Certain US stocks have a structural pattern of exceeding consensus EPS quarter after quarter.
Using historical surprise data and a consensus correction model,
we identify stocks most likely to beat again — before the next earnings release.
Korean small-cap stocks often have no consensus coverage — so we build estimates from scratch. The US is the opposite: consensus data is abundant and reasonably accurate. The alpha comes from predicting how much a stock will beat it.
Wall Street consensus MdAPE is approximately 8.8% — already quite precise. This system doesn't try to outforecast the street on absolute numbers. Instead, it identifies stocks with a structural history of repeatedly exceeding expectations. Our corrected model achieves MdAPE of 6.32%, versus consensus at 7.27%.
Historical surprise patterns, corrected estimates, and consecutive beat streaks are combined to classify stocks into four tiers. Only TIER 1 receives priority analysis.
Corrected estimate ≥ +15% above consensus / 100% historical beat rate / consecutive beat streak ≥ 6 quarters / positive consensus. All four conditions must be met simultaneously.
pred = cons × (1 + clip(median historical surprise × 0.6, ±15%)). Past patterns are conservatively applied, with a noise-floor filter (MIN_EST = 0.20) to block near-zero consensus distortions.
A one-off beat is noise. A stock that has beaten consensus for 6+ consecutive quarters has a structural reason for doing so. We trust the pattern's persistence, not the individual data point.
Stocks with near-zero consensus are excluded from calculation. This prevents loss-reversal candidates from appearing at the top of the list through numerical illusion.
All four conditions met. Structurally high Beat probability. Immediate deep-dive analysis candidates. Full report provided.
Partial conditions met. Beat pattern exists but doesn't reach TIER 1 threshold. Tracked for upgrade.
No notable pattern. Expected to perform in line with consensus. Maintained on monitoring list.
Unprofitable or no consensus available. Excluded from analysis.
Actual quarterly EPS and the corresponding consensus EPS at time of release are collected for US stocks. Surprise rate = (actual − consensus) / |consensus| is computed for each quarter and accumulated as a historical record.
us_stock_earnings DB loadedMedian historical surprise per stock is computed and conservatively applied to the next consensus figure. K=0.6, CAP=±15% prevents overfitting. MIN_EST=0.20 blocks near-zero consensus noise.
Corrected EPS computedCorrected estimate +15%+ / 100% beat rate / streak ≥ 6 / positive consensus — all four must be met simultaneously. Stocks meeting all criteria are confirmed as TIER 1.
TIER 1 stocks confirmedEach TIER 1 stock gets a popup report: predicted vs. consensus EPS, next earnings date, streak count, surprise history bar chart, and quarterly earnings trend. Refreshed every earnings season.
TIER 1 reports loaded to DBThe US Top-Down TIER 1 list is available
on the Investment Code page.
All analysis is provided for reference purposes only. Investment decisions and outcomes are the sole responsibility of the investor.
Estimates are statistically derived from historical data and do not guarantee future earnings performance.