Liquidity Provision
In a handful of the YOLO Games titles, players play against a liquidity pool (or LP).
The liquidity pool is made up of funds contributed by users, who are incentivized to do so due to the LPβs fee. Naturally, this requires a system that favors liquidity providers over time.
To build this system, we consider the LP as just another player to compute risk sizes for optimal growth. We then restrict the funds playable by the player/counterparty to the aforementioned risk sizes.
The best way to do this is using the Kelly criterion.
The Kelly criterion is a formula for maximizing the expected growth of funds based on probabilistic outcomes.
To do so, it uses an equation that considers each player, their odds, and their expected value in the specific game.
A note before we proceed: using the maximum risk permitted by the Kelly criterion does create a risk of the LP going to zero. Of course, we canβt have that happening β so weβll take a fraction of the Kelly criterion to reduce the threat of a severe drawdown. Granted, this will also limit the growth rate.
The Kelly Fractionβ
The Kelly fraction gives us the optimal fraction of funds that should be risked per game in order to achieve the maximum expected growth rate.
To find it, we maximize the expected value of the logarithm of wealth, or, equivalently, we maximize the expected geometric growth rate.
Traditionally, players have used this method to calculate their risk threshold for a given game. But, since we're considering the LP to be a player, we can just use it inversely.
Binary-Outcome Gamesβ
Weβll start with the Kelly fraction for binary-outcome games to create a model for the LPβs growth rate. Ready for some crazy mathematical shit?
Alright, letβs add some variables. p is the probability that the LP wins a round. Accordingly, (1-p) will be the probability that a round is lost. b will be the payout following a successful round.
By winning a round with k% of your funds, your total holdings will become (1+bk), where 1 is the original size of your holdings. If you lose a round (with the funds lost being a), then the total holdings become (1-ak).
Still following? Hereβs how weβll model growth rate:
Where:
- G is the growth rate
- b is the payout from a successful round
- k is the percentage of holdings used
- p is the probability of a successful round
- a is the loss from from an unsuccessful round
Remember, the Kelly fraction is the optimal fraction of funds that should be risked to achieve the maximum expected growth rate.Another way to reach this value is when the first derivative of the growth rate equates to 0.
With this in mind, letβs clean up the previous equation for easier derivation. Weβll remove exponents by taking the natural logarithm of both sides (1), use the multiplicative properties of logarithms (2) and move the exponents to the front of the natural logs (3) to obtain the first derivative with respect to k for both sides of the equation (4).
Let's simplify further:
From here, itβs just basic algebra to rearrange for k.
Finally: