diff --git a/docs/learn/market-creation-guide.md b/docs/learn/market-creation-guide.md new file mode 100644 index 00000000..81835db1 --- /dev/null +++ b/docs/learn/market-creation-guide.md @@ -0,0 +1,284 @@ +--- +id: market-creation-guide +title: Market Creator's Guide +--- + +# Market Creator's Guide + +This is a short guide to selecting topics for prediction markets and formulating +air-tight, bullet-proof descriptions and resolution criteria. + +## What Makes a Good Topic? + +The quality of the topic of a prediction market plays a significant role in +attracting liquidity providers and informants to the market. Uninteresting, +unclear or opinionated topics will deter users from participating. Well-phrased +markets on curious topics will flourish. + +When selecting a topic for a market, the creator must avoid trivial or +exceedingly difficult questions or illegal topics. For example, "Will SpaceX +send a manned mission to Mars by 2026?" is likely trivial. Topics may be +presented in in the form of a question or a statement. Contentious issues are +welcome, as long as resolution criteria are clear (see below). Questions are +preferred for contentious issues. + +The market's duration and time frame has implications on engagement. If the +market runs for a long time, users are required to take into account economic +uncertainties like the price stability of the base asset when making a trade or +providing liquidity. They are also taking on risks of information shock. If the +market runs for too short a time, accurate predictions are difficult to +make.[^1] The ideal length depends on the topic: For sports markets, shorter +durations are acceptable. Markets on political or economic trends should at +least run a couple of month to properly aggregate information. + +[^1] Zeitgeist currently doesn't allow markets with a duration of more than a +year. + +For example, "Will SpaceX send a manned mission to Mars by 2030?" requires +participants to lock up funds for a significant time period. If the base asset +against which the bets are taken is a speculative asset itself, this bears +incalculable risks for those who hold outcome tokens. + +If a market runs too long, it might be a good idea to split the market up into +smaller steps. For example, a market on a manned Mars mission may be split into +markets betting on the success or failure of Starship test launches or unmanned +missions. + +## What Makes a Good Description? + +The landmark features of a good description are clear language and unambiguous +_resolution criteria_, which real-life developments map to which prediction +market outcomes. For every possible real-life development, the resolution +criteria must specify an outcome token which represents that outcome. + +There is one exception to this rule. There are always so-called _long shots_, +which are so unlikely that it makes no sense to bet on them makes no sense, but +which may occur after all. For example, a market on the Champions League Finals +might have one outcome for each team. But there's also the chance that the game +is delayed until after the market is resolved or even cancelled. Any event with +a probability of less than $1\%$ is considered a long shot and should generally +not be made part of the market. If a long shot occurs, the market is instead +voided.[^2] + +[^2] As of 0.3.9, Zeitgeist doesn't yet implement voiding markets. + +The resolution criteria may also specify which source the oracle uses to resolve +the market. This isn't always necessary, but it is of paramount importance if +the topic of the market is contentious and may result in conflicting reports. +Bad or unreliable sources must be avoided. The data used to resolve the market +should be easily reproducible for all participants in the market. + +When creating a scalar market instead of a categorical market, the market +creator must define a lower and upper bound for the values. A badly configured +range doesn't hinder the resolution of the market (recall that if the reported +value falls outside of the market's range, SHORT resolves to 1 or 0 depending on +where the value ended up, and a similar rule applies for LONG), but has +implications for the informants and the quality of the predictions: If the range +is too wide, there's too little price resistance to create a good prediction or +sufficient reward for informants; if the range is too narrow, small fluctuations +or information shocks can ruin the liquidity providers. A good balance must be +struck here. + +## Advisory Committee's Standards + +The list below contains some guidelines that Zeitgeist's Advisory Committee +prefers to use: + +- Please spell-check your description and title. In the age of AI, there's + little excuse for presenting incomplete sentences, gross spelling mistakes or + missing punctuation. In the worst case, these can result in distortions of the + market creator's intent. +- Let AI review your market. This can help identify missing outcomes, loopholes + or lack of clarity. +- Specify contentious issues as questions rather than statements. +- Specify a time zone for each date that is relevant to the resolution of the + market. Prefer UTC unless the topic is clearly related to a particular locale. + Dates without time zones can lead to confusion. ISO 6801 + (YYYY-MM-DDTHH:MM±HH:MM), on the other hand, is bullet-proof. +- Prefer 24 hour clock to 12 hour clock. +- Define an outcome token for every possible outcome except long-shots. +- Resolution data must be reproducible and verifiable as possible. Avoid sources + which are not available for the entire duration of the market. Avoid + newspapers whose articles are paywalled and API calls which require a paid + key. +- Prefer API calls to manually pulling data. +- Prefer round values for dates and other data (April 1, 00:00 preferred to + March 31, 23:59). + +## Examples + +### Will the US Raise the Debt Limit by $1.5 Trillion or More by June 2, 2023, 19:00 (UTC)? + +> One of the most pressing current events is the potential default of the US +> government on its payments. This could carry significant implications for +> global economies worldwide. US bonds are viewed as the gold standard of +> safe-haven assets, and any failure to make payments could trigger major +> macroeconomic events. Will the US increase the debt limit by $1.5 trillion or +> more by June 2, thus ensuring the security of bond payments? +> +> The outcome will be reported based on major US news sources following +> government announcements. The market will resolve to YES if there is an +> announcement that the debt ceiling is increased by $1.5 trillion or more by +> June 2, 2023, 19:00 (UTC). If there is no such announcement until June 2, +> 2023, 19:00 (UTC), or there is an announcement mentioning a lesser increase +> than $1.5 trillion or other measures taken to avert default, then the market +> will resolve to NO. + +Although well phrased, this market might be considered to be ambiguous. The U.S. +avoided default by +[suspending the debt limit](https://home.treasury.gov/news/press-releases/jy1517). +This allows essentially infinite spending. But does a suspension count as a +raise of the debt limit? + +A solution would be to view this a case of the +[XY Problem](https://xyproblem.info). Does the market creator ask _how_ the +government default is avoided - or _if_? If the _how_ matters, the market +description should clarify how suspensions of the debt limit are handled. +Otherwise, rewriting the question to something like "Will the U.S. government go +into default by June 2, 2023, 19:00 (UTC)?" might be a good idea. + +### Major AI-Related Incident in 2023 + +> Within the next 12 months, will there be a verified news report of an +> AI-related incident causing at least $100 million in damages or impacting the +> personal data of over 1 million individuals? +> +> 1. The incident must be directly related to an AI system, such as an AI-driven +> recommendation engine, autonomous vehicle, or natural language processing +> system. +> 2. A verified news report from a reputable source (e.g., BBC, CNN, or The New +> York Times) must cover the incident, detailing the AI's involvement and the +> extent of the damage or impact on personal data. +> 3. The report must specify that the damages amount to at least $10 million +> (USD) or that the incident has compromised the personal data of over 1 +> million individuals. +> 4. The incident must occur within the 12-month window from the market's start +> date. + +Resolving this market is difficult. Scanning all media for reports on such +incidents is time consuming. Even worse, it is virtually impossible to disprove +that such an incident has occurred. + +It is also unclear what constitutes damage. For example, on May 22, 2023 an AI +generated image of an explosion near the Pentagon was distributed on the +internet, +[causing a $500 billion swing of the S&P 500](https://twitter.com/KobeissiLetter/status/1660664125574217731). + +### NHL Finale + +> As we approach the highly-anticipated Stanley Cup final, set to kick off on +> Saturday, June 3 at 8 p.m. ET (Sunday, June 4 at 00:00 UTC) in Las Vegas, a +> pivotal question emerges: Who will seize the advantage and triumph in the +> first game of this series - the Golden Knights or the Florida Panthers? Both +> teams have demonstrated determination and exceptional performance throughout +> the season and playoffs, now coming face-to-face in the most prestigious +> hockey championship. This prediction market endeavors to forecast the outcome +> of this vital opening game, a significant milestone in the quest towards the +> revered Stanley Cup. +> +> The resolution criteria for this prediction market will be determined by the +> victor of the first game of the Stanley Cup final, be it the Golden Knights or +> the Florida Panthers. The market will resolve in favor of the Golden Knights +> if they secure the initial victory, and likewise, it will resolve in favor of +> the Florida Panthers if they prevail. The outcome will be drawn strictly from +> the game's result after regulation time and any necessary overtime or +> shootout, excluding any external factors such as disqualifications, +> suspensions, or unforeseen events. The official NHL game report will serve as +> the primary source of validation for the outcome. +> +> This analysis and the formation of this prediction market are grounded in the +> insights provided by Adam Proteau in The Hockey News (Proteau, 2023). +> +> Reference: Proteau, A. (2023). NHL Stanley Cup Predictions: Do The Panthers Or +> Golden Knights Win It All? The Hockey News. Retrieved from: +> **https://thehockeynews.com/news/nhl-stanley-cup-predictions-do-the-panthers-or-golden-knights-win-it-all** + +This market is well phrased, correctly cites its sources, contextualizes the +market using an introductory paragraph and even specifies the official game +report as official source for the oracle. One way this description could be +improved is by detailing at what date the market will resolve ambiguously if the +game is cancelled or delayed. + +### ASTR/USD at the end of Q1 2023 (UTC) + +> On February 17, 2023, Sony Network Communications announced that it will be +> co-hosting a web3 incubation program with Astar Network on NFTs and DAOs, +> sending the spot price of ASTR/USD soaring over 10c (see +> [Bitcoin News](https://news.bitcoin.com/sony-and-astar-network-launch-web3-incubation-program-for-nft-and-dao-focused-projects/)). +> +> What will the spot price of ASTR/USD be at the end of Q1 2023 (UTC) according +> to CoinGecko? The exact API call that will be used is: +> https://api.coingecko.com/api/v3/coins/astar/history?date=01-04-2024-00-00-00. + +Note that this description specifies a clear API call which is used to resolve +the market. + +### Will @realDonaldTrump tweet before the end of April 2023 (UTC)? + +> Former US President Donald Trump was banned from Twitter on Jan. 8, 2021 +> following tweets about the Insurgence of Jan. 6, 2021 (see blog.twitter.com) +> and reinstated on Twitter in November 2022 after a popular vote conducted by +> Elon Musk. NBC News suggests that Donald Trump prepares for his return to +> Facebook and Twitter. +> +> Will @realDonaldTrump tweet before the end of April 2023 (UTC)? The market +> will resolve to YES if the account sends any tweet. The market will resolve to +> NO otherwise. In particular, the market will resolve to NO even if +> @realDonaldTrump retweets or likes a tweet or sends a direct message without +> tweeting. It is irrelevant whether or not the tweet is later deleted. + +While this market is well-phrased, the result is not easy to verify due to the +clause: "It is irrelevant whether or not the tweet is later deleted." + +### Will a Nuclear Bomb be Detonated in 2023? + +> This prediction market aims to predict whether a nuclear bomb will be +> detonated in the year 2023, as per Coordinated Universal Time (UTC). A nuclear +> bomb is defined as a weapon that uses nuclear reactions of fission or fusion +> to release large amounts of energy. This can encompass a range of nuclear +> weapon designs such as fission bombs, boosted fission bombs, thermonuclear +> weapons, and neutron bombs. +> +> In addition, any nuclear tests conducted by a recognized government body or +> other credible entity, including clandestine or non-state actors, will be +> taken into account, provided these detonations meet the nuclear definition +> above. +> +> For this market, a detonation is considered to have occurred if there is +> physical evidence of a nuclear explosion such as seismic readings, radiation +> detection, or satellite imagery confirming an event. Official statements from +> governments or credible international organizations confirming such a +> detonation will also be accepted as evidence. +> +> Exclusions: +> +> - Failed detonations: Any test or attack that fails to result in a nuclear +> explosion as defined above will not count as a detonation. +> - Dirty bombs: This market does not count detonations of "dirty bombs" or +> radiological dispersion devices, where conventional explosives are used to +> disperse radioactive material. +> - Nuclear Power Plant Accidents: Accidents or malfunctions at nuclear power +> plants or other nuclear facilities, such as meltdowns, do not count, as +> these do not involve the intentional detonation of a nuclear bomb. +> - Nuclear Propulsion Detonations: Any explosion involving a device using +> nuclear propulsion, such as those used for space travel, will not be counted +> as these are not intended as weapons or tests of weapon technology. +> - Subcritical Tests: Tests involving nuclear materials that do not reach a +> critical mass and thus do not result in a nuclear explosion, known as +> subcritical tests, will not be considered for this market. +> - Decommissioning Explosions: Explosions used in the decommissioning process +> of nuclear weapons or facilities, which may involve nuclear materials but +> are not themselves nuclear detonations, should be excluded. +> +> Resolution details: +> +> The market will resolve to "Yes" if there is evidence of at least one nuclear +> bomb detonation occurring in the year 2023 (UTC). The market will resolve to +> "No" if no such evidence is provided by the end of the year 2023 (UTC). +> Verification will be based on publicly available evidence and credible news +> reports or official statements. + +This market is indeed very detailed and tries to ensure that all conceivable +corner cases are specified. The language is very clear and dates are equipped +with time zones, avoiding any confusing. Note, however, that this market +description is awkwardly long. diff --git a/sidebars.js b/sidebars.js index 487bb475..1fc85687 100644 --- a/sidebars.js +++ b/sidebars.js @@ -22,10 +22,11 @@ const sidebars = { items: [ "learn/prediction-markets", "learn/liquidity", - "learn/market-rules", "learn/using-zeitgeist-markets", "learn/governance", "learn/court", + "learn/market-rules", + "learn/market-creation-guide", "learn/futarchy", "learn/comparisons", {