Virtua Fighter 5 REVO · v1.10 · Character Matchup Winrates
Dural vs Lei Fei
Matches counted265
Win rate60.8%
Widest swing-17.0 pts
Survive correction0
Period2026/03/26 – 2026/09/23
Where it was fought
Dural's win rate on each stage, against this matchup everywhere| Stage | Stage effect | Win % | Change | p | Matches |
|---|---|---|---|---|---|
| Low fence | 75.0% | +14.2 | 0.229 | 16 · | |
| Breakable full fence | 72.0% | +11.2 | 0.226 | 25 · | |
| Open | 66.7% | +5.9 | 0.563 | 21 · | |
| Half fence | 63.0% | +2.2 | 0.804 | 27 · | |
| Octagon | 61.5% | +0.8 | 0.931 | 26 · | |
| Breakable half fence | 59.4% | -1.4 | 0.865 | 32 · | |
| Single wall | 57.1% | -3.6 | 0.679 | 28 · | |
| Full fence | 54.8% | -5.9 | 0.473 | 31 · | |
| Rectangle | 50.0% | -10.8 | 0.169 | 34 · | |
| Fence and open | 43.8% | -17.0 | 0.151 | 16 · |
p < 0.05 on its ownStill there after correcting for how many were tested
Stage against stage
Dural's win rate on the row's stage, minus the same on the column'sAll ten shapes at once: chi-square 7.12 on 9 df, p 0.625
| Stage | Full | Break. full | Half | Break. half | Low | Wall | Mixed | Open | Rect | Oct |
|---|---|---|---|---|---|---|---|---|---|---|
| Full fence | 54.8%31 | -17.20.187 | -8.10.531 | -4.50.716 | -20.20.178 | -2.30.859 | +11.10.471 | -11.80.394 | +4.80.696 | -6.70.610 |
| Breakable full fence | +17.20.187 | 72.0%25 | +9.00.488 | +12.60.322 | -3.00.833 | +14.90.260 | +28.20.070 | +5.30.695 | +22.00.089 | +10.50.428 |
| Half fence | +8.10.531 | -9.00.488 | 63.0%27 | +3.60.778 | -12.00.416 | +5.80.660 | +19.20.220 | -3.70.790 | +13.00.311 | +1.40.915 |
| Breakable half fence | +4.50.716 | -12.60.322 | -3.60.778 | 59.4%32 | -15.60.286 | +2.20.861 | +15.60.306 | -7.30.592 | +9.40.445 | -2.20.867 |
| Low fence | +20.20.178 | +3.00.833 | +12.00.416 | +15.60.286 | 75.0%16 | +17.90.236 | +31.20.072 | +8.30.583 | +25.00.095 | +13.50.369 |
| Single wall | +2.30.859 | -14.90.260 | -5.80.660 | -2.20.861 | -17.90.236 | 57.1%28 | +13.40.392 | -9.50.498 | +7.10.575 | -4.40.743 |
| Fence and open | -11.10.471 | -28.20.070 | -19.20.220 | -15.60.306 | -31.20.072 | -13.40.392 | 43.8%16 | -22.90.163 | -6.20.680 | -17.80.261 |
| Open | +11.80.394 | -5.30.695 | +3.70.790 | +7.30.592 | -8.30.583 | +9.50.498 | +22.90.163 | 66.7%21 | +16.70.226 | +5.10.716 |
| Rectangle | -4.80.696 | -22.00.089 | -13.00.311 | -9.40.445 | -25.00.095 | -7.10.575 | +6.20.680 | -16.70.226 | 50.0%34 | -11.50.373 |
| Octagon | +6.70.610 | -10.50.428 | -1.40.915 | +2.20.867 | -13.50.369 | +4.40.743 | +17.80.261 | -5.10.716 | +11.50.373 | 61.5%26 |
Better on the row's stage — deeper at 5 and 10 pointsBetter on the column's stagep < 0.05 on its ownStill there after correcting for how many were tested· fewer than 80 decided matches on one side of the comparison
How significance is decided
- Each gap is two win rates compared with a pooled two-proportion z test, over decided matches only.
- The 95% interval is Newcombe's hybrid score method, which keeps its shape near 0% and 100% where a symmetric interval does not.
- q is p corrected for how many tests were asked at once, by Benjamini–Hochberg at a 5% false discovery rate. A figure is called real on q, never on p alone.
- Standard errors are widened by a measured design effect, because one player appears in many matches and those matches are not independent of each other.
- Stages are assigned at random, so a gap between them is the stage's own effect and not a difference in who chose to play where.
Matches counted258
Win rate61.2%
Widest swing-17.5 pts
Survive correction0
Period2026/03/26 – 2026/09/23
Where it was fought
Dural's win rate on each stage, against this matchup everywhere| Stage | Stage effect | Win % | Change | p | Matches |
|---|---|---|---|---|---|
| Low fence | 73.3% | +12.1 | 0.322 | 15 · | |
| Breakable full fence | 72.0% | +10.8 | 0.245 | 25 · | |
| Open | 66.7% | +5.4 | 0.594 | 21 · | |
| Half fence | 65.4% | +4.1 | 0.647 | 26 · | |
| Octagon | 64.0% | +2.8 | 0.766 | 25 · | |
| Single wall | 60.0% | -1.2 | 0.893 | 25 · | |
| Breakable half fence | 59.4% | -1.9 | 0.817 | 32 · | |
| Full fence | 54.8% | -6.4 | 0.435 | 31 · | |
| Rectangle | 50.0% | -11.2 | 0.149 | 34 · | |
| Fence and open | 43.8% | -17.5 | 0.138 | 16 · |
p < 0.05 on its ownStill there after correcting for how many were tested
Stage against stage
Dural's win rate on the row's stage, minus the same on the column'sAll ten shapes at once: chi-square 7.01 on 9 df, p 0.636
| Stage | Full | Break. full | Half | Break. half | Low | Wall | Mixed | Open | Rect | Oct |
|---|---|---|---|---|---|---|---|---|---|---|
| Full fence | 54.8%31 | -17.20.187 | -10.50.419 | -4.50.716 | -18.50.228 | -5.20.698 | +11.10.471 | -11.80.394 | +4.80.696 | -9.20.488 |
| Breakable full fence | +17.20.187 | 72.0%25 | +6.60.611 | +12.60.322 | -1.30.927 | +12.00.370 | +28.20.070 | +5.30.695 | +22.00.089 | +8.00.544 |
| Half fence | +10.50.419 | -6.60.611 | 65.4%26 | +6.00.639 | -8.00.598 | +5.40.691 | +21.60.169 | -1.30.927 | +15.40.233 | +1.40.918 |
| Breakable half fence | +4.50.716 | -12.60.322 | -6.00.639 | 59.4%32 | -13.90.353 | -0.60.962 | +15.60.306 | -7.30.592 | +9.40.445 | -4.60.722 |
| Low fence | +18.50.228 | +1.30.927 | +8.00.598 | +13.90.353 | 73.3%15 | +13.30.392 | +29.60.095 | +6.70.669 | +23.30.128 | +9.30.542 |
| Single wall | +5.20.698 | -12.00.370 | -5.40.691 | +0.60.962 | -13.30.392 | 60.0%25 | +16.20.309 | -6.70.641 | +10.00.446 | -4.00.771 |
| Fence and open | -11.10.471 | -28.20.070 | -21.60.169 | -15.60.306 | -29.60.095 | -16.20.309 | 43.8%16 | -22.90.163 | -6.20.680 | -20.20.202 |
| Open | +11.80.394 | -5.30.695 | +1.30.927 | +7.30.592 | -6.70.669 | +6.70.641 | +22.90.163 | 66.7%21 | +16.70.226 | +2.70.850 |
| Rectangle | -4.80.696 | -22.00.089 | -15.40.233 | -9.40.445 | -23.30.128 | -10.00.446 | +6.20.680 | -16.70.226 | 50.0%34 | -14.00.284 |
| Octagon | +9.20.488 | -8.00.544 | -1.40.918 | +4.60.722 | -9.30.542 | +4.00.771 | +20.20.202 | -2.70.850 | +14.00.284 | 64.0%25 |
Better on the row's stage — deeper at 5 and 10 pointsBetter on the column's stagep < 0.05 on its ownStill there after correcting for how many were tested· fewer than 80 decided matches on one side of the comparison
How significance is decided
- Each gap is two win rates compared with a pooled two-proportion z test, over decided matches only.
- The 95% interval is Newcombe's hybrid score method, which keeps its shape near 0% and 100% where a symmetric interval does not.
- q is p corrected for how many tests were asked at once, by Benjamini–Hochberg at a 5% false discovery rate. A figure is called real on q, never on p alone.
- Standard errors are widened by a measured design effect, because one player appears in many matches and those matches are not independent of each other.
- Stages are assigned at random, so a gap between them is the stage's own effect and not a difference in who chose to play where.
Matches counted7
Win rate42.9%
Widest swing57.1 pts
Survive correction0
Period2026/03/26 – 2026/09/23
Where it was fought
Dural's win rate on each stage, against this matchup everywhere| Stage | Stage effect | Win % | Change | p | Matches |
|---|---|---|---|---|---|
| Low fence | 100.0% | +57.1 | 0.212 | 1 · | |
| Single wall | 33.3% | -9.5 | 0.659 | 3 · | |
| Half fence | 0.0% | -42.9 | 0.350 | 1 · | |
| Octagon | 0.0% | -42.9 | 0.350 | 1 · |
p < 0.05 on its ownStill there after correcting for how many were tested
Stage against stage
Dural's win rate on the row's stage, minus the same on the column'sAll ten shapes at once: chi-square 3.00 on 3 df, p 0.392
| Stage | Half | Low | Wall | Oct |
|---|---|---|---|---|
| Half fence | 0.0%1 | -100.00.157 | -33.30.505 | +0.01.000 |
| Low fence | +100.00.157 | 100.0%1 | +66.70.248 | +100.00.157 |
| Single wall | +33.30.505 | -66.70.248 | 33.3%3 | +33.30.505 |
| Octagon | -0.01.000 | -100.00.157 | -33.30.505 | 0.0%1 |
Better on the row's stage — deeper at 5 and 10 pointsBetter on the column's stagep < 0.05 on its ownStill there after correcting for how many were tested· fewer than 80 decided matches on one side of the comparison
How significance is decided
- Each gap is two win rates compared with a pooled two-proportion z test, over decided matches only.
- The 95% interval is Newcombe's hybrid score method, which keeps its shape near 0% and 100% where a symmetric interval does not.
- q is p corrected for how many tests were asked at once, by Benjamini–Hochberg at a 5% false discovery rate. A figure is called real on q, never on p alone.
- Standard errors are widened by a measured design effect, because one player appears in many matches and those matches are not independent of each other.
- Stages are assigned at random, so a gap between them is the stage's own effect and not a difference in who chose to play where.