Last year, Valve put limitations on Steam trading cards to stop sketchy developers from making a living leeching money from the user-powered trading card economy. Now it’s doing something similar with Steam achievements.
Illustration: Sam Woolley
In a post in Steam’s private developer-focused Steamworks group, Valve announced that achievements – like trading cards – will now be subject to a “confidence metric” that prevents features from kicking in until it’s been determined that a game is legit.
Until games reach that point, they will be limited to a maximum of 100 achievements. Those achievements won’t count toward your total achievement count and can’t be displayed in Steam profile achievement showcases. In addition, those games won’t show up in your library’s total game count, nor will they be visible in profiles’ game collection showcases.
Over the past year, achievement spam games have become a genre unto themselves on Steam. On Steam, achievements can be used to decorate your profile page and create the illusion that you’re an unstoppable MLG 420 pro, and games with a ton of easy-to-earn achievements make that easy to do.
They’re often exceedingly simple, hastily assembled from preexisting assets, and loaded with thousands of achievements that players can earn for doing things such as walking, jumping and breathing. Some Steam users revile them, but plenty of others see them as a fun way to deck out their Steam profile pages.
Valve, apparently, didn’t like the impact achievement spam games were having on the store – or on its precious algorithm.
“As you can guess, fake games were inflating achievement and game counts for users to display on their profiles,” Valve wrote in the Steamworks group. “Our data shows us that an insignificant number of users were taking advantage of this, but the existence of these fake games still confuses our algorithms and users.”
From now on, if a game falls under the new constraints, its store page will display a note about it.
Comments
11 responses to “Valve Adds Limits To Steam Achievements To Fight Rise Of ‘Fake Games’”
Sounds like yet another case of Valve attempting to build algorithms so they can avoid having to do actual work curating their store.
That would seem to be in line with their recent announcement, don’t you think? They’re doing what they said they’re doing.
Yes, very much so. And very much in line with what they’ve always done. Throw a computer at the problem, that’ll fix it for sure. And when it doesn’t, you throw your hands up and say “I guess it can’t be helped”.
Please fix this god-awful edit moderation, it is cancer.
Sure. But anything beyond a set of strict rules (ie. something an algorithm can perform) introduces subjectivity, and subjectivity leads to inconsistency, both of which they’re trying to avoid. An algorithm applies its rules consistently and can be tailored to each individual user of the system; a human doesn’t and can’t. Personally I’d prefer the algorithm, I want control over what I see.
You seem to think that Algorithms by their nature are objective.
Speaking as someone that has worked on machine learning and AI systems, this is not correct. It is trivially easy to insert bias into a system like this just by picking the wrong rules or even using a skewed training set.
Even a non-AI system will have rules that are picked and enforced by a human, who has inherent biases.
The whole idea that AI is completely objective is why Valve, Google, Facebook and so on have made such a mess of everything they throw it at. Machine Learning is not objective, in fact it will see subjective things you don’t even notice yourself, treat them as a feature, and amplify them.
“Personally I’d prefer the algorithm, I want control over what I see.”
This is an incredibly contradictory statement. If you want to be in control of what you see then you definitely don’t want an Algorithm, and you probably don’t want Curation either.
I’m a software developer with extensive algorithm experience myself, I’m quite familiar with how algorithms work. All algorithms apply their rules objectively and consistently because they’re incapable of exercising judgement outside of those rules. As long as the rules are documented correctly, I have the information I need to apply them to get the result I want. AI and machine learning can not make subjective determinations unless the rules allow them to do so, and if the rules allow them to do so then they should be part of that documentation. Software is not capable of exceeding its instructions, unexpected results are the result of either error or human failure to understand the outer bounds of available logic paths.
What Valve is talking about here isn’t complex, these are basic algorithms. A confidence metric is not machine learning, it doesn’t involve fuzzy logic, it’s the outcome of a simple algorithm that measures multiple inputs.
My statement isn’t contradictory at all. Even when the rules for a human are documented correctly, they’re almost never applied consistently or objectively, which has led to the kinds of controversies that Steam, Twitch, even Youtube and Facebook have whenever people intervene to override the default algorithmic behaviour of the system. A documented algorithm that provides me the ability to specify some of its inputs, such as those that Valve is developing, is absolutely the superior choice.
I definitely want Valve to expose algorithms that are well described that I can use to tailor the results to my interests. That simply can’t be done with human evaluation.
They make money by letting everything on including fake games, You could argue that a game storefront that sells fake games has quality control issues but ppl will turn it into a debate about free speech, Valve are doing exactly what they said they would, Trust me it’s not worth getting into with the troglodytes that champion this as “muh freedom”. Just hope that Valve’s algorithms are good enough to cut through the bs.
They only make money on the sale though. The fake games cause such a visibility problem for the actual real stuff that they probably end up reducing Valve’s cut, to be honest.
Good. Now while they’re adjusting the achievement system, how about doing some upgrades to the 10+ year old system….
* Split DLC achievements away from the main game – if you don’t own a required part, the achievements shouldn’t show. Especially for simulators where many expensive DLCs may be available, but people only buy 2 or 3.
* Have a rule against repetitive or pointless achievements. 100 achievements made up of 4 actually unique copied 25 times with the values exponentially increased makes for dull achievements. And leaving the game on the menu for a week is not exactly much of an achievement.
* Achievement Reset Feature. I should be able to reset my achievements so I can try for them again.
* Free weekend and family sharing games should have their achievements shown despite not owning the game. Or should not be included in any statistics.
I felt like there should be a hard limit on the number of achievements games actually can have paired. 100, plus an extra 20 for each paid DLC would clamp down on it pretty drastically.
You then rank Achievements by how hard they are to get. This would be down to the Devs.
You then assign each achievement with XP for ur profile, with a total of 150 XP with 25XP per paid DLC. This would mean you could end up with difficult and interesting achievements worth more XP.
A hard limit based on paid DLC would be incompatible with living titles that release new content as a stream. Team Fortress 2, for example, has 520 achievements built up over more than a decade of content, none of which was released as paid DLC.
It seems to me that the only issue caused by a lot of achievements is when they’re shown as a sum value on user profiles. Rather than limiting the number of achievements, that could be more easily fixed by only showing percentages instead of counts. That way, a game with a few hundred achievements is actually at a disadvantage rather than an advantage because it takes more achievements to get the same percentage as another game with fewer achievements.