Have you ever asked an AI the exact same question twice and gotten slightly different answers?
You're not imagining it.
That's just how large language models work.
The Randomness Problem#
Every AI model has a setting called temperature. Think of it like adding a small amount of randomness to each response.
Even with a low temperature, the AI can make slightly different choices each time it analyzes the same stock.
For example:
- First run: Bullish – 74% confidence
- Second run: Bullish – 68% confidence
- Third run: Neutral – 58% confidence
Nothing changed—not the stock, not the market, not the AI model.
The AI simply took a slightly different path.
That's perfectly normal for AI, but it's not ideal when you're researching investments.
After all, you wouldn't ask one friend for advice and immediately make a big financial decision.
Our Solution: Ensemble Runs#
Instead of relying on a single AI analysis, Valarn runs the entire research process three separate times.
Each run is completely independent.
Each produces its own market outlook and confidence score.
Then we compare the results.
Think of it like asking three experienced analysts instead of one.
If all three return a Bullish outlook, that's a much stronger signal than just one AI reaching that conclusion.
If two are Bullish and one is Neutral, the overall outlook is still Bullish—but you'll also know there wasn't complete agreement.
And if all three disagree?
That's the AI politely saying, "This one's a head-scratcher." Markets do that sometimes.
Why It Works#
Running multiple independent analyses helps reduce random AI noise.
Instead of reacting to one lucky—or unlucky—response, Valarn looks for consistency.
The result is research that's more stable and more trustworthy.
In our testing:
- Standard analysis produced the same overall outlook 81% of the time when repeated.
- Ensemble Runs increased that to 94%.
That's a 13-point improvement in consistency.
Fast Enough to Use Every Day#
You might think running three analyses would take three times longer.
Fortunately, it doesn't.
Valarn runs all three analyses at the same time.
Most Ensemble Runs finish in 35–60 seconds, only about 40% slower than a standard run.
When Should You Use It?#
Ensemble Runs are a great choice when:
- You're researching an important investment.
- Market conditions are especially volatile.
- The standard analysis has only moderate confidence.
- You want additional confirmation before making your own decision.
A standard run is usually enough when:
- You're researching many stocks quickly.
- The AI already has very high confidence with strong agreement.
- You're doing early-stage research and exploration.
Understanding Convergence#
Ensemble Runs include a Convergence Score.
This tells you how much the three analyses agreed with each other.
- 90–100: Strong agreement. All three reached the same outlook.
- 60–89: Moderate agreement. Two out of three agreed.
- 0–59: Low agreement. The AI found mixed evidence and no clear consensus.
Two analyses can both return a Bullish outlook with 80% confidence, but they aren't equally convincing.
A Bullish outlook with a Convergence Score of 95 is much stronger than one with a score of 62.
That's why Valarn shows both numbers.
Understanding how much the AI agrees with itself is often just as valuable as the confidence score.
The Bottom Line#
AI is incredibly powerful, but it isn't magic.
Small differences between runs are completely normal.
Instead of hiding that reality, Valarn embraces it.
By combining multiple independent analyses into one final market outlook, Ensemble Runs reduce random noise, improve consistency, and give you a clearer picture of what the AI is seeing.
And yes... sometimes even three AI analysts can't agree.
If you've ever watched financial news, you'll know that's actually pretty realistic.
Ensemble Runs are available on all Pro and Team plans.
Valarn
Founder
Valarn Research Team
