s tricky with the collisions. For me they take longer to solve than a normal block due to two different factors: I know the will take longer to solve so I get stressed because I need to be faster than normal to compensate, and on top of that I need to double check that I actually said the right things when jumping around between blocks. So all in all I think I spend more than twice the time to solve a collision. That also translates to strss after the collision since I need to make up for lost time. I was very happy that there were no collisions in my last attempt :slight_smile: