We test every signal before you trust it.
Every night the Barqish engine replays 19,563 coin-days across 26 major coins since Jul 2024, using only what was known on each day. Then it checks what the price really did over the next 7 and 30 days. It keeps what works, drops what does not, and shows you both.
The evidence ranking, tested out of sample
Each quarter since Jul 2025, the model was rebuilt using only earlier data (with a 30-day gap so no answer could leak in), then used to rank the next quarter it had never seen. Bars show how often each band was higher 30 days later. This was a hard period for altcoins, so the base rate is low; what matters is the gap between the bands.
| Evidence band | Setups | Higher after 30d | Avg 30-day move | Median | Chance of a 20% dip | Higher after 7d |
|---|---|---|---|---|---|---|
| Top fifth | 1,735 | 45.8% | +0.2% | -1.3% | 32.8% | 49.6% |
| Upper | 2,765 | 42% | -3.3% | -3.3% | 34.3% | 45.2% |
| Middle | 2,536 | 36.2% | -6.5% | -5.8% | 40.1% | 41% |
| Lower | 2,178 | 34.4% | -6.4% | -5.4% | 33.3% | 42.5% |
| Bottom fifth | 1,539 | 35.1% | -5.0% | -5.7% | 27.7% | 43.3% |
| Any day | 10,753 | 38.5% | -4.5% | -4.4% | 34.3% | 44.1% |
| Test quarter | Setups | Top fifth, avg 30d | Bottom fifth, avg 30d | Result |
|---|---|---|---|---|
| Jul 2025 – Oct 2025 | 2,366 | -17.5% | -3.4% | bottom won |
| Oct 2025 – Jan 2026 | 2,366 | -8.9% | -30.3% | top won |
| Jan 2026 – Apr 2026 | 2,366 | +7.9% | -3.9% | top won |
| Apr 2026 – Jul 2026 | 2,330 | -0.7% | -8.1% | top won |
| Jul 2026 – Sep 2026 | 1,325 | +32.5% | +19.8% | top won |
The latest quarter is still in progress, so it has fewer setups.
What did not work, and why we changed it
Most apps sell a classic trend checklist: buy when price is above its averages, momentum is rising and the coin is beating Bitcoin. We scored every day with that checklist and checked the results. Over this period it had no reliable edge: chasing strength was often punished. So the Barqish ranking no longer rewards it. The checklist still describes the chart on the desk, but the odds you see come from the evidence model.
| Checklist grade | Setups | Higher after 7d | Higher after 30d | Avg 30-day move |
|---|---|---|---|---|
| A | 2,469 | 44.6% | 43.3% | +1.1% |
| B | 2,658 | 41.8% | 41.5% | +0.1% |
| C | 4,027 | 44.6% | 43.9% | +1.9% |
| D | 3,602 | 45.3% | 42.8% | +1.7% |
| F | 6,807 | 49.7% | 45.5% | +2.3% |
| Any day | 19,563 | 46.1% | 43.9% | +1.7% |
What the evidence rewards
Each signal gets a weight equal to how strongly it lined up with the next 30 days. Green pushes a coin up the ranking, red pushes it down. The count shows in how many of the 5 test quarters the signal pointed the same way, which is how we tell a real pattern from luck.
Today’s evidence ranking
Screened coins from the daily radar, ranked by the same model. Not-permissible coins are left out.
Loading today’s ranking...
Live calls, not just backtests
Since 24.09.2026, 75 real Barqish radar calls have been graded 24 hours later and 63% held. Every one is public on the track record and the receipts wall.
How the test works
- Point in time. On each test day the engine sees only prices up to that day. Swing highs and lows count only after they are confirmed.
- Walk-forward. The model is rebuilt every quarter on older data only, with a 30-day gap, and judged on the next quarter it never saw. No result on this page is in-sample except the classic checklist table.
- Transparent model. A weighted score of eight plain signals. No black box: every weight is shown above.
- Honest outcomes. Higher or lower after 7 and 30 days, the average and median move, and how often a 20% dip happened along the way.
- Recalibrated nightly as new days are added.
- Limits. No fees or slippage. Spot only, no leverage. Days across coins overlap and move together, so treat the numbers as evidence, not guarantees. Markets change regime; past behaviour does not guarantee future results.
Educational research, not financial advice. Crypto is volatile and you can lose money.