Every parameter change and research finding documented. Most recent first.
Retrained entry model on 60K samples from April Geyser data using profitability labels. AUC 0.66, profitable at threshold 0.75 (+41.66 SOL on test set, 60.9% WR).
| param | from | to | reason |
|---|---|---|---|
| entry_model | model1_entry_universe.pkl (Jan) | model1_entry_april.pkl | January model scores all April tokens around 0.005 — useless. Retrained on April data with profitability labels: buy at frame, hold until 30% gain or 15% drawdown or 60 frame timeout, label=1 if profitable. |
| entry_threshold | 0.70 | 0.75 | Sweep on test set: 0.65 broke even, 0.75 gave +41.66 SOL on 1,007 trades with 60.9% WR. Higher than 0.80 had fewer trades but similar PnL. |
Tested 288 parameter combinations on 287K April trades using January-trained models. Every combination loses money. Best: -16.73 SOL.
Exit model trained on January data has 8% WR on short holds. Effectively disabling it and raising entry bar.
| param | from | to | reason |
|---|---|---|---|
| exit_threshold | 0.45 | 0.95 | Exit model sell_signal has 11% WR live. Every sell_score range loses money. Raised to 0.95 to effectively disable ML exits — only staleness, timeout, and drawdown rules remain. |
| entry_threshold | 0.50 | 0.70 | Entry scores 0.50-0.55 have 16% WR, 0.55-0.60 has 14% WR. Only 0.70+ entries show 22% WR. Raising threshold cuts entries by ~75% but improves quality. |
| stale_seconds | 30 | 60 | Stale exits at 30s caused -8.0 SOL losses. 3 tokens pumped 400%+ after we stale-exited. Doubling to 60s gives tokens more time to recover from quiet periods. |
| min_hold_frames | 0 | 15 | 27% of exits happened within 3 frames (8% WR, -1.28 SOL). Only 100+ frame holds had 50% WR. Forcing minimum 15-frame hold prevents instant sell-at-loss. |
Reverse-engineered 548 tokens that gained 100%+ from 3 days of Geyser data. Results contradict January patterns.
Beat used flat 0.0012 SOL per buy. Current market median is 0.000025 SOL — 48x lower.
Connected to BlockRazor Geyser, ran full pipeline on live pump.fun data for the first time.
| param | from | to | reason |
|---|---|---|---|
| paper_mode | N/A | true | First live run — paper trading only, no real transactions. |
| BlockRazor | N/A | connected | Geyser gRPC connected to geyserstream-tokyo.blockrazor.xyz:443. Ping OK. |
Autonomous experiment loop improved exit frame accuracy from 6.9% to 23.9%.
| param | from | to | reason |
|---|---|---|---|
| exit labels | symmetric ±3f | asymmetric -12/+1 | Penalizing early exits more than late exits nearly tripled frame accuracy. Model holds longer instead of firing early. |
| exit depth | 8 | 16 | Deeper trees with more estimators consistently improved accuracy: 8→10→12→14→16 each added 1-2%. |