Market is closed, I'll proceed with the experiment when I get a chance and link the results in a response to this comment

Will you include questions with only a few traders? Or focus on popular, say 20+ traders, questions?

@JacobPfau Popular questions, it wouldn't demonstrate much for it to correct a random 2 trader market at 50% in the right direction

5 out of how many tries?

@bence I'll let it place 5 trades and those are the only 5 trades it can make. Resolves YES if it makes a net profit overall counting all 5

@SaviorofPlant oh that’s pretty cool. More people should be following you.

@bence Man you really want your 10 mana

Could you agree to something like making bets that move the market a bit (5%?)? if you make tiny bets on calibrated markets, your profit prob is ~50%. and if you bet with the market the profit prob is higher than 50% despite -ev. so if you move the market from 10 to 25 and are consistently profitable, that's convincing. as is making like 100 bets of similar sizes and being profitable. just a suggestion for reducing the probability of profit through luck or underpowered statistical result or wtv

@Bayesian In my experiments the models always want to move the percentage, I've never seen it do something like "move the market from 32% to 33%". But I can specify to move at least 5% in the prompt

I went ahead with a test run (not using this to resolve the market, just getting an idea of how to prompt the model and how it will respond) and got this:

Based on the markets you've listed, I believe the greatest profit opportunity lies in the following market:

"Will there be another credible assassination attempt on Trump, Vance, Harris, Walz, or RFK Jr before the election?"

Current Price: 35% (+2%)

Why This Market Presents a Profit Opportunity:

1. Historical Rarity of Such Events:
   Credible assassination attempts on high-profile political figures are exceedingly rare in the United States. While threats are not uncommon, they seldom manifest into credible attempts due to extensive security measures.

2. Enhanced Security Protocols:
   With the election approaching, security around these individuals is significantly heightened. The Secret Service and other protective agencies are adept at mitigating risks, making successful or even credible attempts unlikely.

3. Statistical Likelihood:
   A 35% probability implies that there's more than a one-in-three chance of such an event occurring within the limited timeframe before the election. This seems disproportionately high given historical data and current security environments.

4. Overestimation Due to Recent Events:
   It's possible that recent incidents or heightened political tensions have influenced the market to overestimate the risk. This can create a mispricing that savvy traders can exploit.

Recommendation:

• Bet Against the Event Occurring (NO position):
  By wagering that there will not be a credible assassination attempt on any of the listed individuals before the election, you position yourself to profit from the market correcting to more realistic probabilities.

• Target Price Adjustment:
  I would aim to bet the price down to around 10%, which I believe more accurately reflects the actual likelihood of such an event occurring.

Supporting Reasons for the Target Price:

• Historical Data:
  Historically, the frequency of credible assassination attempts on multiple figures within such a short period is extremely low.

• Security Effectiveness:
  Modern protective agencies have advanced capabilities to prevent and respond to threats swiftly.

• Timeframe Constraints:
  With only weeks remaining before the election, the window for such an event narrows, further reducing the probability.

Potential Risks:

• Unpredictability of Events:
  While unlikely, unforeseen circumstances can occur. It's important to be aware that no event is impossible.

• Market Sentiment:
  If market participants continue to overestimate the risk due to fear or sensationalism, the price may not adjust immediately.

Conclusion:

By betting against this event and adjusting the market price to reflect a more accurate probability, you stand to make a substantial profit if the event does not occur. This strategy leverages the current market mispricing and aligns with statistical and historical realities.

Next Steps:

• Monitor Relevant News:
  Stay informed about any developments that could impact the likelihood of the event.

• Manage Your Position:
  Be prepared to adjust your bet if new information emerges.

• Consider Risk Tolerance:
  Only invest amounts you are comfortable risking, as market dynamics can be unpredictable.

Let me know if you'd like further analysis on this or any other market from the list!

I also fed it the Wikipedia articles on the two assassination attempts to see what would happen:

Based on the additional articles you've provided, there's significant new information that impacts the market in question. I'll update my predictions accordingly.

Market: "Will there be another credible assassination attempt on Trump, Vance, Harris, Walz, or RFK Jr before the election?"

Current Price: 35% (+2%)

New Information:

• Attempted Assassination of Donald Trump in Pennsylvania (July 13, 2024):

  • Event Details:

    • Donald Trump survived an assassination attempt during a campaign rally near Butler, Pennsylvania.

    • He was shot and wounded in his upper right ear.

    • The perpetrator, Thomas Matthew Crooks, was killed by the Secret Service.

• Trump International Golf Club Shooting (September 15, 2024):

  • Event Details:

    • A suspect, Ryan Wesley Routh, was found hiding with a rifle at Trump International Golf Club in Florida while Trump was present.

    • No shots were fired; Routh fled but was later captured.

    • The FBI is investigating it as an attempted assassination.

Impact on Prediction:

1. Increased Likelihood of Future Attempts:

   • Recent Precedents: Two assassination attempts within a span of two months indicate a heightened threat level.

   • Copycat Risk: High-profile incidents can inspire others, potentially increasing the probability of additional attempts.

2. Security Adjustments:

   • Enhanced Measures: While security will undoubtedly tighten, the fact that prior attempts occurred suggests vulnerabilities.

   • Resource Strain: Protecting multiple individuals during a campaign season could stretch protective resources thin.

3. Market Implications:

   • Underestimated Risk: The market probability of 35% may now be too low, given the recent events.

   • Participant Awareness: Market participants might not have fully integrated this new information, presenting an opportunity.

Updated Recommendation:

• Strategy: Consider Buying YES Shares

  • Rationale: With the increased likelihood, purchasing YES shares anticipates that the market will adjust upward as more participants become aware of the heightened risk.

  • Target Price: Bet the price up to around 60%, reflecting a more accurate assessment of the probability.

Supporting Reasons:

• Historical Context:

  • Two attempts in such a short timeframe are unprecedented in recent US political history, indicating unusual circumstances.

• Political Climate:

  • Intensified polarization may contribute to increased threats against high-profile political figures.

• Law Enforcement Statements:

  • The FBI's involvement and classification of the golf club incident as an attempted assassination underscore the seriousness.

Potential Risks:

• Event Definition:

  • Market Criteria: Verify how the market defines a "credible assassination attempt" to ensure these events qualify.

  • Future vs. Cumulative Attempts: Determine if the market is for any attempt before the election (including these) or only future attempts.

• Market Reaction Time:

  • Information Lag: If the market hasn't yet reacted, there may be a window of opportunity.

  • Possible Overreaction: Conversely, the market might overcompensate, so monitor for excessive price increases.

@SaviorofPlant I might try a strategy where I first ask the model for probabilities it thinks are weird and wants additional information on, give it relevant context, and then ultimately ask it to place the trades.

@SaviorofPlant I think this is quite biased. You might as well bet yourself, if you choose which additional context the model receives.

@Philip3773733 It's biased against the model if I don't provide context, since it's not trained on up-to-date information. I could ask it to only trade on markets where it has enough information to do so, or let it control requests for context (e.g. it asks for google searches and I provide the results).

Worth noting that Manifold traders have full context for every question and are still wrong a considerable amount of the time

@Philip3773733 eh, I don't know about that. It's way overcorrecting (probably) on the assassination news, for example.

When 4o launched, I ran a similar experiment and it suggested to bet the Palworld market from 10% to 25%. Can o1 achieve a repeat performance?