Unintended Consequences: Game Economy Management

Intro

In economics, we have this awkward concept we call the law of unintended consequence (TLUC hereafter).¹ Economists use the phrase liberally, ignoring the fact that most laypeople don’t fully internalize what TLUC even is. Indeed, the layperson’s definition can probably be summed up as negligence. One might think that an unintended consequence is the result of a gross misunderstanding or, worse yet, intentional negligence. When economists use the word, they’re typically referring to a secondary outcome caused by a treatment that was not modeled or previously understood. In fact, oftentimes an unintended consequence is, in and of itself, a discovery.

I made one such discovery this past week all thanks to TLUC.

Scanning, Fees, and Fixed Costs

One of the most common actions (transactions in blockchain speak) in the Star Atlas game economy allows a player to scan for an item called an SDU. This action is repeated dozens and even thousands of times each day by some players. There’s a cooldown period between each of these scans that lasts about 2 minutes. Furthermore, SDUs are an incredibly valuable resource in the economy. As result, 38% of all transactions coming from the SAGE game economy are due to scanning.

Each transaction sent on a blockchain must be paid for in order to guarantee its legitimacy. For the most part, each scan transaction has the same fee whether someone is scanning with a $15 ship or a $100,000 mega fleet. I call this the fixed cost problem in blockchain gaming.² In an effort to combat this problem and reduce fees paid by players we asked ourselves: What could be done to reduce the number of transactions a player encounters each day?

The number one solution in terms of ease-to-implement and bang for our buck — i.e., the solution that has a high benefit to cost ratio — was time dilation. Simply slow down actions and increase rewards proportionally. Since we knew scanning was our biggest vector for fees, we decided to slow it down by a factor of 6x. This had the potential to reduce fees across the ecosystem by a staggering 31.5%. So we did just this and, believe it or not, it worked! We saw a reduction in fees of 31.42% in the game economy.

The Unintended

While our intervention succeeded in reducing the fee burden faced by players, we also saw that the total number of SDUs produced was increasing. This was not our intention. The whole point of time dilation is that it slows things down without impacting aggregate production and, consequently, prices. While most of our metrics changed in expected ways — one stood out.

After slowing scanning down, we did see a nominal decrease in the number of scans performed by each player. In an attempt to get to the bottom of this mystery, we scaled the total scans per user per day by the dilation factor (6). This told us what the relative change in scans was compared to the pre-period.³ We found that the number of scans had actually increased by 50%. Thus, once we adjusted by the expected decrease — the 6x time dilation — we saw that the number of scans increased relative to where they were in the pre-period.

We had the unexpected outcome of increasing overall production of SDUs even though, theoretically, this was not supposed to happen! I’ve posted a chart below that shows the total number of scans before and after the update. You’ll notice that the adjusted line diverges massively from the unadjusted line after the update.

An explanation

So what’s the deal? Why are players able to fit more scans into the day after the change? It’s a story of efficiency. Before the update, players had to scan every 2 minutes or so. That turned into 12 minutes after the update. We went from a somewhat engaged mechanic to an idle mechanic. Because the idle mechanic was easier to execute, scans in real terms went up. To make it a little clearer I’ll use an extreme example.

Suppose we say a player is “engaged” with the game when they’re waiting to scan again. For example, a player scans and they have to wait 1 minute before they can scan again. They come back in 5 minutes and scan again. They’re only engaged for 20% of the time since they could have scanned during those other 4 minutes.

Using this framework, suppose the scan time is 1 minute in scenario A and 24 hours in scenario B. In scenario A, you have to actively click a button every minute for 24 hours — or 1,440 times a day. In scenario B you have to click exactly once. Humans have to eat, drink, and sleep. It’s almost impossible for someone to be equally as “engaged” in scenario A (clicking every minute) compared to scenario B (where a single click engages a user for 100% of the day).

While the above example is a bit extreme, it’s essentially what’s happening in our case too. Players all of a sudden require less physical presence and work in order to get the same reward. As a result, they have a higher likelihood to be engaged in the game. We did not account for this when we designed the update. There are other potential explanations not described here like technical limitations of the blockchain, but the principle is the same — engage time went up due to a decrease in the effort required to engage.

Conclusion

Unintended consequences happen every day. In fact, there’s probably an unintended consequence of every change we make regardless of importance. Most of the time they go unnoticed. In the case I described above, it was a bit more obvious. And, while we’re monitoring the situation carefully, it’s really fun to see such a popular economic concept in the wild.

1. Rob Norton has a brief write up here.

2. 5 cents is a much larger percentage of a $15 fleet’s income than it is of a $100,000 counterpart.

3. This is similar to the way economists measure real prices. They can measure the relative increase in an item’s price by applying a factor like the Consumer Price Index.