Geography Snap : The Quest for Sub-2s

Hello, dear JetPunker, and welcome to the most complicated Geography Snap blog I've made so far ! As you've probably read the title, you know that today, we're gonna talk of the sub-2s quest on Geography Snap. Or more precisely, on the Flags Easy gamemode. So grab a coffee, sit back, relax, and enjoy !

Yes I prefer to make you wait a bit more, because I'm gonna use specific unofficial Geography Snap vocabulary along this blog, so it's better for you to know what I'm gonna talk about :

- Category : Categories are what you select before playing a game. These are "Countries", "Capital Cities", "Flags", "Flag-Shapes", "Shapes" and "Biggest Cities".

- Difficulty : Difficulty is what you select before playing a game too. It can be set to "Easy", "Medium" or "Hard", each changing the number of tiles you'll have to match (20, 40 or 60 respectively), and the countries that can appear).

- Game mode : A game mode is a combination of one or several categories, along with a difficulty level.

- World Record : A world record is a first place in the leaderboard of whatever game mode.

Right, now time to really dive into the subject, and yes that really is the third time I use this verb in this blog, and talk about the history of the Flags Easy game mode World Record !

Now, I'm not Stewart, so I don't know the full history of the Flags Easy World Record, but I will do my best to tell you everything I know about it. So, the oldest piece of information I have is that Stewart had done a 3.7s on October 28^th 2020 (which is quite an old piece of information indeed), and Changedaworld beat that World Record in with another 3.7s, on March, 27^th 2021, with his old account, confusingly named "changetheworld". He also tied it again the next day, because why not. Then, nimbru tied him on August, 11^th 2021, but I can't tell if it was the World Record at the time. It's useless to know that anyway, since 6 minutes and 8 seconds later, he managed to get a 3.5s. And that was definitely the World Record, because I've got a screenshot of the Flags Easy leaderboard on September, 18th 2021, and nobody at the time managed to do a 3.4s, or even a 3.5s or a 3.6s. Wait what, you wanna see the screenshot ? Alright then...

And no, the 3.4s is not an edit, he managed to do it the next day, on August, 12^th 2021. But then I said "Allow me to introduce myself", and I got a 3.3s on November, 3th 2021. Also, quick side note : at the time, I thought nimbru's World Record was there since a long time (I imagined that to get 3.4s instead of the four other 3.7s, he had to beat it early, and then improving it, 0.1s by 0.1s), although he actually broke it only 83 days before. Then, I got a 3.2s on December 10th 2021. I really don't know why I decided to beat it back then, because this securing strategy (beating his/her own records to make them last longer) is definitely not one I use (however it's the one that Changedaworld uses). But on February, 2^nd, Changedaworld, with his Changedaworld account, tied this record. Another quick side note : although the time shown is rounded down to the tenth of a second, the actual time is calculated down to the hundredth of a second, so when I'm referring to a tie, I'm actually talking of a tenth-of-a-second tie, and if I say that "A tied B", that actually means that A is placed first and B second. Anyway, Changedaworld's 3.2s lasted until May, 17^th 2022, when I tied him. And later that day, he tied me again. But Changedaworld really wanted this World Record, so he beat himself, with a 3.1s, then a 3.0s, and finally a 2.9s on May, 19^th 2022.

Wait, what ? A sub-3s ? Well yeah, Changedaworld was the first player to do a sub-3s on any game mode. But I didn't give up. On May, 31^st 2022, the same day I beat the World Record in Flags Medium and Hard, one day after I beat the World Record in Countries Easy and Capitals Easy, and one day before June, I broke the World Record with a 2.8s, but wanted to secure it, so I managed to do a 2.6s on June, 3^rd 2022. This record lasted for quite a while, actually. Now, on June, 17^th 2022, I asked Changedaworld if he thought that a sub-2.5s was possible. To which he answered : "maybe, depends on how fast you can click"... Anyway, on September, 13^th 2022, he tied the 2.6s, but this was short living, considering I did a 2.5s the next day. But Changedaworld wouldn't give up, and broke that with a 2.4s on September, 17^th 2022, which I tied four days later, on September, 21^st 2022. Changedaworld tried to beat this record many times. On October, 16^th 2022, he did get a 2.4s two times, but was placed 2^nd two times too. Then, on October, 21^st 2022, he did get a 2.4s again, but was still placed 2^nd. And finally, on October, 27^th 2022, he was placed first by doing a 2.4s for the fourth time. Then, I tied him on November 11^st 2022, but later that day, Changedaworld did the first 2.3s. But we still didn't talk about a sub-2s in this game mode back then. 6 days later, while I was in the school bus, I tied his 2.3s, without having to do it 36 times to get first. On December, 4^th 2022, which is more recent than any of the previous dates I mentioned, while I was trying to match 1000 tiles in the fastest time possible, I got a 2.1s. Now, a bit of context : I was out of the bathroom, and while in the shower, I told this to myself :

"Well, back in May, I said that 2.7s would be done during the summer holidays, 2.6s during October or November, and 2.5s in 2023. And then, it wouldn't be improved anymore. Seems like I was wrong, so now I'll say we'll do everything to get 2.2s before the end of the year, maybe even before Christmas, then 2.1s will be done in March or April, then 2.0s will be done in a very long time. The sad thing is that, now, a 0.2s improvement for Flags Easy is impossible, it can only be improved 0.1s by 0.1s now."

Because, indeed, I think of Geography Snap even when I'm in the shower. So yeah, getting that 2.1s was really surprising, and, if you want to know, no, I didn't beat my Personal Best for matching 1000 tiles in the fastest time possible. But why did I wait until May 2023 to submit this blog, you might ask ? Well...

Here it is : The first 2.0s ever ! It has been done on May, 24th 2023. We have never been that close to a sub-2s ! Before we discuss about whether or not this time could be possible, let's first see a table that summarizes a lot of what I said :

And here is a chart showing the known evolution of the Flags Easy World Record :

As you can see, the evolution is very regular. You could think that the improvements done would be more and more sporadic, but that's not the case.

But anyway, let's now talk of whether or not a sub-2s could be done in the future.

To answer the big question, we first need to know how many hundredths of a second we are away from a sub-2s. Indeed, as I said earlier, times are calculated down to the hundredth of a second. According to Stewart, the current 2.0s is actually a 2.08s. This means that we are only 0.09 seconds away from a 1.9s. That looks like nothing, but it's actually 4,3 % of the total time. But we actually cut off 63,2~65 % of the WR in 1 year, so why not ?

In order to get a 1.99s, the number of centiseconds (cs) between two clicks has to be low. Needless to say, no error can be done. The current lowest time ever done with one error is 2.47s, very far from a sub-2s. So that means that only 20 clicks (because of 20 tiles) have to be done. But it's not as easy as that. As you can expect, we can't just divide 1.99 by 20 and go "Ok, we just have to click a tile every 0.0995 seconds". The timer starts at the moment the grid appears, and the grid is different every time. There is a total of 670 442 572 800 different possible layouts of grids in Easy difficulty. So that means that there is always a delay between the start of the timer and the first click. In average, that delay, that is called visual stimulus is around 20 cs, but with training, it can be a bit lower, down to 14 cs. On my side, I would say that it's more around 20 cs, which means that in order to get a 1.99s, each click should be separated by 1.79/20= 0.0895 seconds. But can clicks that fast be considered multi-touch (touching the screen at multiple spots at once) ? Well, no, and here is why :

Thanks to a little trick, it's possible to show a timer when playing a Word Search. How ? Well, just add "?timer" at the end of the url of any Word Search you're playing. I've been speedrunning some Word Searches, especially the #38, in order to see if a sub-2s would have been possible. Indeed, it happened. I managed to get a 1.7s, plus I recorded it. The #38 puzzle has 11 words, and finding a word is 2 clicks, therefore there is a total of 22 clicks. But how can a sub-2s be done on this and not Snap (yet) then ? Well, because of two things :

- First, the timer starts at the moment the first word is founded, therefore the 1.7s only accounts 20 clicks, and starts at the moment the 2nd click is done.

- And second, the grid is the same every time, which removes all the randomness.

But yet, 21 clicks have been done in only 1.7s, which is ~0.081 seconds between each click (in average), and that is less than the 0.0895 seconds mentioned earlier. Why 21 clicks and not 20, you might ask ? Well, because the gap between the 2nd click and the 3rd has to be short too, because the timer starts at the moment the 2nd click is done. If you project that in Geography Snap, you could say that the 1st click of a game is actually the click done on the "Play" button.

So anyway, all that to say that 0.0895 seconds isn't considered multi-touch on JetPunk. But this doesn't mean that a sub-2s is doable, because there is another factor that has to be taken into account...

Luck is by far the biggest threat to overcome the 2 seconds barrier, especially the layout of the grid. The layout of a grid in an Easy game mode looks like this :

When playing on a phone, this grid can be subdivided like this :

The red area symbolizes the tiles that are optimal for the left hand, and the blue area is the same thing, but for the right one. As you can see, the leftest and rightest column are exclusive to the left and right thumbs, respectively. And so, for an optimal layout, these situations have to not happen :

As you can see, both the green and blue pairs are problematic. As we insinuated earlier, the leftest column isn't in the range of the right thumb, and the same goes for the rightest column with the left thumb. That isn't optimal for a good time, because, in order to match the green pair, for example, the right thumb has to move to the left, click, and then go back to the right, which costs time. The same goes for the blue pair with the left thumb too. But that's not the only problematic scenario :

The problem of this grid is that there is no "close pair". A "close pair" (short form : CP) is when the two tiles of the same pair as next to each other. There are 2 types of close pairs : horizontal and vertical. Have a look at this.

Obviously, the horizontal close pair (HCP) is the red one, and the vertical close pair (VCP) is the yellow one. In fact, vertical CPs can sometimes not be useful, if it's on the leftest or rightest column, for example. But why are close pairs so important, you might ask ? Well, because they are easily recognizable, of course ! Getting a good time in Geography Snap is not only about clicking speed, it's also about the fast recognition of the tiles, and having the two tiles of the same pair next to each other is very useful for that. It's undoubtedly easier to recognize two tiles that are next to each other than two tiles that are completely opposite of each other. But what's the chance of having one horizontal CP, or two, or one horizontal CP and one vertical one or any number ? Well, I'm not good enough in maths to calculate that, so instead, have a look at this table :

Basically, I played 100 games of Flags Easy, and for each one, noted the number of horizontal & vertical close pairs. In this table, the rightest you go, the more HCPs there are, and the "downest" you go, the more VCP there are. So for example, if you look at the square that says "7", that means that out of the 100 games I did, 7 of them had 2 HCPs and 1 VCP. So as you can see, the upper-left we go, the darker the squares are. Getting no HCP or VCP is very common, and getting one HCP seems more common than getting one VCP. I don't have a real explanation, but this might be because a vertical 5*4 grid can hold up to 10 HCP, and only 8 VCP. But anyway. In order to get a good time, at least 3 HCP are needed. Looking at this table, this only happens 1/20th of the time though. This might not seem a lot, but 20 games can actually be done pretty fast. At maximum game playing pace, they can be done in 2 minutes.

Remember, I said that a 1.7s in a Word Search was possible because the clicks to do are the same every time.. So why not do that in Geography Snap ? Why not just play games over and over, with the same clicks on the same tiles every time ? Well, let's calculate the chance of finishing a game that way. When clicking the first tile randomly, you have a 1/19 chance of clicking the tile that pairs it, because there are 19 tiles remaining. Then, when clicking on the third tile, you have a 1/17 chance of clicking the right tile because there are 17 tiles remaining and so on and so on until the last pair. So there is a 1/19*17*15*13*11*9*7*5*3*1 chance, or a 1/654 729 075 chance to finish a game by randomly clicking. Considering roughly 1 200 000 Geography Snap games have ever been played since its creation, this idea is not viable.

Well, with all of that, it seems like a sub-2s would be possible, but just not yet. It's like everything, time is the key. Before the first 2.3s was done, seven 2.4s had to be done before. Before the first 2.2s was done, five 2.3s had to be done before. But wait, the first 2.1 was done before the first 2.2, right ? Surely that cannot happen with the 1.9, right ? The first 1.8s cannot be done before the first 1.9s, right ? We're not even sure that 1.9s is possible, so 1.8s is just out of the question, right ? Right ? ...

Wrong.

After I wrote 95% of this blog, I decided to play a bit of Geography Snap. I wasn't expecting anything really, just to make my brain not overheat from writing this complicated mess. And that's where the magic happened. Ladies and gentlemen, today, May, 28th 2023, the first EVER sub-2s has been done ! And it's not a 1.9s, it's a 1.8s ! So that means that the first 1.8s has been done before the first 1.9s, which, as I am writing this, has never been done yet ! This sounds as ridiculous as it is, really. You'll find a link towards the video of the 1.89s in the pinned comment. I really hope that you enjoyed reading this blog as much as I enjoy writing it, and we'll see again very soon, 'cause multiple blogs are meant to be submitted in June. Until then, good bye !