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.
It surprises me that more experiments run on game data don't seem to hit the mainstream. Perhaps I am looking in the wrong places, but I've seen only a few research reports based on ingame data and those are hard to scrounge up themselves.
Video games are one of the great crucibles where everyone is born equal by default - and yet we have massive variance in wealth distribution by the end of it. You always have a select few who end up with enough power and own all the wealth.
- 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.