Pre-launch· Token not live yet — no rewards or treasury until launchWhat this means
← Model
Deeper decision trees

TRAIN v0.05

Deeper trees (depth 3), more of them (300) with a lower learning rate.
Completed

Definition

Model type
Decision-tree ensemble
Signals used
Market signals
Technical details
max_depth
3
subsample
0.8
n_estimators
300
learning_rate
0.03
min_samples_leaf
10
Random seed
1005
Config hash
e5bc765062b13e31
Created
05 Oct 05:12
Completed
05 Oct 05:12
Artifact
models/v0.05/model.json
sha256 9e90dc54dae50eb4…

Dataset & compute

Dataset
Solana Launch Dataset v4
1,170 snapshots · 235 launches · sha256 ebbadfe5c81d
Training runs
1
Compute
TRAIN training server (CPU)
GPU hours
—
no GPU used
CPU time
8.4 s
Provider cost
$0.00
nothing purchased

Benchmarks

held-out test set · change vs v0.04
Spotting crashes (−70% within 1h)
0.914
Spotting survivors (still traded after 1h)
0.882
Spotting 2× runs (within 1h)
0.833
Calling the 1h direction (down / flat / up)
55.1%
Spotting creator dumps (within 1h)
0.894
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 0.972
n=145
collapse_6h
—
n=0
reach_2x_1h
—
n=145
survival_1h
AUC 0.801
n=145
survival_6h
—
n=0
survival_24h
—
n=0
migration_24h
—
n=0
trajectory_1h
acc 95.9%
n=145
trajectory_6h
—
n=0
creator_exit_1h
AUC 0.786
n=51

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
This version was trained before TRAIN started measuring which signals each model relies on.

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 05:12 · 00:00:04 · 1,025 examples · $0.00

Timeline

  1. 05 Oct 05:12
    TRAIN v0.05 not deployed; v0.01 stays live
    Reason: no shared benchmarks yet; keeping the live model.
  2. 05 Oct 05:12
    Benchmark completed: TRAIN v0.05
    Crash-spotting score: not enough held-out data yet (n=65).
  3. 05 Oct 05:12
    Benchmark started: TRAIN v0.05
  4. 05 Oct 05:12
    Training completed: TRAIN v0.05
    00:00:04 · trained on 1025 examples · $0 compute cost
  5. 05 Oct 05:12
    Training started: TRAIN v0.05
    Solana Launch Dataset v4 · decision-tree ensemble · TRAIN training server
  6. 05 Oct 05:12
    TRAIN v0.05 queued
    Deeper trees (depth 3), more of them (300) with a lower learning rate.