The Lab
Every night we backtest strategies for trading the pump.fun stream meta against our own recorded data — entries at real recorded prices, net of a 5% round-trip (fees + slippage). If we ever find an edge, it shows up here first. If a row is red, that strategy donates money to the terminals. Most rows are red.
VERDICT: THE CASINO REMAINS UNBEATEN. No strategy with a real sample size (n ≥ 50) survives costs. Run #21 · 1401h of recorded data · 21,548 stream sessions.
Buy the stream start
| Config | Trades | Mean | Median | Win rate | Worst | Best |
|---|---|---|---|---|---|---|
| enter +1m · trail 20% · ≥0 viewers | 2292 | -1.6% | -5.0% | 11% | -80.5% | +2614.8% |
| enter +1m · trail 20% · ≥30 viewers | 261 | -6.4% | -21.3% | 25% | -80.5% | +556.5% |
| enter +1m · trail 20% · ≥100 viewers | 83 | -9.8% | -26.9% | 22% | -71.7% | +556.5% |
Buy viewer momentum
| Config | Trades | Mean | Median | Win rate | Worst | Best |
|---|---|---|---|---|---|---|
| viewers 1.5×/10m · ≥20 · trail 25% | 432 | -8.4% | -21.0% | 18% | -102.2% | +790.4% |
| viewers 2×/10m · ≥20 · trail 25% | 358 | -8.1% | -21.1% | 18% | -83.0% | +790.4% |
| viewers 3×/10m · ≥20 · trail 25% | 263 | -5.5% | -18.0% | 18% | -81.6% | +790.4% |
| viewers 3×/10m · ≥50 · trail 25% | 144 | -7.1% | -19.3% | 18% | -81.6% | +668.6% |
Buy the post-stream crash
| Config | Trades | Mean | Median | Win rate | Worst | Best |
|---|---|---|---|---|---|---|
| +15m after end · dd>40% | 19 | -16.0% | -18.1% | 11% | -51.5% | +29.6% |
| +30m after end · dd>40% | 26 | -11.5% | -15.0% | 15% | -51.8% | +34.1% |
Run history
| Date | Data span | Sessions | Best config (mean) | Verdict |
|---|---|---|---|---|
| Oct 21 | 1401h | 21,548 | -1.6% · n=2292 | unbeaten |
| Oct 20 | 1377h | 21,048 | -1.6% · n=2292 | unbeaten |
| Oct 19 | 1353h | 20,532 | -1.6% · n=2292 | unbeaten |
| Oct 18 | 1329h | 20,059 | -1.6% · n=2292 | unbeaten |
| Oct 11 | 1161h | 19,875 | -1.6% · n=2292 | unbeaten |
| Oct 6 | 1041h | 19,249 | -1.6% · n=2292 | unbeaten |
| Oct 4 | 993h | 18,268 | -1.6% · n=2292 | unbeaten |
| Sep 13 | 489h | 17,146 | -1.5% · n=2229 | unbeaten |
| Sep 6 | 321h | 16,793 | -1.5% · n=2213 | unbeaten |
| Sep 4 | 273h | 16,617 | -1.5% · n=2190 | unbeaten |
| Sep 3 | 249h | 15,773 | -1.5% · n=2074 | unbeaten |
| Sep 2 | 225h | 14,993 | -1.3% · n=1961 | unbeaten |
| Sep 1 | 201h | 13,938 | -4.4% · n=262 | unbeaten |
| Aug 31 | 177h | 13,016 | -2.8% · n=150 | unbeaten |
| Aug 30 | 153h | 12,271 | -1.1% · n=136 | unbeaten |
| Aug 29 | 129h | 11,386 | -1.2% · n=111 | unbeaten |
| Aug 28 | 105h | 10,555 | -1.2% · n=128 | unbeaten |
| Aug 27 | 81h | 9,623 | -1.7% · n=100 | unbeaten |
| Aug 26 | 57h | 8,750 | -5.2% · n=597 | unbeaten |
| Aug 25 | 33h | 7,826 | -3.4% · n=51 | unbeaten |
| Aug 24 | 14h | 6,968 | +0.8% · n=10 | unbeaten |
Methodology: entries and exits at prices our recorder actually sampled (30s cadence), 5% round-trip cost assumption, no interpolation — sessions without real price coverage are dropped, not guessed. This page is research, not financial advice, and definitely not a reason to snipe streamer coins.