Original Research

Guru Picks vs AI Scanner: Performance Data Compared.

We tracked 12 stock gurus' publicly posted picks for 6 months and compared the results against Banana Farmer's AI scanner signals over the same period. Some gurus performed well. Most didn't. The scanner was more consistent but less exciting. Here's the full data.

Study period: September 2025 through February 2026. Last updated: March 2026.

Key Findings

Guru stock picks averaged a 54% five-day win rate compared to the AI scanner's 80%. But the distribution mattered more than the average: the top 3 gurus outperformed the scanner on raw returns, while the bottom 5 would have lost subscribers money. Consistency was the scanner's biggest advantage. The gurus were a mixed bag.

Top-Line Results

1.Average guru accuracy: 54% at five days, vs 80% for the scanner. Guru accuracy ranged from 38% to 71% across the 12 tracked accounts.
2.Top 3 gurus averaged +6.2% per pick at five days (vs the scanner's +4.51%). When good gurus are right, their conviction picks produce bigger moves because they identify narrative-driven catalysts that algorithms miss.
3.Bottom 5 gurus averaged -1.8% per pick at five days. Three of the five had accuracy below 45%. Subscribers following these picks lost money net of subscription fees.
4.Timing decay hurt guru picks by 1.3% on average. We measured the price move between when a guru likely entered (estimated from chart patterns) and when subscribers could enter (after the public post). Subscribers got a worse average entry price on 78% of picks.
5.Scanner signals had 3.2x lower variance in returns. The standard deviation of five-day returns was 4.1% for scanner signals vs 13.2% for guru picks. Consistency vs conviction.

How We Measured This

We selected 12 stock gurus with significant followings (50K to 2M followers) on Twitter/X and YouTube who post specific, timestamped stock picks publicly. We excluded gurus who only post in paid Discord channels because we couldn't verify the timing of their picks. For each guru, we tracked every public pick with a specific ticker and direction over 6 months.

What counted as a “pick”

A post that named a specific ticker with a directional call. “AAPL looks strong here, adding to my position” counts. “Tech sector looks bullish” doesn't. We recorded 847 total guru picks across the 12 accounts over the 6-month period, averaging about 71 picks per guru. Some gurus posted 3-4 picks per week. Others posted 1-2 per month.

How we measured returns

For guru picks: the closing price on the day the pick was publicly posted was the baseline. We measured 1-day, 5-day, and 1-month returns from that baseline. For scanner signals: the closing price on the day the Ripe badge appeared was the baseline, measured at the same intervals. Both use closing prices, no intraday highs or lows.

The timing problem

Gurus often buy before they post. The pick goes public hours or days after the guru's actual entry. We estimated the guru's likely entry by looking at the price action before the post. This “timing decay” is one of the most important findings: by the time subscribers see the pick, 1.3% of the move has already happened on average. The scanner doesn't have this problem because scores update every 15 minutes and are visible to all users simultaneously.

What Guru Picks Look Like

Guru picks had a recognizable pattern across all 12 accounts. They tend to favor narrative-driven stocks (companies with compelling stories), recent runners (stocks already up 20-50% in recent weeks), and small-to-mid caps where the guru's audience can move the price. The average guru pick was a $4.2B market cap stock that had already gained 18% in the preceding 30 days.

Characteristic	Guru Picks	Scanner Signals
Avg market cap	$4.2B	$6.8B
Prior 30-day return	+18.3%	+8.7%
5-day win rate	54%	80%
5-day avg return (all)	+1.9%	+4.51%
5-day avg return (wins)	+8.4%	+5.64%
5-day avg return (losses)	-5.8%	-3.12%
Return std deviation	13.2%	4.1%

The most interesting number: guru winning trades returned +8.4% on average, higher than the scanner's +5.64%. When gurus are right, they're more right than the algorithm. The problem is they're wrong 46% of the time, and their losing trades averaged -5.8% (vs the scanner's -3.12%). The wider spread between wins and losses, combined with the lower accuracy, gives guru picks a lower expected value per trade despite the bigger individual wins.

Why Gurus and Scanners Perform Differently

The performance gap comes down to three structural differences between human stock picking and systematic scanning. Neither approach is strictly better. They have different strengths and failure modes.

Timing: simultaneous vs delayed

Scanner signals appear at the same time for all users because scoring runs on a fixed schedule. Guru picks reach subscribers after the guru has already positioned. That 1.3% average timing decay eats directly into subscriber returns. A guru with a genuine 60% win rate might deliver only 52% accuracy to subscribers because of the price slippage between the guru's entry and the subscriber's entry.

Coverage: narrow vs wide

The average guru follows 50-200 stocks actively. The scanner scores 9,287 assets every 15 minutes. The scanner catches momentum building in stocks that no guru is watching because they're too small, too boring, or outside the guru's sector expertise. Some of the best momentum signals come from stocks nobody is talking about yet. Those are invisible to gurus who follow the same popular names as everyone else.

Emotion: present vs absent

Gurus are human. They develop biases toward stocks they own. They hold losers too long because admitting a bad pick publicly is embarrassing. They chase recent winners because their audience rewards conviction. An algorithm has none of these problems. It scores every stock the same way every time. That's boring, but it removes the behavioral biases that degrade human stock-picking accuracy over time.

But gurus have one genuine edge

The best gurus identify narrative catalysts that algorithms can't quantify. A guru who deeply understands biotech can evaluate an FDA filing better than any scanner. A guru with industry contacts might know about a partnership before it's priced in. Three gurus in our study had five-day win rates above 65%, and their picks had an average return of +6.2%. These gurus earn their subscription fees. The challenge is identifying them before you subscribe, not after.

Limitations

This analysis has important limitations. We want to be honest about what the data proves and what it doesn't.

Sample size. 847 guru picks across 12 accounts over 6 months is a reasonable sample, but 6 months of market data captures only one market regime. During a bear market or a crash, results could invert. Some gurus specifically shine in volatile or declining markets. Our study period (September 2025 to February 2026) was largely bullish.

Selection bias in gurus. We chose gurus with large public followings who post picks openly. This excludes gurus who only share picks in paid, private channels. It's possible that the best-performing gurus keep their picks private and our sample skews toward less disciplined public posters.

Conflict of interest. We publish this on the Banana Farmer website. We benefit from the narrative that scanners outperform gurus. We've tried to present the data fairly (including that the top gurus outperformed our scanner on raw returns), but you should weigh our incentive when evaluating these results.

No position sizing data. Gurus often allocate more capital to their highest-conviction picks. We weighted every pick equally. A guru who bets big on winners and small on speculative picks would show better portfolio-level returns than our equal-weighted accuracy metric suggests.

Builder's Perspective

ABM

Aaron Browne-Moore

Founder, Banana Farmer

I follow three gurus myself. Two of them are in the “top 3” category in this study. They teach me things about specific sectors that no algorithm can replicate. I don't follow them for picks. I follow them for context and education.
The mistake I see traders make: treating guru picks as orders to execute rather than ideas to evaluate. If your guru says “buy TSLA” and you buy TSLA without checking the chart, the daily trend, and your own risk parameters, you're not trading. You're following.
Scanners and gurus can coexist in your workflow. Use the scanner for systematic, data-driven candidate generation. Use gurus for qualitative context on sectors you don't understand. But never let someone else's conviction replace your own analysis.

Important Disclaimer: This analysis is for educational and informational purposes only. It does not constitute financial advice or a recommendation to subscribe to or cancel any service.

Past performance does not guarantee future results. Guru performance and scanner performance vary by market conditions. We have a conflict of interest as the publisher and creator of one of the compared products.

Trading involves substantial risk of loss. See our full risk disclaimer.

Frequently Asked Questions

Common questions about guru vs scanner performance

Do stock gurus actually beat the market?

Some do, most don't. Academic research from CXO Advisory tracked 68 market gurus from 2005 to 2012 and found an average accuracy of 47%, slightly worse than a coin flip. However, accuracy alone doesn't determine profitability. A guru with 40% accuracy who cuts losses quickly and lets winners run can still be profitable. The problem is that most gurus report accuracy selectively, showing their winners and quietly deleting their losers.

How does AI stock scanning compare to guru picks?

In our analysis, AI scanning produced more consistent results with lower variance. Banana Farmer's Ripe signals showed an 80% five-day win rate across 12,450 signals. The guru picks we tracked averaged 54% accuracy at the same time horizon, but with wider distribution: some gurus hit 70%+ while others were below 40%. The scanner is more boring but more predictable. Gurus have higher upside potential when they're right, but higher downside when they're wrong.

Why do stock guru picks often underperform?

Three main reasons. First, timing decay: by the time a guru announces a pick to subscribers, the initial move has often already happened. Second, crowding: thousands of subscribers buying at the same time inflates the entry price and creates artificial demand that reverses when subscribers sell. Third, selection bias: gurus tend to pick stocks with good stories, not necessarily stocks with the best risk/reward setups based on data.

Should I cancel my guru subscription and use a scanner instead?

Not necessarily. Some gurus provide genuine educational value beyond their picks: market context, risk management frameworks, and sector expertise. If your guru teaches you to think independently, keep them. If your guru just sends tickers with entry prices and you follow blindly, that's a dependency, not a strategy. The best setup for most traders is a data-driven scanner for finding candidates plus your own analysis for making the final decision.

How did you measure guru pick performance?

We tracked publicly posted picks from 12 stock gurus on Twitter/X and YouTube over 6 months (September 2025 through February 2026). We recorded the closing price on the day the pick was publicly announced, then measured 1-day, 5-day, and 1-month returns. We only counted picks with specific tickers and directional calls (long or short). Vague commentary like "I like this sector" was excluded. Full methodology is in the article.

Is this analysis biased because Banana Farmer published it?

Yes, we have a conflict of interest. We're comparing our product against guru services that compete for the same audience. We've tried to be fair by acknowledging that some gurus had excellent track records during the study period, by including our own limitations, and by recommending that traders use multiple information sources. But you should read this with appropriate skepticism and verify the methodology yourself.

About This Research