Fifty — Everyone knows 2048, but what about Threes!?

As the end of the year approaches, it suddenly occurred to me that 2024 actually marks the 10th anniversary of the once wildly popular game, 2048.

However, few may know that 2048 is actually a copycat of the truly brain-burning game, Threes!

History

Before diving into the specific gameplay of Threes! (known in Chinese as 《小三传奇》), let's first revisit how 2048 became the true “side chick,” stealing Threes!'s spotlight and winning the hearts of many.

The original idea for Threes! began taking shape back in 2013. Although the prototype was conceived in just one night, developers Asher Vollmer and Greg Wohlwend spent the next six months gradually refining their concept. After a 14-month development cycle, Threes! was finally released on iOS on February 6, 2014, priced at $1.99 (now increased to $5.99 due to inflation).

Although I didn't have the means to pay for it at the time, I had a jailbroken iPad 2, so I was able to enjoy the game for free (I later purchased it legitimately on Google Play using a Taobao gift card, which felt like buying the two developers a cup of coffee).

But the good times didn't last long. Because the game's mechanics were quite simple, the cost of imitation was relatively low. On February 19 of that same year, a developer successfully ported Threes! to the browser[2], allowing players to enjoy the creators' innovation without paying the $1.99.

Then, on February 27, developer Xun Yang released 1024! on the App Store, with the promotional tagline "No need to pay for Threes." This version was also later ported to the browser.

On March 9, 2014, developer Gabriele Cirulli created the web-based game 2048[3] and made its source code available on GitHub under the MIT license[4]. And then, as you all know, 2048 went viral. The development team behind Threes! expressed confusion over 2048's popularity, as these developers did not acknowledge borrowing Threes!'s core idea. They attempted to critique 2048's game design and even wrote a 45,000-word post detailing their 14-month development journey. Vollmer even tried to have these clones removed from the App Store but ultimately did not succeed.

Of course, the original Threes! eventually received the recognition it deserved, including but not limited to[5]:

Apple Game of the Year 2014!!
Apple Design Award 2014 Winner
The top 10 games of the generation (Eurogamer)
The 30 best games of the decade (Tom's guide)

Rules and Difficulty

Here, I will explain the differences between Threes! and 2048 from the perspective of 2048—the core distinctions really boil down to just three points:

The arithmetic rule in 2048 is 2 + 2 = 4, 4 + 4 = 8, and so on; whereas in Threes!, it's 1 + 2 = 3 (or 2 + 1 = 3), 3 + 3 = 6, 6 + 6 = 12, and so forth.
In 2048, when you swipe, a tile will keep moving in that direction until it either hits another tile that cannot merge or reaches the wall; in Threes!, a swipe can only move tiles by one space at most.
In 2048, newly generated tiles can only be 2 or 4; in Threes!, new tiles can range from 1 up to 1/8 of the highest tile on the board, and the next tile to appear is shown as a preview above the screen.

It is precisely these three rule differences (especially the third one) that lead to a significant gap in difficulty between the two games. Every time I start a game of 2048, I can almost always complete it within 10 minutes, with a success rate of over 90%.

So I just started another round. I opened the webpage at 3:21 PM, and after taking the screenshot, I checked the time—it was 3:30 PM.

But what about Threes!? Let's look at the score distribution of this player who just achieved 6144. You can see that even for a high-level player who has reached 6144, their average score still falls between 384 and 768.

Source: Zhihu, username "梦说东西," dated June 25, 2024
https://www.zhihu.com/question/22722456/answer/106174512

As for me, over the past 10 years, I've played several hundred rounds, achieved 6144 twice, and just now played a round where I scored 1536—which is about my average performance lately. With a little understanding of the rules, we can equate beating 2048 with reaching 1536 in Threes!.

Clearly, a game with over a 90% success rate is much easier than one where "average performance" is the norm. Therefore, we can conclude that Threes! is significantly more difficult than 2048.

Game Strategy

2048

Let's start with 2048. I believe the only key strategy in 2048 is to arrange the high-value tiles in an S-shaped pattern from largest to smallest. To achieve this, we have the following measures:

Keep high-value tiles clustered in a corner.
Pay attention to blocking positions in the bottom row.
Avoid forming a 2x4 or 3x4 layout.

The first point is easy to understand—it's just another way of describing the S-shaped arrangement, so I won't elaborate further.

For the second point, let me briefly illustrate with a diagram:

Clearly, whether you move right or up at this point, the 1024 tile will break away from its corner position, violating the first principle of the strategy. Those who have played 2048 know that once a high-value tile leaves the corner, it's very difficult to restore it to its original position, significantly increasing the risk of game over. To prevent this, immediately after merging a new high-value tile, you should use low-value tiles like 2 or 4 to fix the relative positions in the bottom row. This is what we call "blocking."

The third point usually signals that the game situation is deteriorating (2x4) or nearing its end (3x4), as shown in the diagram below:

At this point, you're forced to move upward, breaking the previously formed S-shaped pattern. Moreover, if not handled properly afterward, it can easily lead to game over. Therefore, it's crucial to avoid such situations at all costs.

But honestly, these are the main challenges in 2048. With enough practice, clearing the game isn't actually that difficult.

Threes!

Now let's talk about Threes! — it's a whole different story. There are two key points to note:

The S-shaped strategy is not optimal, but high-value tiles still need to be concentrated in the corners.
Balancing the tradeoff between merging 1s and 2s and merging other tiles.

Regarding the first point, some readers familiar with 2048 might be surprised — isn't the legendary S-shaped tactic effective? Personally, I don't think so. In fact, more aggressive players have even proposed a so-called "edge strategy" to improve flexibility (see [6]). I prefer a strategy somewhere in between — keeping high-value tiles as close to the corner as possible while maintaining a certain degree of flexibility, as shown in the example below:

This resembles forming a triangle shape with the high-value tiles. Why is this better? It mainly comes down to the second and third differences between Threes! and 2048.

Most obviously, the S-shaped strategy doesn't handle randomly generated high-value tiles well. If you use the S-shaped strategy, the first and second rows are usually fixed. If a tile with a value of 1/8 of the highest tile appears, it's often difficult to move it to the second row to merge with the corresponding tile in the first row (as shown in the image below). In contrast, a triangular arrangement of high-value tiles makes this much easier to handle.

Does this make your head spin? Does it feel like this round is about to end?

Secondly, an unwritten rule also makes the triangular strategy superior to the S-shaped strategy — new tiles always appear in one of the empty spaces on the opposite side of the swipe direction (e.g., if you swipe left, the new tile will appear in one of the empty spaces on the right). This means that if you anticipate a sufficiently large tile is about to appear, you can quickly "integrate" it into your existing triangular formation, avoiding the awkward situations that occur with the S-shaped strategy and increasing your chances of survival in such scenarios.

Regarding the second point, I don't have much experience that can be easily put into words, but my personal feeling is to prioritize merging 1s and 2s. This helps avoid situations where tiles are "misaligned" (i.e., larger tiles are separated by lower-value tiles, often 1s and 2s). However, there is one exception: when there's a long and volatile chain of merges involving tiles of 3 or higher (e.g., involving four or more elements, such as 3 → 3 → 6 → 12) that could easily break after an unrelated move (e.g., 3 → 3 → 6 → 12 becoming 3 → 1 → 6 → 12). The finer details of handling such situations can only be mastered through practice, developing a "feel" that guides your decisions.

Art, Music, and Easter Eggs

As a game that took 14 months to develop, Threes! naturally put a lot of effort into elements beyond the gameplay itself. For instance, the game includes both a day mode and a night mode (and this was back in 2014!). The soundtrack is also incredibly enjoyable, and just this year, a new night mode soundtrack was added (I only found out while writing this post and immediately decided to finish writing so I could go experience it).

If you take a close look at the tiles in Threes!, you'll notice that each one corresponds to a different character. You can read a brief introduction for each character the first time you unlock it, or by checking the main menu. For example:

The game also includes several Easter eggs:

When you unlock the 6144 tile, you might occasionally catch a glimpse of the 12288 tile's shadow on the main screen. For example:

If you swipe right on the "Special Thanks" screen, you'll find abstract drawings of all 12 characters:

When you truly beat the game (by merging up to the 12288 tile), a special Easter egg video will play, showing your total playtime. I won't spoil what the video looks like here. You can reach this ending by playing legitimately, watching someone else's playthrough, or ~~modifying the memory~~.

Reflection

As the creators of Threes! themselves said: "Threes grows with you and you'll grow with Threes." I believe this game truly lived up to that sentiment. Thank you, Threes!, for adding some color to my childhood and youth. If I ever find the time in the future, I will continue striving toward the goal of reaching 12288.

At the same time, I can't help but feel a sense of regret and powerlessness over how this brilliant idea was borrowed and overshadowed by others—this is another reason why I wrote this post. Even ten years later, such phenomena still exist and will likely continue into the foreseeable future. Nevertheless, it's important to have some words to remind people that, at the very least, she once existed in this world.

If you've made it this far, perhaps you already understand what's in my heart.

——Merry Christmas