Can Bitcoin Overcome the Price of Anarchy – Bitcoin Magazine

by James · August 25, 2022

This is an opinion piece by Shinobi, a self-taught Bitcoin educator and tech-savvy Bitcoin podcast host.

The Lightning Network as a payment routing network has many similarities with the Internet itself. You must be connected to the network, payments are routed from a source node on the network to a destination node just like data packets on the internet and it requires an uninterrupted connection from the source to the destination. It also has a huge difference – the requirement for liquidity. On the Internet, as long as the bandwidth is available (ie the pipes are not “clogged”), you can send an infinite amount of information along a route as long as you have enough time to wait for it to come through. However, lightning channels can be depleted, as they actually require moving money from one side of a channel to another to route a payment, and eventually they will run out of money on one side and push everything to the other.

This creates a necessary balancing act between the use of the network in the present to forward payments for individual users and the health of the network in the future in terms of the ability to forward payments for other users. Every time someone routes a payment through a particular channel, they increase the likelihood that the channel they used will not be able to process payments in the same direction for other users in the future.

Essentially, users trying to en masse strategies to benefit themselves in terms of guaranteeing payment delivery can have negative effects on the overall liquidity distribution of the network and actually reduce the likelihood of individual users’ payments reaching their destination. Essentially, whichever strategy is dominantly used by end users to choose routes for their payments will have systemic effects on the entire network. In the negative sense, ie how individual behavior has degrading effects on the system as a whole, this dynamic is known as “the price of anarchy.”

Rene Pickhardt has engaged in a line of research to develop heuristics useful for improving the reliability of payment delivery across the Lightning Network. One strategy to achieve the goal that has emerged from this research is referred to as “Pickhardt payments.” Currently, the most widely used strategy across the network is to prioritize route selection based on the lowest toll. This works pretty well for small payments, but not so much for larger amounts. Intuitively, the reason should be obvious: such low-cost routes are widely used which tend to push liquidity in one direction, leaving less available. The effect this has on other small payments taking the same route is small until they approach exhaustion, but for larger amounts the chances of success are lower.

Pickhardt payments work by prioritizing reliability over cheapness, making educated guesses about the probability of a payment succeeding over various potential paths it could take. Just like the low-fee dominant prioritization strategy, over time, when a node tries to make payments and sees some fail, it will update its assumptions about the probability of payment success and over time improve its accuracy. This should help prevent nodes in swarms from always draining the same channels, because their view of the network in terms of reliability will evolve uniquely over time.

An important part of choosing a path is assessing the direction in which liquidity flows in a channel. Is it balanced both ways? Is it mainly one direction? In his latest research looking at the dynamics of the price of anarchy, Pickhardt noted his realization that based on public gossip data it may be possible to estimate the rate of drainage in channels, how balanced or unbalanced the flow through it is and further improve the reliability of estimates of payment success or failure along certain routes. By estimating this correctly, you can look at a channel and guess which direction has a high probability of making a payment and which direction has a low probability.

Another aspect of Pickhardt payments is to optimize for both reliability and low fees. By modeling things to study the price of anarchy dynamics of the Lightning Network, it was discovered that optimizing for both reliability and fees lead to one of the worst network externality costs or the highest price of anarchy. This appears to create the largest frequency of channel depletions across the network of all lane selection strategies.

Now these effects do not exist in a vacuum or without counterbalances. Routing nodes on the network are also actors who have tools at their disposal and can adopt strategies to optimize flow control and counterbalance this. Routing nodes can change fees to discourage pushing liquidity to one side of a channel, i.e. if most payments flow one way, they can charge higher fees for that and lower fees for going the other way. Nodes can open or close channels, creating new connections to meet higher demand. Nodes can also rebalance channels, pushing liquidity from one channel out into the network and back into another channel to change the liquidity distribution in that channel. Nodes sending payments may also choose and use different routing strategies when they observe that the current one leads to frequent payment failures.

I’m sure people reading right now are thinking something along the lines of, “Who cares, the market will sort itself out, Lightning is a market-driven system.” Lightning is an almost entirely market-driven system, but it is not as simple when analyzing dynamics as the price of anarchy. Users of the network are not going to manually analyze routing algorithms, pick and choose what to use with each payment; They are going to use tools and software that automate all of this and hide it in the background. This makes this type of research important to the overall health of the network. There must be a way to enable end users to engage in the network in a selfish manner, prioritizing their own interests, without degrading the performance of the network as a whole.

Modeling how these two dynamics interact, the node dispatching strategies and node routing mitigation strategies is incredibly important to develop strategies for both classes of users to balance and optimize the overall health of the network and the reliability of payments for individual users. Routing data between different devices is a long-solved problem in computer science, on which the Lightning Network builds heavily, but the dynamics of liquidity constraints add a new facet to the entire field of research around reliable routing of information.