Loading data...
Loading data...
This engine compares each stock's current technical state (price, momentum, volatility, volume) against hundreds of historical states of that same stock only — surfacing stocks whose current setup closely resembles states that historically preceded a positive move.
Not enough data right now to show a list. The list only includes stocks with at least 8 real, non-overlapping historical matches and a meaningfully positive median 20-day return following similar states.
Every number here is computed from real historical prices only — no AI guessing, no fabricated data. But this is not a calibrated forecast: the engine has not yet gone through walk-forward validation against a period it wasn't built from, and nobody has measured whether, say, a "73% historical positive rate" actually plays out near 73% going forward. Treat this as real historical pattern research, not a buy recommendation.
Every stock that entered the Top 10 gets recorded on the real day it entered, and after 20 real trading days we check whether it actually moved the way history suggested — no number here is computed in advance, only after it actually happened.
No prediction has gone through 20 real trading days yet — tracking runs automatically from here on; check back in a few weeks for the first real result.
Every stock that enters the Top 10 (and isn't already held) is virtually bought with $10,000. It sells when the real price reaches the historical target (the median return) — a win — or drops 7% from entry — a loss. No real money moves; this exists purely to check whether the historical edge actually holds up when traded mechanically.
10
Open positions
—
Win rate (0 wins / 0 losses)
$-1,079
Total P&L (realized + unrealized)
+$0
Of which realized (closed trades)
—
Avg. return per closed trade
Currently open positions:
| Stock | Entered on | Similarity | Entry price | Current price | Target | Stop-loss |
|---|---|---|---|---|---|---|
| NCLH | 2026-09-14 | 90/100 (15 matches) | $14.82 | $14.15 (-4.5%) | $16.08 | $12.84 |
| MAR | 2026-09-14 | 87/100 (15 matches) | $334.69 | $336.46 (+0.5%) | $354.68 | $308.34 |
| MPWR | 2026-09-14 | 84/100 (15 matches) | $1234.46 | $1205.41 (-2.4%) | $1357.37 | $1049.29 |
| HAL | 2026-09-14 | 88/100 (15 matches) | $35.84 | $33.69 (-6.0%) | $37.89 | $31.29 |
| CVNA | 2026-09-14 | 82/100 (15 matches) | $69.16 | $65.48 (-5.3%) | $74.95 | $58.79 |
| CCL | 2026-09-14 | 85/100 (15 matches) | $22.75 | $21.91 (-3.7%) | $24.30 | $20.06 |
| DECK | 2026-09-14 | 83/100 (15 matches) | $81.27 | $78.68 (-3.2%) | $88.09 | $69.48 |
| HLT | 2026-09-14 | 82/100 (15 matches) | $306.19 | $303.97 (-0.7%) | $320.96 | $280.65 |
| EW | 2026-09-14 | 85/100 (15 matches) | $84.37 | $88.46 (+4.8%) | $88.86 | $77.29 |
| CNP | 2026-09-14 | 82/100 (15 matches) | $39.11 | $38.37 (-1.9%) | $40.71 | $36.14 |
No stocks currently clear the statistical bar to display. The list refreshes automatically — check back later.
For each stock, the engine builds a daily "state vector" (returns, RSI, MACD, distance from moving averages, volatility, volume) using only data that was known up to that day(no look-ahead), then searches that same stock's own history for days whose state vector was statistically similar to today's. Matches that fall too close together in time (within 10 trading days) aren't counted separately, so a single rally doesn't get counted as dozens of "independent" examples.
Beyond technicals: the vector also includes real institutional activity (13F holdings, quarterly resolution) and fundamentals/growth (current ratio, debt-to-equity, ROIC, revenue/EPS growth). Fundamentals exist in our data only at annual resolution(not quarterly) and with no real report-publication date, so a report is treated as "known" roughly 75 days after its fiscal year end, and carries a materially lower weight in the overall similarity score than the daily-updated fields.
"Analog quality" (0-100) is different from "similarity" — it also accounts for how many matches were found, how much data was available, and how many matches occurred in a market regime similar to today, so a single high-similarity match doesn't look more reliable than it actually is. For the full methodology and its limitations, see the methodology page.