Pre-launch· Token not live yet — no rewards or treasury until launchWhat this means
← Model
Launch structure, deeper trees

TRAIN v0.19

Same signals with deeper trees, to learn interactions such as 'bundle still holding while snipers exit'.
Live

Definition

Model type
Decision-tree ensemble
Signals used
Market + wallets + launch structure
Technical details
max_depth
3
subsample
0.8
n_estimators
300
learning_rate
0.03
min_samples_leaf
10
Random seed
1019
Config hash
72393ed8021cf9e9
Created
05 Oct 17:17
Completed
05 Oct 17:21
Artifact
models/v0.19/model.json
sha256 d57e130156e2f3d2…

Dataset & compute

Dataset
Solana Launch Dataset v11
7,225 snapshots · 1,450 launches · sha256 1a237aeb2f86
Training runs
1
Compute
TRAIN training server (CPU)
GPU hours
—
no GPU used
CPU time
207.7 s
Provider cost
$0.00
nothing purchased

Benchmarks

held-out test set · change vs v0.18
Spotting crashes (−70% within 1h)
0.955
Spotting survivors (still traded after 1h)
0.957
Spotting 2× runs (within 1h)
0.866
Calling the 1h direction (down / flat / up)
56.9%
Spotting creator dumps (within 1h)
0.880
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 0.993
n=814
collapse_6h
—
n=0
reach_2x_1h
AUC 0.919
n=814
survival_1h
AUC 0.956
n=814
survival_6h
—
n=0
survival_24h
—
n=0
migration_24h
—
n=0
trajectory_1h
acc 97.4%
n=814
trajectory_6h
—
n=0
creator_exit_1h
AUC 0.817
n=202

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
Trades per wallet−0.008
Trades, last 5 minutes−0.005
Net SOL into the curve−0.003
Creator's current holding−0.003
Back-and-forth wallets−0.001
Survival · 1h
Net SOL into the curve−0.144
Creator's current holding−0.013
Buy share of volume since launch−0.012
Volume acceleration−0.006
Volume since launch−0.005
Creator's launches in prior 24h (sampled)−0.005
Reaches 2× · 1h
Holders (from trades)−0.046
Price change, last minute−0.029
Net SOL into the curve−0.028
Creator's current holding−0.011
Holder growth, last 5 minutes−0.009
Buy share of volume since launch−0.004
Creator exit · 1h
Creator's current holding−0.105
Net SOL into the curve−0.049
Age of the launch−0.026
Buy share of volume since launch−0.018
Website set at launch−0.016
Market cap−0.014

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
05 Oct 17:17 · 00:03:24 · 6,411 examples · $0.00

Timeline

  1. 05 Oct 17:21
    TRAIN v0.19 deployed — now live
    average score across 5 tests 0.850 vs 0.843 for the live v0.18
  2. 05 Oct 17:21
    Benchmark completed: TRAIN v0.19
    Crash-spotting score regressed 0.956 → 0.949 vs live v0.18.
  3. 05 Oct 17:21
    Benchmark started: TRAIN v0.19
  4. 05 Oct 17:21
    Training completed: TRAIN v0.19
    00:03:24 · trained on 6411 examples · $0 compute cost
  5. 05 Oct 17:17
    Training started: TRAIN v0.19
    Solana Launch Dataset v11 · decision-tree ensemble · TRAIN training server
  6. 05 Oct 17:17
    TRAIN v0.19 queued
    Same signals with deeper trees, to learn interactions such as 'bundle still holding while snipers exit'.