Recording Unknowns Separately: Uncertainty in the Tenth Research Log

Unknowns in the Screening Funnel

Today's latest scan (snapshot) shows a total of 519 Binance USDT spot trading pairs, 271 within the market cap research range, and 74 passing all hard filters (Figure 1). Among the reasons for not passing hard filters: 168 with risk tags, 208 with market cap outside the research range, 32 not yet in a bottom area, 16 with insufficient or unstable liquidity, 1 with almost no trading, and 4 with too short a listing time.

These numbers tell us that the rule engine considers 74 targets eligible to enter the next round, but "passing hard filters" does not equal a conclusion, nor does it mean they will all be approved by AI review. In particular, note that the statistics on reasons for not passing may overlap; we only record the categories given by the engine and do not perform merged calculations.

The Potential Board Is Not an Approval List

Figure 2 shows the potential scores and below its peak declines for the top 10 on the potential board. YGG potential score 68, below its peak decline 91%; CITY 62, decline 75%; ASTR 60, decline 81%; CHR 58, decline 83%; MBL 57, decline 84%; ILV 57, decline 82%; CELR 56, decline 73%; MAV 55, decline 86%; LAZIO 54, decline 72%; GMT 54, decline 84%.

The potential board is the result of rule engine scoring, not AI selection. These scores only reflect the quantification of rules on bottom patterns, valuation, volume, and other dimensions; a large below its peak decline does not mean cheap, nor does it mean it will rise in the future. Among the top 10, only ASTR and CELR currently appear in publicly available review-approved signals, both with 4 buy votes from 4 reviewers; other coins do not yet have corresponding review approval records. No one should treat the potential board directly as a buy list.

Disagreement and Red Flags in the Review Council

In the past 24 hours, the review council sent 42 cases to review, approved 27, red-flag vetoed 11, had fewer than 3 affirmative votes, and 1 under review (Figure 1). In the red flag categories: 9 with announced delisting or listed for monitoring, 1 project halted, 1 recent large unlock.

The stance distribution of the four reviewers (Figure 3) shows: Fundamentals analyst 5 buy, 1 wait, 0 avoid; Skeptic 6 buy, 0 wait, 0 avoid; Technicals 4 buy, 2 wait, 0 avoid; Risk officer 25 buy, 5 wait, 12 avoid. The Risk officer's 12 avoid votes are significantly higher than the other three, indicating that from a risk control perspective, a considerable number of targets are deemed unacceptable. This disagreement is exactly why we keep the "avoid" vote: it makes uncertainty in AI judgment explicit rather than using majority votes to conceal differing opinions.

Track Record and On-chain Data Too Thin to Draw Conclusions

On-chain large transfer data over the past 24 hours shows: 10 transfers into exchanges, amount 0.69 million USD, involving 5 coins; 3 outflows from exchanges, amount 0.25 million USD, involving 1 coin; 119 transfers between whales, amount 17.13 million USD, involving 25 coins. These amounts and counts are very small, so their reference value is limited.

In terms of public track record, there is only 1 closed signal, with a 100% win rate, average per trade 9.89%, median 9.89%, maximum drawdown 0%, and holding days 4.4; 13 signals are still open, with 0 closed and 0 profitable in the last 7 days; exit reason was a take-profit after rising N% once. Since the closed sample is only 1 trade, this 100% win rate has no statistical significance, and we cannot claim the method is effective based on it. The profit-loss ratio is currently null and cannot be calculated. We report these "not yet conclusive" numbers truthfully rather than hiding them.

What This Means for Users

Today's theme is uncertainty. We choose to record unknowns separately: 74 passed the screening funnel, but most of them do not have public review-approved results; among the top 10 on the potential board, most also show no trace of AI approval; the Risk officer gave 12 avoid votes, indicating that even in AI review, clear disagreement arises between different roles; there is only 1 closed signal, and any conclusion about win rate or return is premature.

For users, these limitations mean that the data you see is only a process, not a result. We make failures and uncertainties public to prevent anyone from mistaking the "research log" for "trading signals." If the sample continues to grow, we will update these numbers in subsequent logs, but that is all we can say today.