There is also a discussion about it over at HN: https://news.ycombinator.com/item?id=7295850
Games have gone from selling at a 50% markup in the 80s to actually being sold at a loss, as a loss leader, in retail. This means retail has lost control over this part of the value chain. In the 90s he worked at Microsoft, and Microsoft would sell to resellers who would sell to stores, and which warehouses sold out would be all the information they would have. They had no insight into OEM accounts or anything outside the United States. Microsoft commissioned a study, found Windows was being used on 30 million machines in the US, but was the #2 product, being outsold by Doom, made by a 12-person company in Mesquite, Texas. How did this make any sense? It took years to build up enough distribution strength to go from #7 to #6 in the word processing category. Doom didn't have a 500-person sales force. It had a completely different approach. There were structural changes that were affecting the relative value of business functions. Whole categories within typical corporations were essentially being made obsolete. Communications to your customers were on an exponentially decreasing cost curve. Everything associated with talking to a customer or delivering a product to a customer. And this would be true whether you're doing digital products or physical products. It would just be more obvious when you also had low cost for incremental production. All the costs associated with building a video game tend to be associated before you sell your first copy. Selling the copy + 1 is relatively cheap.
They went against the movement towards outsourcing -- find the cheapest person in the world to do the job. They decided to use the expensive talent. Talent just means, the ability to be productive. In the software space, it's relatively easy to look at relative productivity of people within that space. IBM in the 1980s shipped 1000 debugged lines of code per developer per year. On Half Life 1, one developer was shipping 4,000 lines of code per day. ( :O ) And that's without even arguing about whose lines of code are more useful. And his lines were more useful than somebody developing a 3270 terminal emulator for OS/2.
They assumed the same variation in productivity was probable true across most roles, and designed Value around that idea. So, how do you attract, and retain, the most highly productive people in the world?
So, this had certain consequences:
1. Valve is not a publicly traded company. Control around decision making now involves 3rd parties. At Value, only the customer matters. There's no notion of a board making decisions or a distribution channel insisting on a product should fit some category.
2. People don't need titles. Titles are the enemy of something that makes them more productive. Often solutions in one generation that are different from the previous generation. An explicit organizational structure, captured in titles, makes it less likely for them to have the insights to build the next generation of whatever is going to be exciting.
3. Management is a skill, it's not a career path. Everybody is a mixture of individual and group contribution. There's a set of tasks related to project organization and keeping things going. And usually, people refuse to do it twice in a row, on back-to-back projects, because it's very much a service job. "My job is to entirely define myself in terms of the productivity that I enable in other people. That's a very stressful job and it's hard to measure your own productivity. People say, hey, Jay, you should do it again, and Jay says 'screw you guys!'." So we look for some younger sucker to give the job to, who thinks it's authority within a hierarchy related to decision-making, and then finds out that it's, oh, working really really hard to make other people more productive. They have no QA department. They have no marketing department. They have one guy who calls himself VP Of Marketing because it confuses people outside the company if he doesn't tell them that when he's talking to them. But everyone in the company talks to customers.
4. Everyone designs their own office space. The company went from 1 person 1 office to desks on wheels, so someone can pick up and join another group in 15 minutes.
5. If you're not making quantitative predictions, you're doing it wrong. Anybody can explain anything after the fact.
6. You have to be really aggressive about firing people. It's sink or swim. They have a term, "beaten wife syndrome", for people that come in from other industries and are used to any time they show initiative, somebody's going to leap out and smite them. It takes 6-9 months for people to adapt.
They conceived of single-player games being like feature films, while multi-player games being like sports. But that model broke down in free-to-play games. The incremental value of an audience member is greater than the incremental cost of making that person an audience member. So free-to-play games sound like suicide but they aren't.
In games with auction houses, someone will play your game 20 hours a week for 4 years, and end up with no value.
10 times the content comes from the user base (for Team Fortress 2) as comes from Valve.
The most anybody has earned in a single year is $500,000. (Broke PayPal in the first 2 weeks.)
Some employees from other game companies would make more money selling content in their games than they made as employees at their company.
They started to see inflation, deflation, users creating their own currencies. In Korea you have to make the equivalent for a W4 form for the income they players make in the game. Financial crises would erupt in the TF2 hat exchange. They didn't know what to do so they hired an economist.
Steam is a curated store but they are going to move away from that because they are not trying to create artificial shelf space scarcity. They want to change it to a network API publishing model. People can connect their own stores to the Steam backend.
He is working on getting prediction markets easy to set up and use. Within 3 months of prediction markets going up, prediction markets will be vastly more reliable than game publisher's own projections for their own sales of their own products.
One of the biggest mistakes game developers make is mistaking a genuinely entertaining action by a player as hacking. The example game developers tell each other is when Lord British was killed after his first speech in Ultima Online, they rolled the game back, rather than recognizing that it was the coolest thing that had ever happened in Ultima Online, and they should have let that instance turn into whatever an instance turns into after you kill off Lord British. A general property of social system is people have to have confidence in the future. If property rights are provisional, then your value of your property changes accordingly. If you look at communities where the homicide right is high compared with North America, you'll see that lack of security ripple throughout the society.
Releasing source as a goal to increase productivity hasn't been successful. It's too hard for most people to add value.
Communication and access to data are becoming exponentially cheaper, and one of the consequences of that is that the moral hazard agency costs in information asymmetry. Or: you can't lie to Reddit. Reddit's ability to detect bullshit is insanely high.
He doesn't like consoles. Closed platforms increase the "chunk size" of competition, which is a bad idea, and the cost of entry to the market, so people who have good ideas, it's a lot harder for their productivity to be monetized, and it also delays standardization. Most roles in corporations are rent-seeking behavior on top of friction. Standardization is a great way to to eliminate the ability to rent-seek. The internet does a better job of organizing a bunch of individuals than General Motors or Sears does. Corporations look like pre-internet ways of organizing production and allocating capital.
3D printing is making the physical world look like video games: on-demand making of custom items on a global basis. The lessons learned in the video game space today are going to be true for a much wider set of industries tomorrow.
- Red HatSoftware Engineer, 2013 - present
Linus Torvalds: Software and Process Patents Don't Make Sense | Muktware
I think copyrights, and especially patent disputes, get really nasty really quickly. Especially on patents, it's like winner take all on co
Linux Today - Fedora 17 heads up: gnome-shell for everyone!:Fedora 17 he...
Internet.com - The Network for Technology Professionals. IT Professionals. IT Management; CIO Update; Security; Storage; Server; Networking;
Linux Today - Richard Stallman's Personal Ad:Richard Stallman's Personal Ad
Login. Manage My Profile. Register. Click here. Click here. Linux Today: Linux News On. search.internet.com. Linux News Sections: Blog - Dev