How Kasiel Tracks Its AI Stock Analysis Accuracy (Publicly)
Detailed explainer of how Kasiel publicly tracks every AI stock analysis verdict, the auto-scoring methodology at 30 days, and why this matters for evaluating AI stock analyzers.
Why public track records matter
Most AI stock analyzers publish 'backtested performance' — historical simulations of how their strategy would have performed if you'd followed every signal. These numbers are typically wildly optimistic because of survivorship bias (only the live products are tested), look-ahead bias (the model sees data it wouldn't have had at the time), and overfitting (the model is tuned to historical results). Treat backtested numbers as marketing, not evidence. What's much more credible: a live, timestamped, public track record where every call is published before the outcome is known.
How Kasiel publishes every verdict
Every public Kasiel analysis (Tournament Winner, Monthly Gem, or any analysis the user marks public) is auto-tweeted to @kasielanalysis on X the moment it runs. The tweet includes the verdict, confidence score, key levels, and a link to the full report. The timestamp is on the tweet — it can't be retroactively edited. This means anyone can verify the call was made before the outcome was known.
How outcomes are scored
At fixed maturity intervals (7 days, 30 days, 45 days, 60 days, 90 days, 180 days, 365 days), each verdict is auto-scored against real market prices. The scoring is deterministic: BUY/STRONG BUY where the stock rose more than 5% is Right; SELL/STRONG SELL where it fell more than 5% is Right; WAIT where the stock moved less than 5% is Right. Direction was correct but magnitude wasn't strong is Close. Direction was wrong is Wrong. The 30-day result is automatically tweeted when each call matures, regardless of outcome.
What the scoreboard looks like
The /results page shows the full live scoreboard. Top wins, top losses, win rate, average return per call, leveraged-equivalent returns at 3x. You can filter by verdict type, by ticker, by date range. Every individual call is clickable — you can read the original analysis and see how it played out. There's no curation; nothing is hidden because it embarrassed us.
Why this is unusual
Most AI stock analyzers either don't publish individual call outcomes at all, or publish them selectively. Some publish portfolio-level returns that don't tell you whether individual calls were any good. Some publish a 'win rate' without explaining how wins are defined. Kasiel publishes everything, deterministically, automatically. The track record is the credential — if our public calls earn your trust, that's the credibility we've earned. If they don't, no marketing claim should override what you can see.
Limitations to be honest about
The track record so far covers a few months — meaningful but not a full market cycle. Performance in a bull market doesn't guarantee performance in a bear market or sideways tape. The 30-day scoring window is a short horizon; longer holds may produce different outcomes (which is why we also track at 90, 180, and 365 days). And the win rate is calculated across all matured calls, not weighted by position size or by your ability to actually execute at the levels we publish. Use the track record as a useful signal of system quality, not as a guarantee of your individual returns.
Try AI Stock Analysis Free
Get a free institutional-grade Lite analysis on signup. No credit card required.
Analyze a Stock