Arbitrage And Equilibrium In The Team Fortress 2 Economy

When we were designing Path of Exile ( http://www.pathofexile.com ) we had an explicit design goal to create a complex barter economy. To that end we decided to remove gold (quite strange for an Online RPG).

We added what we call internally "currency items" which are homogeneous items that have an inherent use. For example, an item that rerolls the magic properties on another item. There are 22 of these types of items of increasing rarity.

Because they all have an inherent use, there is demand for almost all of them. Because there are so many, few people have a true understanding of what the current exchange rates are.

The rest of the items in the game are randomly generated, and are hard to pin down accurate prices on, especially with so many competing currencies. Experts can take advantage of the uninformed to trade up in value and frequently do.

You might ask why do we want to have an economy like that?

There is a large class of players for whom trading becomes a kind of end game. If your economy is complicated and interesting there is room for a huge range of skill levels, room for true expertise in the economy. The feeling of making a trade in which you gained advantage over another player is addictive and really fun.

My notes below (disclaimer- never took econ, took very few stats classes):

- Barter still exists even though players are more sophisticated. Author expected to find one item becoming defacto currency, but that does not happen.

- Because barter exists, you can express the price of items relative to other items in the game, ie 1 lazer gun = 2 hats. It is theoretically possible to start with 2 hats and end up with 6 through multiple trades of items.

- Author then extrapolates a model for an economy like TF2 that has 35k items and 600M relative prices.

- TF2 is not in equilibrium, meaning lots of opportunity for arbitrage. Arbitrage opportunities are greatest when new items and sales are introduced.

- Assuming this "dis-equilibrium", author adjusts the model by accounting for frequency of trades, ie 1 lazer gun = 2 hats but may only be traded a few times, but lets say a new item, earmuffs, are added to the game and conversion rate is 3 earmuffs = 2 hats. This might be traded many times by players due to its introduction and inherent value to them.

- You can graph the value of these items over time to see if they can be arbitraged.

- My knowledge of statistics is rather weak, but it looks like if you compute the error rate, ie square root of the differences between normal model and adjusted model, you can monitor with real time data when arbitrage opportunities arise.

Grand summary: Gaming companies produce a ton of real time transaction data, and with this data it is possible to monitor economy dynamics in new ways that economists have previously been unable to do. If players had access to this data they can arbitrage more effectively, but more importantly for game designers it might be worthwhile to be knowledgeable about these effects so they can keep players engaged in the long term.

I think it would be even more interesting to examine trade strategies in a game like Eve online. There's a giant simplification that occurs in games like World of Warcraft and now Team Fortress 2, where the only cost of arbitrage is the cost of finding out where opportunities exist. There are still interesting things to examine in games where transactions are instantaneous and safe, like how much risk people are willing to float by holding stock in intermediate goods, but the really interesting thing to me is examining communities where moving items around has real costs associated.

I would expect that the TF2 economy tends towards an equilibrium that is as simple as Yanis describes. This is probably not true in a game such as Eve. If one item is cheap to produce in on sector of space, and expensive in another, the equilibrium will reflect this fact: "arbitrage" opportunities will exist, and be stable, and be known by another name, "trade routes". There are real costs in transporting goods in Eve, for example the time cost of having a human pilot navigate a freighter, and the risk of being attacked by pirates.

In Eve you can have perfect information on the price and volume of transactions that occur, the movement of goods, the sovereignty of territory in which there are markets, the corporate affiliations of the people who operate firms and trade routes, and the wars and conflicts that arise between individuals and sovereign states. That sounds like an amazing playground for someone who is interested in liberal international relations, free trade and pacification, and any number of things that go beyond pure economics.

These posts are fascinating. In college I minor'd in econ, and I always had this feeling that economics grossly simplifies things, so it's fascinating seeing how economic principles play out in a world where there is such control over all the factors.

That there is a this kind of economy in TF2 is amazing to me. TF2 is the last great game I've ever played, but it's old in terms of video game lifespan and I feel that it's kind of a niche among shooters; you always hear of people playing Call of Duty/Halo...I can't think of a single friend or acquaintance who has ever mentioned TF2 in passing.

It's such a great, nuanced game that I thought it would attract the kind of players who would frown upon ingame purchases...and the fact that it's free also made me assume that most players would be the non-paying type. Glad to see that Valve's dedication to the game has some payoff to them.

It wasn't until he added his picture to the sidebar that I realized the author, Yanis Varoufakis, was the basis for the Heavy in Team Fortress 2.