When it comes to providing liquidity to decentralized markets, especially those supported by Automated Market Makers, Impermanent Loss (IL) is critical to keep in consideration. IL (or the simple possibility of it happening in the future) can impact (positively or negatively) the incentives of those thinking about investing in a project.
Despite its significance, we have found little (or practically no) information about this phenomenon in Prediction Markets. We believe that there are important nuances that should be considered by cryptoasset investors and traders.
This blogpost is the beginning of a series of posts around the incentivization of Prediction Markets, and in this first instalment, we will describe the above-mentioned nuances in an introductory way and explain important concepts and aspects that will help minimize Impermanent Loss in Prediction Markets. We’ll also use a couple of examples to help us highlight these nuances.
The commonly understood definition of Impermanent Loss is that it “is the temporary loss of funds occasionally experienced by Liquidity Providers volatility in a trading pair”. It is impermanent because even though there may be volatility in the asset values, the only factor that can make permanent the loss is the removal of the assets by the Liquidity Providers (LP). However, when it comes to Prediction Markets, the reasoning is slightly different... In Prediction Markets, we refer to IL as the possible loss generated by the variation in the “probability of success of possible events” in the market in relation to the amount originally deposited.
The main difference is between the event that they are taking reference on: on one hand, the asset price is given by an event in which traders have the possibility to influence it, but on the other, the price indicates the probabilities of success of an event that will be known at market expiration, and traders have no control of that outcome (we’ll talk about this later).
Before delving further into Impermanent Loss in Prediction Markets, consider the following: although there are differences between IL’s application in exchanges and prediction markets, we could define Impermanent Loss in general terms as the “potential opportunity cost of being a Liquidity Provider”.
Impermanent Loss in Prediction Markets
Let’s clarify two important differences between IL in Exchanges and Prediction Markets: the first one is the market expiration. The other difference is the market reproducibility. Let’s explore these…
When we refer to market expiration, remember that exchange markets aren’t designed to conclude. IE. There is no explicit temporal restriction to them. Therefore, when it comes to exchange markets, we always have time on our side to allow the market to regulate itself, and thus allow the loss to become impermanent.
When we refer to market reproducibility, we mean the number of times a pool is repeated throughout the crypto space. In the case of a pair of assets such as DOT / USDT, there are most likely dozens of pools, but the number of companies or projects creating a specific prediction market is considerably smaller.
These two factors have important yet distinctive implications for Impermanent Loss. Firstly, as mentioned above, the existence of a market (pool) expiry implies the possibility of a total loss of one of the asset’s value (in the case of the option that represents the event that occurred). (Also, take into consideration market maturity, whereby the closer the market is to its expiry, the more well formed its results/opinions will be).
Second, the variations in the value of the pool’s assets are due to the direct action of the individuals in that pool in relation to the market that they are reflecting.
This means that the variations in the values of the shares of a pool will vary only by perceptions of the market in question, and not by adjustments with respect to other pools (arbitrage). This already gives us an idea of how different the structure of the impermanent loss can be in prediction markets compared to what is observed on exchanges.
So how do we assess impermanent loss in a prediction market?
Before we can answer that, we need to clarify some things:
1- Purpose of liquidity
First of all, we have to take into account the purpose of Exchanges and Prediction Markets, i.e. to know the purpose of providing liquidity. In the case of exchanges, the purpose is to be able to offer occasional buyers and sellers (traders) the possibility of exchanging two assets with each other, within the parameters of the overall market-related price. We could label the purpose of these markets as “monetary”.
With prediction markets, what is sought is to be able to provide the means for a set of individuals to interpret signals related to a certain event and buy assets that reflect the most possible outcome for these events, which results in the formation of a set of probabilities from among the possible options. In other words: Its purpose is to provide information based on incentives.
2- Value origin
As the impermanent loss will vary given the discrepancy of the ratio of assets originally contributed, it is important to identify the nature of the assets that make up the pool. In the case of exchanges, we find a price ratio determined by two present signals, tangible but generally not very predictable (except for specific assets that have a fixed value).
On the other hand, in the case of Prediction Markets, we find values that reflect the probability of success for a set of scenarios related to a particular event. Although these values depend on future events, the results of the prediction markets are usually quite close to the final results (read about this here). There is also another implication: in the first case, the action that buyers and sellers of the asset make can change the event in question (asset price), while in the second the buys and sells of shares have no influence on the final outcome. This can provide significant uncertainty.
It is at this point that we can finally tackle the question: How can the risk of Impermanent Loss in Prediction Markets be mitigated?
And the general rule for now is: Understand the market. Although, it’s not always that simple. There are two fundamental aspects for markets understanding an IL mitigation: signal perception and fee awareness.
Remember that prediction market shares have no influence on the result of the event per se, this means that if we are in the presence of a trend of buys/shares of a specific asset while the other(s) experiences the opposite situation, there is the possibility that some external event triggered a signal in the market.
As mentioned in a previous post, the worst-case loss happens when traders are very certain of the outcome and eventually change the market probability of the realized outcome to 1, so the liquidity provider needs to have the ability to interpret the market in order to know when is convenient to add liquidity, in which markets, and when they’ll need to extract their liquidity from the market. To get a better understanding, allow me to classify signals in three different classes:
The first categorization of a signal has to do with its veracity. We live in an era where Fake News has populated the web and, in many cases, this is used by entities with the intention of harming or benefiting from a possible event.
This classification of signals, which we could also call temporal/timeless, has to do with the definitive character of the signal that is being sent. Take, for example, a market that has been created to predict the winner of one of the NBA conferences (basketball): In our case, the result of the first game of the playoffs represents a reversible (or temporary) signal, since later games provide results that may contradict or validate the previous signal. Alternatively, if we were covering an internal election of a political party, and one of the candidates announces that they will withdraw their candidacy, we have an irreversible (or timeless) signal.
This last classification is related to the strength of the signal in terms of its diffusion. Generally, market signals are propagated both by the market itself and by other entities that use its data. This type depends on the quality and quantity of sources that propagate the signal. Here, “quality” refers to the weight that these entities carry in popular opinion.
These categories are complementary to each other, and allow us to generate a classification grid based on three main questions:
- Is this information real?
- Will this signal have another one that comes as a result of it?
- Who provides the information?
Another thing to keep in mind is market maturity, which we have mentioned previously. If a market is close to its end, the strength and irreversibility of the signal is going to be higher than in a recently opened market, because there is less time to revert it.
Consider some examples to fit into these three categories…
Back to political elections: Depending on where you are, it is possible that behind the elections are people with particular interests who would benefit from the results. This can cause certain sectors to start spreading fake news with the intention of helping to produce the scenario that suits them. So we have a false signal being broadcast.
Now, who spreads it? It can be a simple publication by anyone on a social network, which would imply that it does not have much power of dissemination, or it could be issued by companies that own mass media, in which case it would be widely spread due to the amount of reach and influence that these companies have on different media such as TV and social networks..
Perhaps the reward that arises from the imposition of fees on traders is the most powerful and well-known tool that an LP has to combat IL. This, added to the gap between supply and demand, are the reasons why LPs are profitable at the end of the day. However, that gap between supply and demand is not always positive. This depends on factors such as market volatility, the type of Market Maker used, and the amount of liquidity in the market.
Although AMMs are implemented to provide these incentives to Liquidity Providers, there are certain market situations that act as shocks, generating loss windows for LPs and profits for external agents (traders).
As this is a factor that can be controlled by the creator of the Scoring Rule, we set out to create Rikiddo in order to provide LPs with the highest quality of incentives possible. The main purpose of this Market Scoring Rule is to dynamically adjust the fees to encourage trades in periods of low volatility and to discourage trades in periods of high volatility and, in this way, to be able to maintain as stable a volume as possible.
In summary, being aware of how fees are handled is very important for the liquidity provider, since these are what regulate the incentives of people who want to enter the market, controlling the input and output flow.
We have to bear in mind that the concept of impermanent loss is different when talking about exchanges and prediction markets, mainly due to the fact that the state (value) of the asset in question is fed back in the markets created on exchanges, but the same is not said for prediction markets.
On the other hand, the interpretation of signals in prediction markets is important, because it is highly possible that after a certain point of market maturity (that is, as it approaches its end), the signals become clearer, making the market irreversible, rendering useless any intention of waiting for the market to revert back to the state it was in previously.
And keep in mind; although this post has covered possible losses on exchanges and in prediction markets led by impermanent loss, it is important to remember that Liquidity Providers experience gain far more than they experience loss, but these losses are important concepts in the effective management of funds contributed to the Liquidity Pool.
Throughout this post we have talked about the primary differences between Impermanent Loss on exchanges and in prediction markets. We’ve considered important concepts such as Market Expiration and Reproducibility, and a high level overview of how they work. In future posts, we will be delving deeper in to these concepts and seeking to understand Impermanent Loss in a more programmatic and systematic way.