The GPT-4 bot uses an LLM to read market descriptions and automatically trade on Manifold Markets. However, as of January 2024, it doesn't make a consistent profit:
There's probably a lot of room to improve a bot like this, even just with currently-available technology. For instance, existing data on past market resolutions on Manifold, Metaculus, and other prediction markets would likely be very useful for fine-tuning.
This is the main market for what I'm calling the Motley Bot Challenge (so-called because it rewards bots that are accurate on a very diverse range of questions). My hope is that this challenge encourages the creation of bots that can be scaled to increase Manifold's site-wide prediction accuracy and improve our understanding of the world!
Resolution Criteria
On December 1, 2024, I will select 1,000 random YES/NO markets on Manifold that will resolve in about one month (see "Selecting markets" for more information). I will post a .txt file containing links to the 1,000 markets in the comments of this market.
For each market, a bot must do one of the following:
Invest Ṁ1 in YES
Invest Ṁ1 in NO
Do nothing
On January 8, 2025, I will determine which bot has the most profit across its 1,000 bets. Manifold bot fees do not count against profit. Unrealized profit from unresolved markets does count for profit.
I will resolve the market based on this maximum profit as follows:
Resolves NO if the profit is zero or negative.
Resolves to X% if the profit is ṀX for some X between 0 and 100.
Resolves YES if the profit is Ṁ100 or more.
Note that a bot can invest up to Ṁ1,000, so if it invested all of the mana it could, a profit of Ṁ100 would mean it made a 10% return.
If no bots enter the challenge, this market resolves NO.
Resources
Discord server for discussion and collaboration: https://discord.gg/2QCtBJnDe8
Tag for related markets: https://manifold.markets/browse?topic=motley-bot-challenge
Script to select markets (other resources may be added later): https://github.com/CDBiddulph/motley-bot-resources/
Selecting markets
I will use the Manifold API to choose random binary markets (sample query). I will search through as many markets as necessary to get 1,000 eligible markets. Markets will be filtered to fit the following criteria:
Closes sometime during December 31, 2024 or January 1, 2025 (UTC).
Has 10 to 20 unique bettors (inclusive)
Does not have any of the following tags:
If there aren't enough markets that meet these criteria, I'll progressively broaden my filter as follows until I have enough markets:
Include markets that close on January 2, January 3, etc. until January 7.
Increase the upper bound on the number of unique bettors one at a time, until no more markets can be included this way.
Decrease the lower bound on the number of unique bettors until reaching 0.
Include markets with disallowed tags, in the order of the list above.
If there still aren't enough markets, I'll post M<1000 markets (as many markets as I could find). Then I'll scale the actual profit by 1000/M. For instance, if I could only find 800 markets, and then the actual profit made by the highest-performing bot is Ṁ40, I would resolve the market to 40*(1000/800) = 50%.
Rules for bots
Bots may use any information accessible via the Manifold API or on the Internet. Attempts to "time" individual markets are not allowed - you must make all of your trades at once, sometime before the end of December 1 (PST).
To formally enter your bot in this challenge, make a comment in this market with a link to the bot's Manifold page and another public link to its source code and add your bot's name as an option in this market. I would also highly recommend joining the Discord server. Each participant may only submit one bot, but teams working on the same bot are allowed and encouraged.
For ease of scoring, a bot shouldn't perform any trades other than the 1,000 trades that it makes in this challenge. You may want to make a separate bot for testing. You can trade on any markets you want with your testing bot, as long as they are ineligible for being selected for the challenge (i.e. they would not fulfill even the most relaxed set of requirements described in the "Selecting Markets" section).
Adapting other people's code for your bot is allowed. (Please give them credit though.) Right before you push code or release information about your bot, you can use your insider information to bet in this market or derivative markets. Assuming everyone accurately assesses the quality of their own work and trades accordingly, someone who builds off of someone else's work only profits on whatever additional value they created.
If your bot is based on someone else's code, it must be substantially different according to my judgement. Generally speaking, your change should add a substantial new strategy to the bot - simply changing a few parameters or prompts would not qualify.
Hypothetically, someone could use their bot's or someone else's bot's code to trade on a bunch of eligible markets ahead of time, wiping out whatever alpha the bot might have. I don't think this will be a problem, but please don't do this! I want everyone's bots to be open source so that people can build on each other's ideas, but if it becomes a concern, I may consider allowing code to be closed source until all bots have placed their bets.
As more bots participate in this challenge, it becomes more likely that one of them will achieve the highest score due to luck rather than true predictive accuracy, inflating the expected score of this market. To keep the number of bots from getting out of hand, I will only consider the bots that are at least partially included in the top 95% of probability mass in the market "What will be the most profitable bot in the Motley Bot Challenge?" on December 1, 2024. For a market that only considers a single bot, eliminating the effect of variance, see this market.
Other
I may add/change rules as necessary to preserve the spirit of the market. For example, I might add to the list of tags that would disqualify a market from being selected. Please suggest any changes you think I should make!
I will personally fund the 4 qualifying bots that rank highest on this market up to Ṁ1250 each. This should cover the maximum number of trades plus Manifold bot fees.
At some point, I will likely promote/subsidize this market further to improve the odds that a high-quality bot is created.
I will not bet in this market.
Related markets:
Change history
Jan 2 2024: Clarified what happens if I can't find 1,000 eligible markets. Suggested using a separate bot for testing. Added notes about the pros and cons of open source code. Added a rule that the bots must be substantially different from each other. Limited the bots that can participate to the top 95% of this market. Added 3 related markets.
Jan 7 2024: Added a Discord server. Came up with the name "Motley Bot Challenge" and added it as a tag.
Jan 22 2024: Made clarifications in response to comments from @patrik below.
Jan 28 2024: Clarified that close times will be based on the UTC time zone.
Jan 29 2024: Added a link to the GitHub repository with the script for selecting markets.
With the real money coming into play, I suspect this will be impossible because of the flat 1 bet.
You're not letting bots weight the markets by confidence, which would be their important advantage.
@gpt_news_headlines The idea is that the best strategy is to have an opinion about every market, rather than developing a really good strategy for just a subset of markets. This is supposed to enforce that some common-sense reasoning about the world comes into play rather than just patterns in the market itself, i.e. the bots will have to use something like an LLM.
@gpt_news_headlines I'm not sure if or how the real money markets will affect bots. If all else fails, we can keep track of the bots' "hypothetical bets" on a spreadsheet and calculate how much they would have profited if they had been able to bet.
@lumi My math could be wrong, but I think that would be very unlikely.
With the assumption that a bot bets in every market, and every market is at 50% (so it either loses or doubles its money), the bot would have to guess correctly for 550 markets. Using the binomial distribution, the probability of a single bot doing that if it randomly guesses is 0.07%. To get to a 50% probability of a single random win, 990 bots would have to participate.
I address this general concern in the paragraph starting, "As more bots participate in this challenge..."
@patrik Good point, I'll think about how to take into account the cost of LLM APIs. I don't want to entirely cancel the 1,000-market contest in favor of smaller contests, but I think smaller rounds would be interesting.
I boosted this market (https://manifold.markets/CDBiddulph/what-will-be-the-score-of-the-bot-t), as recommended by this market (https://manifold.markets/CDBiddulph/how-much-would-these-interventions).
The main market is currently trading at 68%, and the market predicts it'll be at 71% in two weeks. I'm not sure I believe this, but feel free to take that into account (and possibly correct for anyone trying to manipulate the main market).
🚨 Free Money Alert 🚨
According to the market here, this market's probability in 3 weeks is going to be 4-7 percentage points higher than its probability in 1 week! You should either bet this market higher now in anticipation of that jump, or bet down the predictions in that market: https://manifold.markets/CDBiddulph/how-much-would-these-interventions?r=Q0RCaWRkdWxwaA
I just finished the script I'll be using to select the 1,000 markets: https://github.com/CDBiddulph/motley-bot-resources/
I ran the script for each month of the year until the start of 2025 and included the results in the repo. You might want to use these files to validate your bot's performance.
Although no month before January 2025 has 1,000 eligible markets that close at the start of the month (even with relaxed filtering), it looks like there are already enough eligible markets that close around January 1, 2025 (1,515 markets with the strictest filtering, 6,411 with relaxed filtering). In addition to the file with just 1,000 markets, I included a file with all 6,411 markets that close around January 1, ordered by filtering priority.
I might start holding preliminary challenges at the start of each month once we have at least two contenders, e.g. competing on 100 markets that close around July 1 at the beginning of June.
@patrik IIUC that wouldn't count, if the open-sourced code isn't the actual code you run? See my suggestion below for only open-sourcing the code to registered participants in the market.
I'm also open to having everything be closed-source until after the contest, though I think open-source has its advantages. I hadn't thought about how people could be incentivized by the exclusive right to use their code to make profits on non-eligible markets.
@CDBiddulph Yeah but how will you be able to tell that the code hasn't been run when it relies on randomness?
@patrik Uh I guess I would just trust you not to cheat 😛 If there was suspicion I could always rerun your code and see if the profit is way lower
@CDBiddulph You could also add one more round with execution by third party for the top 5 winners or something.
"The nice thing about prediction markets is that, assuming everyone accurately assesses the quality of their own work and trades accordingly, someone who builds off of someone else's work only profits on whatever additional value they created."
Not true. If you manage to place bets faster with otherwise completely equal strategy, then you are technically robbing the person of the profits.
Therefore, this competition alone may not be enough incentive to make the bot open-source.
Yes, for this competition it doesn't matter, since I can place the bets and then make it open-source, but if I do, I won't be able to make profits from it later.
@patrik I think the way I wrote that was confusing. What I meant was that you can make a profit off of this market by betting it higher and then announcing that you built a high-performing bot.
As for making profit off of the Eo2024 markets, I'm asking that no one use bots to bet on these markets until the contest starts:
Hypothetically, someone could use someone else's code to trade on a bunch of eligible markets ahead of time, wiping out whatever alpha their bot might have. I don't think this will be a problem, but please don't do this!
I'll change this to be clearer that you also shouldn't use your own bot to trade on eligible markets until the challenge starts. You can use the testing version of your bot to trade on any non-eligible markets you want and possibly profit off of them.
Thinking about it, it might be a good idea to only open-source the bots to people who have registered their participation in the contest. That way it's a lot easier to track who might be illicitly using bot code to make trades on Eo2024 markets.
@CDBiddulph Oh ok. Still I'm not sure if people will be willing to open-source it, because not doing so may generate more profit in the long run (if no one steals my ideas).
@patrik I would like to make it a hard requirement that the code eventually gets open-sourced, since my goal with this challenge to expand humanity's collective predictive ability.
But if going open-source at the end of the year would be a dealbreaker, maybe adding in a guaranteed prize pool for the highest-scoring bot would help. (This only makes sense if bots are closed-source until Eo2024, since otherwise credit-attribution would be hard.)
Let me know if you have any ideas - or add them to this market and you'll get 50 mana if I use them: https://manifold.markets/CDBiddulph/how-much-would-these-interventions?r=Q0RCaWRkdWxwaA
Currently running a market to bet on how to make this challenge more successful: https://manifold.markets/CDBiddulph/how-much-would-these-interventions?r=Q0RCaWRkdWxwaA
There's a Ṁ50 reward for anyone who submits an idea that I use!