The Jon System


A typical 2-card system requires a total of 2,704 images.

If that’s not hard enough for you, a 3-card system takes 140,608.

The claim I’m about to make will either sound absurd, naive, delusional, or anticlimactic (if it’s already been done and I just haven’t run across it yet). If it happens to be the last one, then please excuse my “wow this is super amazing” tone. It was a breakthrough for me, at least.
I suppose if it’s already been done I can’t exactly call it the Jon System, either…
We’ll sort that out later. For now, here’s the crazy claim:

. . .

I’ve created a style of 3-card system that uses fewer images than a standard 2-card system.

. . .

So, I guess you could say that’s it’s not a pure, traditional 3-card system. It involves two encoding steps for every set of three cards. I’ll explain it, and I guess you can decide for yourself whether it counts.

I realized while brainstorming ways to optimize card memorization that every card really contains two pieces of information: the suit and the value. I wondered then if encoding the two separately for each set of cards would prove more or less efficient. It turns out that it was more. A lot more. I’ll do the math in a minute.
The way that I imagined a person would encode them separately would be to first encode the values into an image based on the order that they appear. 2:5:Q would have a unique image, for example. As would J:10:4 or 2:2:A.
Then they would encode the order of the suits as they appear (S:S:H or C:D:S, etc). This could be using predefined images, or (as I’ll discuss later) actions.
Then in recalling you would recall both the values and the suits, so if you had Kh:6s:3d, it would become K:6:3 and H:S:D. You would recall both parts and be able to say put the first, second, or third components from each set together to recall the specific card.

I know I’m getting a bit long winded with the explanation of the basics, and I’m probably just making it more confusing a this point, so here’s that math that I promised.

A Formula, Just For Fun:

To see how many images it would take for an n-card system like the one I’m suggesting, you would raise the number of card values you have to the number n, and then add the number of suits to the power of n, so…

(13^n) + (4^n)

For a 2-Card System:

A standard 2-card system requires 52*52 = 2,704 images. The Jon System requires a total of 185.

And not all of them have to be images, but like I said, I’ll get to that later.

For a 3-Card System:

A standard 3-card system takes 525252 = 140,608 images. The Jon System requires 2,261.

That’s right. A 3-Card Jon System needs fewer images than a standard 2-card system. I think that’s pretty cool, personally.

That thing that I keep mentioning about not using just images pertains to the idea of combining images and actions. I intend to use images for all of the values, but I think that it would work better for me overall if I use actions for the suits. That way I can say “(image for values) does (action for suits) to (image for next three values) which does (action for next three suits) to…”
I just sort of feel like that would be more effective for me.

. . .

So that’s what I’m working on. I appreciate any support, advice, or questions or whatever that any of you feel the urge to share. Thanks for bothering to read this thing. :slight_smile:

I don’t know if it counts as a true 2- or 3-card system, but I think with some practice I can reach similar times with it. I’ll let you guys know how it goes (though it may be a while before I actually time myself memorizing a deck - I have to finish filling out the images first :P).

Also, if there’s a cooler name than “Jon System” (my name is Jon, so you can see how original that one was ;P), please feel free to trash my name and use that one instead. :wink:

. . .


Hey Preo!

I am on the move here, so I didn’t have time to properly analyze your idea, but I find it VERY interesting. However, as far as I understand, don’t you think you would be using two images to encode two cards? Because you still need to cope with 2,704 combinations of two cards. Who uses a 2-card system needs only one image for two cards; you don’t, you need two. Hence, your drastically smaller number of images. Or am I missing something here?

Also, for a 3-card system, they use one for three cards, but you would still be using two. Not as bad as for your two-card system, but still, much slower then “their” 3-card one. No?

I think what makes all this confusing is the necessary distinguishing between number of available single images and number of possible combinations. You are reducing the number of total single images of your system and you calculate it by SUMMING number of combinations of card-numbers with number of combinations of card-suits, but ( and that’s a big BUT), you still have to encode an amount of combinations equal to these previous two amounts MULTIPLIED BY EACH OTHER and not summed. Hence, you need to use more images per two or three cards; therefore your system is still way slower then theirs.

If I am totally off here, i am sorry. I am on the road, but stopped to think about your system because I found it really cool. But I guess I shouldn’t be writing so fast because I didn’t have time to think properly.

It’s just that I am so curious…



Jon, that’s great that you are working on new systems. This is a great time to be in memory sports, and this is a great forum to explore such ideas. I was thinking a little about this system.

There are basically two different encoding phases here, one to encode your card values (for a 3-card system, this would require 131313 = 2,197 images), and then an encoding phase for defining your suit sequence (which would require 444 = 64 images). Now, whether your images are people, or objects, or actions, or any mixture thereof, doesn’t really matter. You’ll still be using two images for every three card combo, one image giving you your card values, and the other giving you the suit sequence. Two images for every three cards would mean 34 images for a full deck.

Now compare to a block system like Alex Mullen or Johannes Mallow use. They use a system that converts two cards into ONE image, which would require only 26 images for full deck. In their system, they use the same image for A-7 whether the suits are Heart-Club or Club-Heart, which effectively cuts the number of images they have to memorize in half, so 1,352 images. They figure out the suit sequence based on how it is placed on a locus. So, they use fewer images, PLUS have fewer assignments to memorize than what you propose.

You should try your system, though, and let us know how it works. My guess is you would use the Major system (or Dominic) to convert the three card values into an image (maybe A-5-7 would be Talcum powder), and maybe a Heart-Club-Spade sequence could be a Hacksaw (using the initial sounds of the suits), so you might imagine talcum powder being shaken all over a hacksaw. Recall, I think, might be tricky. But best of luck to you.


I think you’ve had good responses so far. I would just add that I’m not clear what you perceive to be the benefit of this compared to, for example, a PAO system. That also uses a complex image and allows you to encode three cards in one image (admittedly made up of 3 bits of info).


Thanks for the input, guys! More than I was expecting to get so soon. I appreciate the feedback. :slight_smile:

Before I get into replying to things, I would like to add the disclaimer that I put this whole thing together pretty late at night and I mainly posted this description so that I could work it out in my head and not forget it all. :stuck_out_tongue: I don’t know if it will work all that well, but it fits the way that my brain works, so that’s why I’m experimenting with it. A personal project, I guess. :wink:

Yeah, there would be two “steps” to encoding each set of cards. I prefer not to use the term “images” because I’m personally using actions for the suits. That would leave you with a single image of the previous image doing the action for the suits to the image for the cards. It’s maybe not that much more efficient, but I find that my brain works very quickly with that setup, so it’s more a matter of fluidity of thought for me.

I’m actually using this primarily as a three card system, partly due to the reasons you pointed out regarding its effectiveness as a two card system.
Again, this mostly comes down to fluidity for me. I would have a lot of trouble working with the huge number of images a three-card system requires, but the action-object setup I’m trying to put together with this system better fits my style, and I have certain “sub-systems” that help me to remember which image is which card-set (such as every set beginning with Q being something from alice in wonderland, etc).

I guess it’s less about image-efficiency for me and more about flow and feasibility.

Not quite. As far as possible combinations, I suppose you’re right, but I have to learn far fewer unique images. And then the use of an AO setup helps turn each “two” images into one image, just utilizing an action and an object.

You’re fine! And I’m flattered that you found my scribblings intriguing enough to go to the trouble of reading them while on the road. :slight_smile:

Wait, you mean there’s a bad time to be in memory sports? :stuck_out_tongue:
Yeah, you’re definitely right about that one - the field is really developing. Thanks for the support. :slight_smile:

And yes, there are two encoding steps. I’ve addressed this in reply to Tammish, but I think it’s worth continuing to discuss. I want to make sure that I’m not desperately justifying it simply because I’m attached to the system. :stuck_out_tongue:
The two steps are those that I’ve put into an AO (action-object) setup, so that they really only become one image apiece. It’s maybe not any more efficient, but as I said before, that part really is mostly for the sake of my thought’s flowing smoothly between the sets.

Also, the systems that you reference are great systems, but they’re also 2-card methods. I think that it comes down to a choice between more images to memorize or more variables to juggle when memorizing. If I weren’t using actions I would agree wholeheartedly that I’m really just using two images per set and it’s no more efficient than before. I think that the action-object thing is what makes it worth trying out from my perspective. Maybe it’ll blow up in my face… No telling until I’ve had some practice, I guess.
Also, I’m using it as a 3-card system (also something I previously mentioned, but I’ll restate it here). Assuming that you treat each action/object pair as one image, that’s when it becomes insanely more efficient. It depends on whether you think that such a pair is really one or two images, I suppose.

I dunno. This is all new territory for me. Half the reason I’m trying this longshot out is just for the experience. :stuck_out_tongue:

Again, thanks for the support. I really appreciate it. :slight_smile:

I’m actually using a “category” system to fill it all out. Each set of card values beginning with 4 is going to be an animal of some sort. If the next value is 3, it’s a bird. Once I get to the last value of the three - for the sake of example we’ll say it’s a Jack - I just have to remember that J, in the bird group, is a “blue jay”. So 4:3:J all together is a blue jay.
Obviously I would work with these until they become automatic (just like with major system associations), but it’s a nice way to help my brain keep it all sorted out (also a benefit of the major system).
It’s more intuitive to me than the major system, though I’m currently using the major system for my 52-image system (which I’ll continue to use until I have this one fully pieced together). We’ll see how it plays out in practice. :wink:

I think the only advantage it really has over PAO in general is that - as you mentioned - it requires one less piece of info. Other than that I don’t know that there’s any clear advantage except that I find it more natural to use personally. I always feel a bit like I’m “forcing” PAO associations. This one just flows a bit better for me.

Thanks again for the feedback, everybody. I’m really feeling pretty good about working with this thing. :slight_smile:


So, please keep us informed on how it goes!



I like it. I see what looks like a definite advantage over PAO here.


Hey Preo!

Man, I was on the road again when I saw LociInTheSky’s comment and I almost got into a car accident! His comment made it all fit in my mind immediately and now I see the advantage of your system. And it is great! Now, we just need to think about the trade-off involved.

I will make a shallow analysis about the table below, but I think that if LociInTheSky and the other experts who use 2-card systems could hop in, we would all profit very very much. I can easily understand the math, but I can’t evaluate the reality involved.

LociInTheSky probably understood it in a split-second, but I did only now and I think that not even you really understood the true advantage of your system. You claim that:

But the problem is: your system is NOT a 3-card system, it is a 1.5 card system!

Please, take a look at the table below:

When we say 1-card or 2-card, we are actually referring to the card/image ratio. So, a PAO system is a 1-card system just like one that uses 52 images to encode 52 cards in 52 loci.

** PLEASE, here we must establish a convention, at least, in the realm of this post. It doesn’t matter if it is a person, an action or an object: it is an image! If someone is better with actions, fine, but for the sake of this objective analysis, actions or objects must be considered as images just like persons. **

That said, all the first three lines of the table show 1-card systems. The fourth line shows your “2-card” version of John system. It uses 2 images to encode 2 cards, so it is also a 1-card system.

But, now, let’s look at the column entitled “Total number of images”. One could argue that based on the card/image ratio and the total number of images, a 52-card system is better – it has the same card/image ratio of a PA or a PAO system, but it needs half or two thirds the number of cards, respectively. I won’t get into this argument because I don’t have any experience to do so.

But, with respect to your “John 2” system, there is not much controversy: in objective terms, it is worse than all other 1-card systems because it has the same card-image ratio, but needs more images than PAO (eve if just a few ones more).

Now, let’s see the remaining lines of the table.

Your “John 3” system encodes 3 cards with 2 images, so it is not a 3-card system like you mentioned, it is a “1.5-card system”. Now, if we use the card/image metric as a proxy for speed when conjuring images (I know there is much more involved, but stick with me in this argument and its relative relevance), we could say that your system is 50% faster than a PAO system! This is fantastic!

But then comes the problem: it needs 2,261 images whereas a PAO system needs only 156. If we use this “total number of images” metric as a proxy for how “toilsome” the system is (please, if you guys know a better term I am all ears), your system is 1,349% more toilsome than a PAO system. Which one is better? Let’s continue before trying to answer.

The next line in the table shows a true 2-card system like the Shadow System. It encodes 2 cards with just 1 image. In terms of “speed”, it is thus 33.33% faster than John-3. In terms of number of cards needed, it needs 2,704 whereas John-3 needs 2,261, so it is 19.6% more toilsome than John-3.

A true 3-card system (is anybody attempting that?) would encode 3 cards with just 1 image, so it would be 50% faster than a 2-card system. But it would require 140,608 images instead of 2,704, so it would be 5,100% more toilsome than a 2-card system.

My conclusion:

I have none for now. I need to think more about that and I would love to hear from the more experienced ones first.

At least, I hope this analysis will help any further discussions.




Well, after thinking a lot about this, I finally came to a conclusion – and I would like to state it here, so that I can move on with my life :slight_smile:

Basically, what Jon is doing is to encode the cards “horizontally”, instead of “vertically” like every other system does. So, for any number of cards that it includes in its “encoding window”, it creates one image for the numbers and one for the suits.

In contrast, vertical systems create one image for all cards in its “encoding window”. The “horizontal” idea is also used in a vertical 2-card system like the Shadow system, but only to provide the consonants and vowels used to create the image when developing the system in the first place. In contrast, the horizontal aspect of Jon System is used to compose a complex image while applying the system.

This is unique and interesting, but also problematic. The problem is the exponential growth you encounter when trying to add cards to the encoding window.

For instance, if including only 2 cards, we need 185 cards (13^2 + 4^2); if including 3 cards, this number goes to 2,261 (13^3 + 4^3); if 4 cards, it goes to 28,817 (13^4 + 4^4).

In contrast, vertical systems of the style of a PA or PAO system grow linearly with respect to the number of images needed. A PA system needs 104 cards (522); a PAO needs 156 (523); a PAOO needs 208 (52*4) and so forth.

However, to further understand this, we need to look at the card/image ratio. When including just 2 cards, the ratio of the Jon system is 1; a simple PA system uses only 104 images for the same ratio (instead of Jon’s 185). In order to get the same ratio of a 2-card system like the Shadow System, a Jon system would need to include 4 cards in its encoding window (a Jon-4 system), thus meaning 28,817 cards instead of just 2,704 (or even 1,352, if you don’t consider the shadows).

An interesting situation happens when comparing Jon-3 with PAO. When including 3 cards in the encoding window, Jon’s card/image ratio is 1.5. A PAO system encodes every card with one image, so its ratio is just 1. That would be a great improvement over PAO, but the hassle to develop such a system is similar to (or even worse than) the Shadow System. You would need to create 2,261 images for Jon-3, whereas the Shadow System needs “only” 441 additional images for a 33.3% better ratio. Moreover, if you don’t consider the shadows, Jon system is in fact way more toilsome than the Shadow System, needing 909 additional images than the latter.

My conclusion is that the Jon System (unfortunately) is not worth it. It is too toilsome in comparison to PAO and too slow (and even more toilsome) in comparison to a 2-card system.

However, the concept is really interesting and it was a great exercise to think about it. Horizontal encoding based on number-suit decomposition may provide other benefits that we not yet foresee.

Keep up the good work Preo!