AI Predicts PUBG Player Placement From Stats and Rankings

An anonymous reader shares a report: Fun as the element of surprise may be, matches in PUBG might be less dynamic than they seem. That's the assertion of researchers at the Department of Computer Science at the University of Georgia, who tested several AI algorithms to predict final player placement in PUBG from in-game stats and initial rankings. As the coauthors explain, each PUBG game starts with players parachuting from a plane onto one of four maps containing procedurally generated weapons, vehicles, armor, and other equipment. To train their AI models, the team sourced telemetry data recorded and compiled by Google-owned Kaggle, an online machine learning community. In total, it contained 4.5 million instances of solo, duo, and squad battles with 29 attributes, which the researchers whittled down to 1.9 million with 28 attributes.

Most players don't rack up any kills, the team notes, and only a small fraction manage to win with a pacifistic strategy. In fact, 0.3748% of the players in the corpus won kill-free, out of which 0.1059% players won without a kill and without dealing damage. They also observed that players who actively traverse maps -- i.e., walk more -- increase their chances of winning; that 2.0329% players in the sample set died before taking a single step; and that with players with fewer kills who prefer to battle solo or in pairs had higher chances of winning compared with players who played in a squad.

People who play better

[phantomfive]

People who have more skill tend to beat people who have less skill? Is that what I am reading?

Re:

[the_saint1138]

The results (and therefore the summary) are pretty full of errors. For example, they proved that distance traversed correlates with increased odds of winning, not that increasing traversal distance would increase a player's odds of winning as stated.

Predictability vs randomness

[Solandri]

You want some random variability in a game, but not too much

• If a game is too random (everyone has an equal chance of winning), then there's no avenue for improvement. The game becomes boring, and people lose interest. (Unless you allow them to place bets, in which case it becomes gambling, and people will go into denial that it's random, making up all sorts of imaginary trends in their minds).
• If a game is not random enough (the best player always wins), then the outcome is predictable before the game even starts. The game becomes boring because there's no point to playing it, and people lose interest.

Somewhere in between there's a balance where the game allows individual differences in capability and improvement to shine through, yet there's enough randomness that underdog players still have a chance to win. If the game's outcome can be completely predicted by an AI, then it won't be a very interesting game for the players. There needs to be some unpredictability for the game to remain interesting. But if the game's outcome can't be predicted at all by an AI, then it's basically a roulette wheel and not very interesting.

Re:

[Lordpidey]

If a game is not random enough (the best player always wins), then the outcome is predictable before the game even starts. The game becomes boring because there's no point to playing it, and people lose interest.

Yeah, chess was a flash in the pan that lasted about six months since there's no random element.

Re:

[phantomfive]

The randomness in chess comes from paths that are too deep to be calculated, but the human (or computer) makes the move in that direction anyway.

Then you pray to the chess god Caissa and hope the odds are in your favor.

What is PUBG?

[Ubi_NL]

Really, is it an abbreviation for a game name? Is it a company? is this really expected to be common knowledge?

Re:

[Bigjeff5]

You do realize you're on a website with the tag line "News for Nerds", right? The odds of someone in this community not knowing what PUBG is is extraordinarily low.

Case in point: considering your first guess was a game name, it's obvious that you yourself are not confused. Nobody else in the comments is confused either.

Do you go around saying "What is Windows? Are we just supposed to know what that means? What about Linux? What is that?" Of course not.

Because this is a website for nerds. Of course we all kn

Re:

[Anonymous Coward]

You do realize you're on a website with the tag line "News for Nerds", right?

"News for Nerds", not "Gabz for Gamerz".

Greatest buggy game

[Chewbacon]

My buddy turned me on to this game. We like the strategy required compared a game where you can simply respawn over and over. He described the game as "better than sex." I reflected on this: I've played maybe 900 PUBG games and maybe won 30 of them; when it comes to sex, my odds of winning are much better than that.

Re:

[mentil]

Nah, sex is more like Dark Souls: you encounter Ornstein and Smough and give up.

scouring kaggle for publications

[hraponssi]

Well, there was a PUBG data analysis competition on Kaggle that ended just 4 months ago. The authors even note sourcing data from Kaggle. And now this appears on arxiv. Surprise.

So I guess it is now trendy to go scrape some best models from a Kaggle competition, re-run them, write some generally vague paper about them, and go pick your week at a nice conference location, score some points in academia. Oh yeah, why didn't I think of this when I was still in that world.

BTW, the best solutions in that competit

Re:

[phantomfive]

It's amazing how many problems for which machine learning has been proposed can be solved more easily without machine learning. It's to the point that any time someone proposes machine learning as a solution to a problem, I roll my eyes.