Pre-launch· Token not live yet — no rewards or treasury until launchWhat this means
← Model
Adds launch structure (bundles, snipers, dev) — retrained on new data

TRAIN v0.25

Same configuration as v0.18 (best so far on the headline benchmark), retrained on the newest dataset.
Completed

Definition

Model type
Decision-tree ensemble
Signals used
Market + wallets + launch structure
Technical details
max_depth
2
subsample
0.8
n_estimators
200
learning_rate
0.05
min_samples_leaf
10
Random seed
1025
Config hash
f5f6ee3233427e54
Created
06 Oct 00:42
Completed
06 Oct 00:45
Artifact
models/v0.25/model.json
sha256 2b01fa030c8565b9…

Dataset & compute

Dataset
Solana Launch Dataset v17
10,305 snapshots · 2,068 launches · sha256 0eb44f139f93
Training runs
1
Compute
TRAIN training server (CPU)
GPU hours
—
no GPU used
CPU time
175.7 s
Provider cost
$0.00
nothing purchased

Benchmarks

held-out test set · change vs v0.24
Spotting crashes (−70% within 1h)
0.944
Spotting survivors (still traded after 1h)
0.972
Spotting 2× runs (within 1h)
0.863
Calling the 1h direction (down / flat / up)
55.9%
Spotting creator dumps (within 1h)
0.882
Spotting crashes (within 6h)
insufficient data
Spotting survivors (6h)
insufficient data
Calling the 6h direction
insufficient data
Spotting survivors (24h)
insufficient data
Spotting migrations (within 24h)
insufficient data

Validation metrics (during training)

collapse_1h
AUC 0.994
n=1224
collapse_6h
AUC 1.000
n=9
reach_2x_1h
AUC 0.917
n=1224
survival_1h
AUC 0.951
n=1224
survival_6h
—
n=9
survival_24h
—
n=0
migration_24h
—
n=0
trajectory_1h
acc 97.5%
n=1224
trajectory_6h
acc 100.0%
n=9
creator_exit_1h
AUC 0.838
n=275

Validation launches are separate from training launches and from the locked test set. Benchmarks above are the public numbers.

What this version relies on

held-out validation launches
Collapse risk · 1h
Volatility, last 5 minutes−0.011
Net SOL into the curve−0.009
Trades per wallet−0.003
Buy share of volume, last 5 minutes−0.002
Trades, last 5 minutes−0.001
Holder growth, last 5 minutes−0.001
Survival · 1h
Net SOL into the curve−0.180
Buy share of volume since launch−0.010
Share of tiny (<$3) trades−0.005
Volume acceleration−0.005
Snipers (first 10s)−0.002
Creator's current holding−0.002
Reaches 2× · 1h
Net SOL into the curve−0.051
Holders (from trades)−0.036
Price change, last minute−0.026
Holder growth, last 5 minutes−0.011
Creator's current holding−0.007
Volatility, last 5 minutes−0.005
Creator exit · 1h
Net SOL into the curve−0.089
Creator's current holding−0.069
Age of the launch−0.018
Buy share of volume since launch−0.011
Buy share of volume, last 5 minutes−0.007
Website set at launch−0.005

For each signal, how much the score drops when that signal is scrambled across launches. TRAIN learns these weights itself from the data; nobody hand-codes them.

Training runs

Attempt 1Succeeded
06 Oct 00:42 · 00:02:55 · 9,081 examples · $0.00

Timeline

  1. 06 Oct 00:45
    TRAIN v0.25 not deployed; v0.19 stays live
    Reason: average score across 5 tests 0.844 vs 0.845 for the live v0.19.
  2. 06 Oct 00:45
    Benchmark completed: TRAIN v0.25
    Spotting crashes (−70% within 1h) regressed 0.955 → 0.944 vs live v0.19.
  3. 06 Oct 00:45
    Benchmark started: TRAIN v0.25
  4. 06 Oct 00:45
    Training completed: TRAIN v0.25
    00:02:55 · trained on 9081 examples · $0 compute cost
  5. 06 Oct 00:42
    Training started: TRAIN v0.25
    Solana Launch Dataset v17 · decision-tree ensemble · TRAIN training server (CPU)
  6. 06 Oct 00:42
    TRAIN v0.25 queued
    Same configuration as v0.18 (best so far on the headline benchmark), retrained on the newest dataset.
  7. 06 Oct 00:42
    Enough new data for another training run
    6,368 new labelled snapshots since Solana Launch Dataset v9.