The world’s own numbers
Every chart on this page is computed live from finished Bannerfall games — no samples, no simulations, and no player named.
Every finished game Bannerfall has ever recorded — 34 of them. "Capped" games are those that reached their turn limit and were decided on points rather than by conquest, so they are counted apart from the games that simply ran long.
| Rank | Faction | Win rate | 95% CI | Games |
|---|---|---|---|---|
| 1 | Sylvan Court | 56.3% | 39–72% | 32 |
| 2 | Vale Nomads | 27.3% | 10–57% | 11 |
| 3 | Stormcrest | 25.0% | 11–47% | 20 |
| 4 | Ironhold | 19.4% | 9–36% | 31 |
The pale band behind each ring is that faction’s 95% confidence interval — how much of its win rate is real and how much is still the small number of games behind it. A wide band means the ranking above it has not been earned yet.
| 5 | Duskborne | 5.0% | 1–24% | 20 |
| 6 | Ardent Legion | 4.3% | 1–21% | 23 |
| 7 | Azure Covenant | 0.0% | 0–30% | 9 |
| 8 | Emberfang | 0.0% | 0–23% | 13 |
34 decided games across 9 weeks. Each ribbon's thickness is that faction's wins in that week; a week nobody finished a game is a real zero, not a gap.
No language-model opponent has finished a turn in the last 14 days, so there is nothing to time. This is "no games", not "instant" — the number appears once one is played.
One square per day, from the first order Bannerfall ever recorded. A square is shaded against how busy the other busy days were — the darkest quarter of them are the darkest squares — so the pattern reads the same whether the world is quiet or at war. A day with no orders is drawn empty; a day that has not happened yet is not drawn at all.
| 2026-08-31 |
|---|
| 199 |
| 2026-09-01 | 969 |
|---|
| 2026-09-02 | 359 |
|---|
| 2026-09-05 | 13 |
|---|
| 2026-09-06 | 1 |
|---|
| 2026-09-08 | 1035 |
|---|
| 2026-09-09 | 2619 |
|---|
| 2026-09-10 | 1111 |
|---|
| 2026-09-11 | 137 |
|---|
| 2026-09-13 | 2 |
|---|
| 2026-09-14 | 195 |
|---|
| 2026-09-15 | 31 |
|---|
| 2026-09-16 | 2 |
|---|
| 2026-09-20 | 3 |
|---|