Hour-Aware Adaptive Risk Management for Autonomous Memecoin Trading: A Multi-Layer Intelligence Framework

We present UDAY (Unified Decentralized Autonomous Yardstick), a measurement framework for autonomous memecoin trading on Solana decentralised exchanges. UDAY measures four yardsticks of system behaviour: filter precision against forward-market outcomes, counterfactual rejection trajectories on tokens the system did not enter, fragility of the realised return distribution, and the realised yield of the deployed variant. The present paper additionally reports the time-of-day structure of trading outcomes across the four yardsticks. Across 190 paper-traded positions over fifteen days (March 29 to April 12, 2026), the streamlined variant achieved a 40.5 percent win rate, mean return of +0.62 percent per trade, and cumulative percentage return of +117.7 percent (sum of per-trade percentage returns; net SOL position-and-loss +0.039 SOL). The return distribution is left-skewed (skewness -1.21) with heavy tails (excess kurtosis 6.61); the system loses on a majority of trades and depends on a small number of large winners to push the mean and cumulative figures into positive territory. A Mann-Whitney U test of three blacklisted UTC hours (2, 13, 23) against all other hours yields U = 1,274, z = -1.23, p = 0.22; the directional pattern (blacklisted-hour mean -17.85 percent against +2.55 percent for other hours) is consistent across the sample but does not reach conventional statistical significance at n = 190. We report this comparison as exploratory rather than confirmatory because the three hours were identified as the worst-performing hours within the same sample used to test them. A counterfactual rejection-tracking system collected 4,874 forward-sample observations across 184 distinct rejection events. At the event level, 17.9 percent of rejected events reached at least a 50 percent drawdown from reference price within the 24-hour follow-up window. At the sample-observation level, 26.0 percent of forward-sample observations recorded the rejected token at half or less of its reference price. The filter stack avoided these forward-realised drawdowns, supporting the working hypothesis that the rejection criteria are net positive against the forward-market evidence base. Fragility analysis shows that removing the top three trades (1.6 percent of sample) renders cumulative return unprofitable, reflecting extreme concentration of profit in a small number of large winners.