I will use this data set:
Gold Aug 26 (GC=F) Stock Price, News, Quote & History - Yahoo Finance
I have asked three LLM models to calculate the 50/50 maximum gold price estimate based on historic data. I asked for a statistical method and a python script. I could not immediately verify the methods and code, because this is a learning exercise for myself. But these are the three outputs I got:
(DeepSeek) >>>
Fetching data for GC=F from 2026-01-02 to 2026-06-30...
Successfully calculated threshold using 121 trading days of data.
========================================
RESULT: The calculated price threshold (X) is: $4722.30
========================================
Based on the last 6 months of historical data:
If the daily maximum gold price is above $4722.30 on any day in July 2026, the bet 'YES' would be correct.
If it stays below $4722.30 for the entire month, the bet 'NO' would be correct.
This threshold represents the 50th percentile (median) of daily highs in the historical sample.
(ChatGPT) >>>
Estimated trading/session days in target month: 23
Downloaded rows: 138
First row: 2025-12-08
Last row: 2026-06-26
As-of Close used: 4078.70 on 2026-06-26
Historical windows used: 102
--- Main result ---
Median historical max multiplier: 1.04783
Estimated 50/50 threshold X: 4273.79
25th percentile threshold: 4143.17
50th percentile threshold: 4273.79
75th percentile threshold: 4510.02
90th percentile threshold: 4965.57
(Claude) >>>
Ticker : GC=F
Last data date : 2026-06-26
Start price S0 (close) : 4,078.70
Calibration window : 123 trading days (~180 cal. days)
Daily vol sigma_hat : 2.1661% (annualised ~34.4%)
Daily drift mu_hat : -0.0477% (annualised ~-12.0%)
July trading days T : 23
Analytic driftless ref : X ~ 4,375
drift mode GBM X (median) Bootstrap X (median)
------------------------------------------------------
historical 4,419 4,352
zero 4,447 4,375
martingale 4,434 4,368
RECOMMENDED X (bootstrap, martingale): 4,368
implied P(July max > X) = 50.0% (target 50%)
July-max 25/50/75 pctile = 4,192 / 4,368 / 4,576
Read the spread of X across drift modes (section 8) as the drift risk
you are taking on. If the Manifold market's implied X sits inside that
spread, your 'edge' is really a bet on gold's drift, not its volatility.
Update 2026-07-31 (PST) (AI summary of creator comment): - Due to an issue with the market setup, the creator proposes resolving the ChatGPT: $4,273 option to 52% YES because its prediction was closest to the truth.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ54 | |
| 2 | Ṁ21 | |
| 3 | Ṁ19 | |
| 4 | Ṁ14 | |
| 5 | Ṁ9 |
People are also trading
Cannot resolve NO it this setup. And N/A does not feel right. My proposal is to resolve 52% ChatGPT: $ 4,273 as YES because its prediction was closest to the truth.
As stated before this was an experiment that is not fully satisfactory.
These ones are better but I'm still not satisfied. Suggestions welcome!
@creator did you mean to make these linked markets? They could all resolve no to my view.
@MRME Yes, this is a slightly failed experiment. I'm switching to placing predictions on the statistics of the previous month.
And some of the algorithms the AIs used were stupid.
@MRME Cannot resolve NO it this setup. And N/A does not feel right. My proposal is to resolve 52% ChatGPT: $ 4,273 as YES because its prediction was closest to the truth.
A more logical, and maybe more fun set of bets for July 2026:
/uair01/gold-price-in-july-2026
/uair01/silver-price-in-june-2026
I'm still experimenting and trying to improve it. Sorry for that.