Chapter 481: Recommendation Algorithm Logic

Although movie market sentiment was loud and Han Han kept being criticized, box office numbers kept rising steadily.

Because Chen Zhixi did not give up propaganda, the movie "Afterwards, No More" performed even better at the box office than in her previous life.

In her previous life, this movie collapsed by day five, earning only a little over 30 million, but this time it earned over 60 million.

Of course, the most explosive was "Transformers 4," with total box office already exceeding one billion, and daily earnings still exploding.

This movie can't compare—it's an IP film with a built-in audience base.

Currently, domestic films truly cannot compete with Hollywood films; the gap is too wide, and a national awakening is needed.

On the international market, "Get Out" has been screening for about a month.

This movie has a 99% fresh rating on Rotten Tomatoes, an IMDb score of 8.2, and a Douban score of 7.8.

It ranks better than 73% of mystery films and 92% of horror films.

This movie can be called this year's highest-rated thriller, and also the first box office dark horse, with global total earnings exceeding $255 million.

$255 million equals roughly $63.75 million for China, or over 440 million RMB—still earning money from overseas markets.

Of course, the cost is negligible, only around $5 million.

Fang Hong still can't easily access this money; he wants to save it to pay off the film base in one lump sum—he's just seven hundred million short.

After "Get Out," "The Purge 4" is about to be released.

Recently, Liu Tianxian went to shoot advertisement covers and attend events for sponsors—she does this every year, and besides, it's only two months until the golden September and silver October season.

She just won an award, and now is precisely when advertisers favor her.

Her appearance sparked a wave of fan adoration; wherever she went, crowds lined the streets to welcome her.

Online, she still trended constantly—she appeared, and trends followed, impossible to suppress; Liu Tianxian now has this kind of influence.

There's no way around it—whether in status or achievements, no other rising stars or celebrities can match her; she can crush them.

Online, people still hate Liu Tianxian, mainly targeting her nationality.

This is unavoidable—the more famous you are, the more people will curse you.

Other rising stars and celebrities are also constantly hyping themselves up—since you trend, I'll buy publicity drafts to split your audience.

Da Mimi's exposure is also extremely high; her new movie "Tiny Times 3" has been released.

This movie can be called her signature work; later, when people mention her, they'll likely think only of this film.

She has no other films anyone remembers.

This movie has also faced waves of audience complaints and low ratings.

In terms of talent, Guo Jinming and Han Han can't compare.

Fang Hong recently went to Chang'e Technology—he must enhance his own strength to race against capital and hot money.

Chang'e Technology's staff has now exceeded ten thousand employees, as the company expanded and hired throughout the year, though most are programmers.

He first visited Chang'e Technology to inspect the engine—the game engine—which has been under development for over three years and is now showing promise.

It was actually built long ago; everything since has been optimization—building it is easy, optimizing it is hard.

It must run smoothly and be user-friendly.

Fang Hong stood with a group of technicians before several monitors, watching a staff member compile and debug "Hero Glory" on the new engine.

"How difficult would it be to fully migrate 'Hero Glory' to the new engine?" Fang Hong asked.

"Very difficult. Migration could cause major instability, requiring large numbers of senior technicians to debug and optimize," Wang Yuanming said.

He was the head responsible for the engine.

"Oh, so it's quite troublesome," Fang Hong said.

"Yes, there are existing tools that speed up migration, but they're unstable too—you still need professionals," Wang Yuanming said.

Fang Hong understood: it's not impossible, just labor-intensive and requiring massive work.

He asked Li Min beside him: "How's 'Honor of Kings' doing?"

"Still in testing and optimization; it should launch public beta around September 1st. Current test results are good," Li Min replied.

It'll be online in just over a month.

Of course, no one knows what will happen after launch, since we don't have Tencent's distribution channels.

But having a first-mover advantage is correct—Tencent's mobile MOBA game won't launch until next year at the earliest.

Of course, Chang'e Technology's distribution channels aren't bad either.

Fang Hong thought for a moment and said to Wang Yuanming: "Proceed with the migration. For games, we must use our own engine."

"Understood!" Wang Yuanming nodded.

The engine isn't just for games—it's also needed for film special effects.

For example, virtual-real scene rendering requires engine processing; currently, all are using U. S. company engines.

Next, Fang Hong inquired about the development progress of "Fortnite."

There's nothing much to learn here—it's just a process of technical accumulation: get the basic technologies right, and launch can only happen after "PlayerUnknown's Battlegrounds" is developed.

This game needs professional designers—the one named Brendan Greene, later hailed as the "Father of Battle Royale"—but inviting him won't automatically let him design "PlayerUnknown's Battlegrounds"; he needs experience. Wait until he finishes "Survival Kings," and his experience will be mature enough.

Inviting certain specialists too early makes it hard to get what you want, because their experience isn't ready—they might just become your test subject.

And when they join another company and create a hit, that's hard to predict.

Like "Survival Kings" and "PlayerUnknown's Battlegrounds" are both dataohasha games, but the latter has far greater influence.

So we still need to wait a while—once "Survival Kings" is out, that'll be enough.

Fang Hong has also sent an invitation email to him and discussed game development; the man has shown interest.

For someone like this, money is enough.

In his previous life, Bluehole invited him—why can't he invite him now?

After reviewing the games, Fang Hong went to the literature division, where Hua Wei was overseeing operations.

It's been half a year since the acquisition of Shengda Literature; Fang Hong originally told him to cut all other messy sections and keep only the male and female channels.

"All other platforms like Tianfang Audio, Yuedu Network, and Jinjiang Literature City have been sold or integrated; the physical division was also sold off—only the male and female channels remain."

Hua Wei reported again: "But they were all sold at a discount."

As long as they're sold, and staff are cut, even the saved operational costs amount to a huge expense reduction.

Fang Hong asked: "How's the operational status?"

"Operating normally, no major changes—we don't compete with anyone, don't even respond to provocations or poaching."

"What can you even get from poaching in this industry? It's not a technical field—editors are always replaceable; even if you lose a top author, giving more benefits to bottom-tier authors quickly brings new talent to the surface. Still, you've got to admit, our ability to develop new talent from within is pretty strong—this year, there's a relatively unknown novelist who's doing quite well."

Hua Wei spoke with confidence. "But we can't let our guard down—we must stay sensitive to external market changes." Fang Hong feared Hua Wei was too focused on his own patch and ignoring market shifts.

Many things and strategies change with the market.

"Understood, I'll pay attention. Daily and monthly active users are both stable."

Hua Wei added: "By the way, Wu Wenfei came to see me."

"Why did he come to you?" Fang Hong asked, puzzled.

Wu Wenfei was the original founder of Qidian and still held some shares.

"He came to discuss his shares—he wants to take charge of the literature division. If we won't let him lead it, he wants us to buy out his shares," Hua Wei said.

It's normal—if he can't run the operation, those shares are useless, completely under someone else's control.

"Isn't he at Tencent? Why come back here?" Fang Hong asked.

"Who knows? Maybe he can't let go—he founded Qidian, after all. It's like a child to him," Hua Wei speculated.

This work has been organized and uploaded by 【79】 In his past life, Wu Wenfei had continuously used Tencent's power to undermine Shengda Literature, aiming to acquire the company and move Qidian over.

In his previous life, Wu Wenfei had been using Tencent's power to strike at Shengda Literature, all to acquire the company and bring Qidian over.

But in this life, Fang Hong has intervened—now it's very hard for him to acquire it again, since Chang'e Technology is stable now, without the crisis Shengda once faced.

And we've carried out a series of reforms—cost-cutting, efficiency-boosting—there's no way he'd go back to Tencent.

Who knows how the market will change later? "Then have him come tomorrow—I'll talk to him," Fang Hong said.

It's better to let Wu Wenfei lead the literature division—he has decades of accumulated experience.

Hua Wei was only a temporary replacement, unfamiliar with this field, merely following Fang Hong's instructions step by step.

Of course, if Wu Wenfei wants to lead this division, Fang Hong will still keep his shares.

With that settled, after chatting with Hua Wei, Fang Hong went to Zhang Yiming—this man had been focused on algorithms, his original strength.

To better implement his algorithm theory, he needed others to align with his thinking.

So he held frequent meetings to convey his ideas to key technical staff.

At this moment, Zhang Yiming was in a meeting; the whiteboard in front was densely covered with algorithm recommendation logic—not code, but Chinese characters.

It was meant to help people understand his recommendation logic and ideas.

Fang Hong entered the meeting room; everyone saw him but didn't greet him—only glanced at him with respect, then returned to listening to Zhang Yiming speak.

In the company, Zhang Yiming was strict with subordinates—if he didn't let you speak, you absolutely couldn't speak.

Zhang Yiming also saw Fang Hong but said nothing, continuing his explanation of algorithm logic.

Fang Hong roughly scanned the logic written on the whiteboard.

The entire process was divided into three stages:

The first stage was recording user behavior to capture user group preferences and recommend content they liked—this was collaborative filtering algorithm.

It determined user preferences through explicit feedback behaviors, implicit feedback behaviors, positive feedback behaviors, and negative feedback behaviors.

Explicit feedback behaviors were clear user actions toward content, such as: likes, follows, comments, etc.

Implicit feedback behaviors were actions not clearly directed at content but still recorded by the platform, such as: viewing duration, dwell time, browsing history, etc.

Positive feedback behaviors indicated user preference for such content, such as: likes, follows, positive comments, long dwell time, etc.

Negative feedback behaviors indicated user dislike for the content, such as: blocking, clicks, "not interested," negative comments, short dwell time, etc.

This algorithm is currently used on Bilibili and Toutiao, performing well—user numbers are growing, and stickiness is high.

The second stage logic was finding other users similar to a specific user.

The system collected user preference data, calculated similarity levels between every two different users, then made recommendations for the target user.

The third stage logic was recommendation based on the neighbor principle—this was extremely complex and existed only in theory.

Fang Hong roughly understood the basic logic—he'd seen top analysts break down TikTok's algorithm in his past life.

But analysis was one thing—actually building it was hard, and technical details were nearly impossible to obtain.

Knowing something didn't mean you could replicate it—the core was always kept secret.

The meeting ended quickly; everyone picked up notebooks and pens, greeted Fang Hong one by one with "General Manager Fang," then filed out.

Only Zhang Yiming and Fang Hong remained.

Zhang Yiming put away his pen and smiled: "General Manager Fang, what brings you here today?"

"Came to take a look—saw you speaking with such depth just now. Your talent impresses me," Fang Hong joked.

"No, no—I'm still in the exploratory phase. Whether it works, how well it's applied, still needs practical feedback," Zhang Yiming waved his hand.

This kind of thing accumulates gradually and innovates—it's not something you achieve overnight.

He added: "But General Manager Fang hasn't shortchanged us on funding—we've been able to develop well."

Since Hero Glory returned, Chang'e Technology's funds were indeed ample—after all, there was still a 200-million-dollar base from the end of last year.

But as development deepened, money was consumed quickly.

"Each has its own role."

Fang Hong said, then added: "Do you know Kuaishou?"

"Yes, I've used their app—they're transitioning to short videos. Isn't Zhang Nan's team incubating something like that?" Zhang Yiming said.

"Yes, this short video format will be a future trend; we must shift our focus to it."

Fang Hong said again: "Short, flat, fast — it fits your original definition of fragmented time."

His purpose in saying this was to get him to pay more attention to short videos, ideally accelerating their development.

Zhang Yiming thought about the thing, but only had a vague concept, since he wasn't a reincarnator and couldn't possibly know everything.

In his past life, TikTok wasn't developed by him; it was developed by Zhang Nan with his team, though the core algorithm belonged to them.

"If you cut some of Bilibili's weird meme videos into clips of only ten or so seconds, it's essentially a different kind of short video," Fang Hong said more specifically this time.

This gave Zhang Yiming a clearer concept.

"The thing you're talking about should be easy to make," he said.

"It sounds easy enough, but the competition and promotion might be troublesome," Fang Hong worried about this.

Actually, the TikTok app isn't complex — any group of programmers could build it; what's complex is the backend algorithm system, the recommendation system.

"I'll go check on Zhang Nan when I have time, see what they've accomplished," Zhang Yiming said.

Only after seeing the results of that thing would he have a clear sense of things.

(End of Chapter)