| Roshi vs King | 49–31 | 61.25% |
| Roshi vs Jin | 35–45 | 43.75% |
| Roshi vs Kazuya | 23–34 | 40.35% |
| Roshi vs Hwoarang | 24–25 | 48.98% |
| Roshi vs Steve | 28–21 | 57.14% |
| Roshi vs Reina | 25–24 | 51.02% |
| Roshi vs Bryan | 21–26 | 44.68% |
| Roshi vs Lee | 27–18 | 60.00% |
| Roshi vs Lili | 20–24 | 45.45% |
| Roshi vs Jun | 22–22 | 50.00% |
| Roshi vs Lars | 19–24 | 44.19% |
| Roshi vs Paul | 20–19 | 51.28% |
| Roshi vs Law | 19–20 | 48.72% |
| Roshi vs Yoshimitsu | 21–15 | 58.33% |
| Roshi vs Nina | 18–18 | 50.00% |
| Roshi vs Azucena | 13–23 | 36.11% |
| Roshi vs Victor | 21–15 | 58.33% |
| Roshi vs Dragunov | 19–16 | 54.29% |
| Roshi vs Eddy | 22–13 | 62.86% |
| Roshi vs Devil Jin | 10–18 | 35.71% |
| Roshi vs Leo | 15–12 | 55.56% |
| Roshi vs Xiaoyu | 6–20 | 23.08% |
| Roshi vs Lidia | 17–8 | 68.00% |
| Roshi vs Heihachi | 13–12 | 52.00% |
| Roshi vs Asuka | 12–12 | 50.00% |
| Roshi vs Jack-8 | 10–13 | 43.48% |
| Roshi vs Kuma | 15–8 | 65.22% |
| Roshi vs Clive | 12–9 | 57.14% |
| Roshi vs Alisa | 7–13 | 35.00% |
| Roshi vs Feng | 6–10 | 37.50% |
| Roshi vs Raven | 7–5 | 58.33% |
| Roshi vs Claudio | 8–2 | 80.00% |
| Roshi vs Panda | 4–5 | 44.44% |
| Roshi vs Shaheen | 4–4 | 50.00% |
| Roshi vs Leroy | 3–5 | 37.50% |
| Roshi vs Zafina | 2–4 | 33.33% |
| Roshi vs Anna | 1–2 | 33.33% |
Limitations
This data is often requested to give insight into which characters you have more trouble with than others, but it is not particularly helpful for that. The main issue is that it is heavily skewed by how strong the opponents you play are.
For example, this data suggests my worst matchup is clearly vs Reina, but that's just because most of those games are vs Yagami.
There is a way to account for this being worked on. The central idea is to assign each matchup a rating vs you which adjusts based on the result, much like the regular rating but also based on the rating of each player. With this, it would give a better summary of how well you perform vs each character.
In the meantime, this page is here to present the data as requested.