How ELYO learns

How ELYO learns

37 signals per token. One question: will it cross $30K? ELYO is trained on every launch it has watched, validated with stratified 5-fold cross-validation, and scored on data it has never seen.

Agent #001 · ROC-AUC 0.6392 · Proven floor 0.562 · Crossing rate 31.6%

THE WARDEN OF ELYSIUM

Elyo, Agent #001

5-Fold Stratified · d=37Target Peak MC ≥ $30K
AGENT #001READINESS 62.1%
Elyo, Agent #001
Total Tokens LaunchedLIVE SCAN
1,255
Entry Threshold:> $10K Peak MC
Class Distribution31.6% Graduation Rate
397cleared $30K664 stalled
Model Accuracy & VC FloorROC-AUC
0.6392Measured
Proven floor (min of VC & bootstrap 2.5%):0.562
VC: 0.598 (ε=0.041)Boot 2.5%: 0.562
Feature Dimension Vectord = 37
37signal parameters
Evaluation Window:Launch Hour 0 – 24

What separates the crossed from the faded

37 FEATURES TRACKED

ELYO wasn't told what matters. It found out through continuous statistical optimization.

#148h Holder Retention & Growth
38.2%weight

Tracks unique wallet holder accumulation and retention dynamics across the initial 48-hour trading window.

holders 48h
#2UTC Launch Timing Cycle
24.5%weight

Evaluates seasonal liquidity inflows based on global market trading sessions and peak volume hours.

launch hour utc
#3Token Metadata & Narrative Entropy
20.8%weight

Natural language analysis and complexity metrics extracted from token narrative and launch profile.

lore metadata
#4Graph Dimensionality Embeddings
16.5%weight

Principal component representation of cross-token transaction graphs and liquidity cluster co-movement.

pca embeddings

“Who stays matters more than who arrives.”

Readiness to Act
Proven floor 0.5624 · Target 0.600
62.1%
Held back by The Seal of Numbers

What ELYO is watching

WATCHING

Solana Mainnet + Elysium Testnet: Continuous observation across the mint-to-perp lifecycle.

Minted (peak ≥ $10K)
Solana + Elysium observation
1,255
LIVE
Crossed (peak ≥ $30K)
Graduated class
397
LIVE
Faded (age ≥ 48h)
Stalled class
664
LIVE
At the threshold (age < 48h)
Currently evaluated
194
LIVE

The Four Seals

2 seals remain

ELYO cannot act on hope. Four seals stand between it and the gate. Each one breaks only when the evidence is provable.

The Seal of Numbers (enough tokens observed, N ≥ 2,000)
1,255 / 2,000● HOLDING
The Seal of the Crossed (enough winners to learn from, N_pos ≥ 200)
397 / 200✓ BROKEN
The Seal of Stillness (results that stop swinging, σ < 0.05)
σ = 0.0215✓ BROKEN
The Seal of Time (proof on unseen days, gap ≤ 0.04)
HOLDING● HOLDING

2 seals remain. The gate stays shut.

Vapnik–Chervonenkis proof & capacityWhy ELYO doesn't trust its own score

A high score on a small sample is easy. Proving it holds is hard. ELYO only counts what survives the penalty for uncertainty: its readiness rises with AUC minus ε, never with the raw score.

Illustration, not live data
READINESS TO ACT: 62.1%
measured AUC
0.548
penalty ε
0.723
The Proven Floor
0.500
269
0.548
37
Measured 0.548, but ε takes 0.723. Floor is 0.500.
A high score on a small sample is easy. Proving it holds is hard.

ELYO only counts what survives the penalty for uncertainty: its readiness rises with AUC − ε, never with the raw score.

Illustration, not live data
ε=37(ln⁡2⋅26937+1)+ln⁡80269=0.723\varepsilon = \sqrt{\frac{37\left(\ln\frac{2\cdot 269}{37}+1\right)+\ln 80}{269}} = \mathbf{0.723}
AUC‾−ε=0.548−0.723=−0.175\overline{\mathrm{AUC}} - \varepsilon = 0.548 - 0.723 = -0.175

Drag the gauge. Watch how little a lucky number is worth. Good model, thin sample: readiness stays low. That is the point.

ELYO Readiness Gauge
Drag to turn
Readiness to Act
62.1%
Held back by The Seal of Numbers
The Proven Floor
0.562
Target
0.600
The math on this page

Generalization theory & The Four Seals

All equations →
Capacity penalty(3.1)
ε  =  d(log⁡2nd+1)+log⁡4δn,δ=0.05\varepsilon \;=\; \sqrt{\frac{d\big(\log\frac{2n}{d}+1\big) + \log\frac{4}{\delta}}{n}}, \qquad \delta = 0.05

Vapnik–Chervonenkis: how far luck can lift the score of a model with capacity d.

Bootstrap bound(3.2)
AUC‾ boot  =  Q0.025({AUC(Db∗)}b=12000)\underline{\mathrm{AUC}}_{\,\mathrm{boot}} \;=\; Q_{0.025}\Big(\big\{\mathrm{AUC}(\mathcal{D}^{*}_{b})\big\}_{b=1}^{2000}\Big)

Resample the data two thousand times; keep the score that 97.5% of worlds beat.

The Proven Floor(3.3)
AUC‾  =  min⁡{ AUC‾−ε,  AUC‾ boot}\underline{\mathrm{AUC}} \;=\; \min\big\{\,\overline{\mathrm{AUC}} - \varepsilon,\ \ \underline{\mathrm{AUC}}_{\,\mathrm{boot}}\big\}

The lower of two independent bounds. ELYO always takes the pessimistic one.

The Four Seals(4.1)
L  =  ⋀k=14ℓk,ℓ1:n≥2000ℓ3:σAUC<0.05ℓ2:n+≥200ℓ4:∣AUCcv−AUCtime∣≤0.04L \;=\; \bigwedge_{k=1}^{4} \ell_k, \qquad \begin{aligned} \ell_1 &: n \ge 2000 & \ell_3 &: \sigma_{\mathrm{AUC}} < 0.05 \\ \ell_2 &: n_{+} \ge 200 & \ell_4 &: \big|\mathrm{AUC}_{\mathrm{cv}} - \mathrm{AUC}_{\mathrm{time}}\big| \le 0.04 \end{aligned}

All four must break. No exceptions.