What We Learned
WHAT THE SYSTEM CAN LEARN
Which sources and repeated patterns are actually useful?
Evidence scores measure how complete the saved research is. They do not claim that a source or research area predicts markets.
How complete each publisher’s saved records are
| Publisher | Stories | Saved checks complete | Exact times | Repeated-item share | Past prediction accuracy |
|---|---|---|---|---|---|
| Reuters | 14 | 100 / 100 | 100% | 0% | Not enough evidence |
| CNA | 2 | 100 / 100 | 100% | 0% | Not enough evidence |
| MINDEF Singapore | 1 | 100 / 100 | 0% | 0% | Not enough evidence |
| World Health Organization | 1 | 100 / 100 | 0% | 0% | Not enough evidence |
A complete saved record is not a truth or forecasting score. It only says whether the required information was preserved.
Open every research area’s saved checks
| Research area | Stories | Saved checks complete | Opposing evidence saved | Exact times | Repeated stories | Helped a prediction |
|---|---|---|---|---|---|---|
| AI Opportunities (SG) | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Biotech & Pharma | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Circular Economy / Waste Management & Recycling | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Cybersecurity & Digital Trust | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Defence & Security | 1 | 100 / 100 | 100.00% | 0.00% | 0.00% | Not enough evidence |
| Energy & Power | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Extreme Weather, Meteorology & Natural Disasters | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Geopolitics & International Relations / Conflict Analysis | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Law & Regulation | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Manufacturing & Automation | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Medical breakthroughs | 1 | 100 / 100 | 100.00% | 0.00% | 0.00% | Not enough evidence |
| Public Policy & Government | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Semiconductors & Hardware | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Singapore News | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Sports Signals | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Venture & Startups | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Water & Utilities | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
| Wellness & Fitness | 1 | 100 / 100 | 100.00% | 100.00% | 0.00% | Not enough evidence |
Results by prediction type
Only individual saved results with the required proof count here. This breakdown is separate from the official overall score.
| Prediction type | Saved records | Fairly checkable | Right | Wrong | Accuracy |
|---|---|---|---|---|---|
| Up · 20 hours | 102 | 2 | 2 | 0 | 100.00% |
| Down · 20 hours | 51 | 5 | 1 | 4 | 20.00% |
| Up · 96 hours | 8 | 0 | 0 | 0 | Not enough evidence |
| Down · 96 hours | 7 | 0 | 0 | 0 | Not enough evidence |
| Up · Check time not saved | 4 | 0 | 0 | 0 | Not enough evidence |
Repeated patterns with the most saved events
| Pattern | Events | Complete 20-hour checks | Ready for comparison | Most common saved move |
|---|---|---|---|---|
| Gold futures — economy or policy news | 8 | 4 | 67 / 100 | Rose · 100% |
| iShares 20+ Year Treasury Bond ETF — economy or policy news | 8 | 1 | 21 / 100 | Not enough evidence |
| Invesco QQQ Trust — economy or policy news | 7 | 1 | 29 / 100 | Not enough evidence |
| Oil futures — company or asset news | 7 | 2 | 48 / 100 | Not enough evidence |
| SPDR Gold Shares — economy or policy news | 7 | 1 | 24 / 100 | Not enough evidence |
| Gold futures — company or asset news | 6 | 0 | 39 / 100 | Not enough evidence |
| SPDR S&P 500 ETF Trust — economy or policy news | 6 | 0 | 22 / 100 | Not enough evidence |
| VanEck Semiconductor ETF — company or asset news | 6 | 3 | 44 / 100 | Rose · 67% |
Different story links do not prove unique information. How much a source added, whether opposing evidence improved a decision and whether a market relationship helped a forecast remain unmeasured until their outcomes are linked.
How varied are today’s sources?
Publisher variety: 37.65 out of 100. The diversity score is the chance that two randomly chosen stories come from different groups. Higher is more varied.
| Publisher | Stories | Share |
|---|---|---|
| Reuters | 14 | 77.8% |
| CNA | 2 | 11.1% |
| MINDEF Singapore | 1 | 5.6% |
| World Health Organization | 1 | 5.6% |
Source locations · variety 36.42 out of 100
- United Kingdom / global wire: 14 stories · 77.78%
- Singapore: 3 stories · 16.67%
- Switzerland / international body: 1 stories · 5.56%
Languages · variety 0.00 out of 100
- English: 18 stories · 100.00%
Wire-service share · variety 34.57 out of 100
- Reuters: 14 stories · 77.78%
- Wire origin not recorded: 4 stories · 22.22%
Research areas · variety 94.44 out of 100
- AI Opportunities (SG): 1 stories · 5.56%
- Biotech & Pharma: 1 stories · 5.56%
- Circular Economy / Waste Management & Recycling: 1 stories · 5.56%
- Cybersecurity & Digital Trust: 1 stories · 5.56%
- Defence & Security: 1 stories · 5.56%
- Energy & Power: 1 stories · 5.56%
- Extreme Weather, Meteorology & Natural Disasters: 1 stories · 5.56%
- Geopolitics & International Relations / Conflict Analysis: 1 stories · 5.56%
- Law & Regulation: 1 stories · 5.56%
- Manufacturing & Automation: 1 stories · 5.56%
- Medical breakthroughs: 1 stories · 5.56%
- Public Policy & Government: 1 stories · 5.56%
- Semiconductors & Hardware: 1 stories · 5.56%
- Singapore News: 1 stories · 5.56%
- Sports Signals: 1 stories · 5.56%
- Venture & Startups: 1 stories · 5.56%
- Water & Utilities: 1 stories · 5.56%
- Wellness & Fitness: 1 stories · 5.56%
Source locations describe the publisher or organisation, not necessarily the country of the reporter or the event. An unrecorded wire origin remains unknown.
Next improvement: do not score whether a source helped a prediction until enough released predictions have properly checked results.
The current run contains a clear macro story and close Bitcoin quotes, yet the candidate remains unlocked because exact historical replay and event-timing proof are incomplete. Closed US markets also make several assets unsuitable for a new +20-hour rule. This is a useful forward lesson: a compelling narrative should not bypass clock quality. The Graph record has 49 investments, 152 headline events and 172 verified response windows, but zero causal claims, so it adds context rather than weight.
Fact and interpretation
Counts, quote timing and market closure are observed. The lesson is a prospective operating rule, not a causal conclusion.
Strongest opposition
Waiting for perfect evidence can miss a move, but the contract protects the public record from unresolvable calls.
What could change this view
Test this lesson prospectively across a declared sample before proposing any reversible workflow change; owner approval remains required.