The Partial Colored Coins is a variant of the Colored Coins Prediction Market that introduces the ability for users to predict using external assets such as Ether, Bitcoin, or USDC. This concept was originally proposed in Parasitic Immune Augur.
The design can be summarized as follows:
Users lock external assets (e.g., Ether) into trade agreements that rely on oracle resolution. Whenever the system reaches a decision point:
- The external assets and their associated claims are converted into a branchable asset (e.g., REP)
- The decision branches - each possible outcome creating a separate path
- Each resulting branch can continue to fork as new decisions arise, allowing the system to evolve indefinitely
An interesting additional benefit we can get with this design is that we could in theory allow any number of any assets being traded on the system. We just need to be able to convert these claims into
Note
- Bob has 100 YES shares on the market worth $50
- Alice has 100 NO Shares on the market worth $50
- Open interest ($100) is sold in open market to get $100 worth of
$REP$ . Let's assume 1$REP = 1$ $, meaning we get 100$REP$ in the sale. - The market forks and the chain is split into YES and NO universes.
- On the YES universe, Bob can convert their 100 YES shares to 100
$REP_{YES}$ . On No Universe, Bob's shares are worth 0$REP_{NO}$ . - On the No universe, Alice can convert their 100 NO shares to 100
$REP_{NO}$ . On Yes Universe, Alice's shares are worth 0$REP_{YES}$ . - Bob believes that the NO universe is correct even thought they lost, so they continue to trade in Universe NO in the future.
-
Alice also believes the NO universe is correct and they continue to trade there. Alice sells her
$REP_{NO}$ for 100$ and realizes profit of 50$.
It also might happen that the steps 7-8 go differenly in a way that both traders disagree and continue to trade in YES and NO universes separately. This can result in both universes to be valuable and end up being used for future markets.
There's a couple big assumptions being made on top of the colored coin assumptions:
-
Open interest can be sold to the same worth of REP: When open interest gets sold on the open market, we get equivalent worth of
$REP$ tokens from the market. -
Asset swap assumption: People are okay having their open interest of some external assets being converted into
$REP$ .
Open interest assumption splits into couple of smaller assumptions:
-
Zero trading fees: When selling assets on open market to
$REP$ , you often have to pay trading fees, and which results in the assets to be worth less than their original value. - Sufficient liquidity: When selling big amount of assets, there might not be enough liquidity on the market and there's a lot slippage.
-
Open interest < REP's market cap: If
$REP$ 's market cap is less than the open interest being sold, it's not even possible to get equal worth in$REP$ as not enough$REP$ exists for that. -
REP needs to maintain its value: After Open interest has been converted into
$REP$ ,$REP$ 's price might drop, causing users to lose value.
When a fork occurs in Colored Coins, we can simply split the
Assuming the market cap of
Note
When Partial Colored Coins forks, it becomes Colored Coins System
In the event of a fork, the system must be able to liquidate large volumes of assets in exchange for
- In a margin call, traders can intervene (e.g., by adding collateral) to prevent liquidation.
- In our system, liquidation is automatic and unavoidable; open interest holders cannot interfere once triggered.
- Margin calls prioritize speed to avoid further declines in asset value.
- In our case, speed is secondary - the priority is to recover
$REP$ in a cost-efficient manner.
The primary goal is to maximize the value recovered for open interest holders. The only significant time-related pressure comes from the opportunity cost of having open interest unresolved.
A Dutch auction offers a strong solution for protocol-driven liquidation, particularly because it is gas-efficient for blockchain environments.
- The auction begins with a high starting price.
- The price gradually decreases over time at a fixed rate.
- Buyers can purchase at any point while the auction is open.
- The auction ends when all assets are sold.
- All participants pay the same price - the final clearing price when the last asset is sold.
Dutch auction ensure that the assets are sold at a market-driven price, preserving capital for the open interest holders.
Warning
TODO: which is the best auction method to use?
TODO: how long the auction should last?
TODO: how much time we need to alert everyone this auction is going to happen?
In prediction markets in general, users can redeem a complete set of outcome shares (e.g., Yes and No) for $1 of collateral. This allows them to exit early, avoiding liquidation and related fees, offering a cost-efficient path that also reduces the system’s liquidation burden - benefiting all open interest holders.
Both the Colored Coins and Partial Colored Coins systems are inherently secure against the Parasitic Interest Problem (described in Augur V2 whitepaper).
Augur V2 addresses this issue by requiring open interest to be registered within the system. This allows Augur contracts to be aware of the open interest and adjust its fee structure accordingly. However, Augur V2 cannot enforce the registration of open interest; it relies on social layer. The community is expected to discourage the use of external systems that exploit Augur V2 without contributing to its security. In essence, the social layer plays a key role in defending against parasitic interest.
To maintain security, Augur V2 also aims to keep its market cap at least 5x the size of its open interest. This ensures resilience even against up to 4x parasitic interest and some volatility of
In contrast, the Colored Coins solution inherently avoids parasitic interest, as the system does not resolve decisions. Since no resolution occurs, no external system can leech off its outcomes. The Partial Colored Coins approach shares this resistance: open interest is exchanged for
However, both systems can be susceptible to parasitic behavior when they include escalation games designed to prevent forks. For example, one could construct a wrapper oracle like:
Use the Colored Coins system with an escalation game, and if the escalation game fails, ask Bob to resolve the market.
This approach reduces Bob's workload, as he only intervenes when the escalation game fails. The trustworthiness of this setup depends on Bob’s valuation of his reputation being higher than any potential gain from exploiting open interest.
Another example is a Market Cap Oracle:
Use the Colored Coins system with an escalation game, and if it fails, resolve based on the highest traded market cap of the forked colored coin.
This mechanism assumes that traders will gravitate toward the fork representing the truthful universe, even if others attempt to manipulate colored coin prices. In theory, price manipulation wouldn’t matter, as rational traders prefer to trade in a universe they believe to be accurate.
However, this assumption can conflict with the greed assumption - that users may act in their own short-term interest, potentially favoring manipulated or parasitic systems if doing so benefits them financially.
The security of Partial Colored Coins relies on ensuring that open interest remains below the market cap of REP. This same principle applies in Augur V2, where the goal is to maintain this inequality to preserve system integrity.
Augur V2 attempts to enforce this by introducing a fee on open interest holders, which is then redistributed to
However, in Augur V2, this fee is charged only once, regardless of how long the open interest is held. This creates a weakness: users can lock up open interest for extended periods without incurring further cost, reducing the incentive to keep open interest below the
A more robust approach is to implement time-based fees - fees that increase the longer open interest is held. These can be dynamically adjusted to maintain a target relationship between open interest and
- The fees are directly taken from the current open interest and thus they reduce the total open interest amount
- The higher the fees are, the less attractive the platform is for traders
One simple method to implement time-based fee adjustments is via a Bang–bang controller:
Here, increment is a tunable parameter that determines how quickly the fee reacts to changes in open interest. The key is to adjust fast enough to be effective, but not so fast that it causes fee volatility.
The actual Open Interest Fee (%) then evolves over time as:
Where:
Fee Changeis the output from the Bang–bang controllerΔtimeis the time delta since the last adjustment
Alternatively, more sophisticated controllers - such as Proportional–Integral–Derivative (PID) controllers - can be used for smoother and more responsive control.
The Target REP Market Cap must always exceed Open Interest. If
Even when
- Augur V2 uses a target of:
- Partial Colored Coins may require a significantly higher multiplier, as in fork scenarios, where the REP market cap and liquidity must withstand large, sudden purchases.
Warning
TODO: Research on which Open Interest multiplier should be used
If the system holds a large number of arbitrary external assets, it must be able to accurately track the price of each one to assess the total open interest. This can be risky, as unreliable or volatile assets may disproportionately affect the calculated open interest. Such distortions are particularly dangerous if the system adjusts its fees based on open interest targets, as Augur V2 does.
An alternative to directly adjusting Augur V2's fees is a dashboard defense approach: instead of trying to enforce a hard cap, a public dashboard could estimate the total value of stored assets and inform users if the system appears insecure. However, this comes with a critical limitation-users who already have assets in the system may not be able to withdraw once the system is undercollateralized, even if they are warned in advance.
It is also possible that the fee adjusting controller does not adjust the fee fast enough and the system can become undercollateralized under the controller as well. In this case the users of the system should also try to exit the system.
Instead of auctioning the assets, we could give all the assets to
Assuming fork does not have an impact on the market cap of
tokens for open interest holders to replace their funds with
Combining these equations result in number of tokens we need to mint for the assets:
It can be seen from the equation that
In the multiple asset case scenario, the open interest need to be known separately for each asset:
We can then get the minting amount needed:
where
We can then get
We then need to have a price oracle for each
We can then deduce how much
This can be solved in closed form when there's one or two assets, but requires a numerical solution for more assets.
We also need to assume that minting
There's an easy attack you can do to against this system. The system requires that we have a reliable price oracle for each collateral. If we allow users to trade with any asset out there, we cannot guarrantee the asset will have a reliable oracle. Given we have an asset
Pros:
- No big liquidation events to crash the market
- Smaller market liquidity requirements: We only need enough
$REP$ liquidity for each asset to measure their fair price - No trading fees
- The costs are socialized to
$REP$ holders instead to open interest holders
Cons:
- We need to have an access to reliable price oracles
- We cannot support arbitary assets
- During the price measurement, there's two sides trying to manipulate the market:
-
$REP$ holders prefer as little as possible$REP$ to be minted - Open interest holders prefer as much as possible
$REP$ to be minted
-
Warning
TODO: research if we can somehow guarrantee that manipulating price oracle is harder than the assets we use for it
The cost to trigger a fork should be high enough to cover the potential cost of liquidating all open interest, helping to keep the system solvent during a fork. To prevent abuse, triggering a fork should also require burning some assets - ensuring it’s never free and discouraging spam or griefing. If the system only burns REP as a defense, an attacker who controls 100% of both REP and OI could still fork at no real cost, defeating the purpose of the safeguard. Forks affect all users and are irreversible, similar to Ethereum’s state growth problem, so their cost should reflect the broad impact. This means the cost should scale with the number of users involved, not just the total value at stake. Ultimately, the cost to traders should increase with the amount of open interest in the system to align incentives and maintain security.
The system's security relies heavily on the liquidity of
The
What is the upper limit bonus we can give to traders that won't break incentives? As the forks are disruptive, there could be some kind of escalation game that results in fork to trigger, similarly to Augur v2. In Augur v2 At least
If we assume then we have at most
The open interest holders pay total trading costs of
If
A happier outcome here is when
Warning
TODO: We could analyze on the impact on the market and markets liquidity requirements assuming theres x rep, y eth locked in unisvap v2
TODO: We could also try to try to measure the loss traders are making and minting enough rep to cover it
Forks in partial-colored coin systems can be highly disruptive. When a fork occurs, all open interest is liquidated, and users are forced to migrate to a new universe. To minimize this disruption, forks should be treated as a last resort - only occurring when no other resolution is possible. One mechanism to help avoid forks is the escalation game, which enables users to coordinate and resolve disputes collectively.
The primary goals of the escalation game are:
- Long-Term Coordination - Enable sustained collaboration among participants over time of the dispute
- Public Commitment - Allow individuals to visibly signal their belief in the correct outcome
- Anti-Hedging Cost - Impose a cost on participation to prevent users from supporting multiple sides
- Incentive Alignment - Ensure users who choose correctly are rewarded more than they lose by participating
- Collective Funding - Support pooled contributions toward a shared resolution goal
- Decision or Deadlock - Include a path to force a resolution or formally declare a stalemate
- Affordable Stalemate - Make “no decision” affordable but not easily exploitable
- Broad Participation - Maximize inclusion by making participation widely accessible
In Augur V2, a fork is triggered when 2.5% of all theoretical
Forking is a last-resort resolution mechanism. It is highly disruptive and intended to be rare. A fork occurs when an outcome in a market successfully fills a dispute bond equal to at least 2.5% of the total theoretical REP.
If the escalation game concludes without triggering a fork, Augur V2 assumes the final staked outcome is correct and resolves the market accordingly.
The escalation games can result in incorrect outcomes. If too few users are willing or able to participate - especially in researching and voting on complex markets - the game may be dominated by poorly informed or misaligned participants. This is especially problematic in high-stakes markets, where the cost of incorrect resolution is significant.
We must prioritize accurate resolution in markets with substantial open interest. These users are paying customers; if they lose money due to incorrect outcomes, their trust in the system erodes. In such cases, user retention and system credibility are at risk.
An escalation game requires a token that is valuable, liquid, and fork - capable. Augur V2 uses
Using external assets in escalation games brings notable advantages: improved liquidity, greater price stability, having externally valuable asset, and increased participant accessibility. Higher liquidity attracts more users, and stable prices make the potential rewards from escalation more predictable. Being able to access externally valuable asset also allows the system to burn that asset and be more confident that attackers lose value even if they own significant portion of
Moreover, since Partial Colored Coin users may already have exposure to
However, this approach comes with trade - offs. If the escalation game ends with a "no decision" outcome, participants are forced to convert the external asset back into
A further complication lies in determining how much of the external asset is needed to trigger a fork. If the external asset is the same as the one used in the open interest, the fork threshold could be set as a percentage of that open interest. But in the Partial Colored Coins system, some markets use
Escalation games are optimistic systems - they assume that the currently winning outcome is correct unless it is actively disputed. If these escalation games resolve correctly most of the time, users may begin to trust them blindly, leading to a phenomenon known as the normalization of deviance: the system appears to work reliably, so people stop scrutinizing it.
In Augur V2, it was assumed that those with the most at stake - users with open interest in a market - would participate in the early rounds of escalation. Their involvement is expected to draw attention to disputes and encourage those with more
One proposed mitigation to this issue is the use of paid reporters who are required to report on every market. If a paid reporter submits an incorrect resolution, others can challenge it through the escalation game. If the challenge succeeds, the reporter is removed from the system, the market is resolved according to the escalation outcome, and the faulty reporter forfeits their position and associated rewards. The system then selects a replacement reporter.
A similar mechanism exists in reality.eth, where a role called the adjudicator serves a comparable function - ensuring correctness and accountability in market resolutions.
- How much to burn (v2 has 20%)
- How much to award to open interest holder (v2 has 0%)
- How much award to people who played escalation game (v2 has 40%)
Warning
TODO: The cost of fork should at min be the cost of the pain caused to open interest holders. However, we don't know how much this is, and if we pay too much, it introduces an attack vector.
TODO: The cost should be higher than the cost occured to open interest holders, however, its hard to measure the cost introduced to traders
Without prior oracles such as escalation games or designated reporter, you can make a bet, and forget about the bet until you want to bet again or withdraw. Prior oracles break this, as they require you to pay attention to the prior oracle resolutions and be ready to dispute their resolutions.