VirtuAnalytics

Novel Insights Through Data & Analysis

Virtua Fighter 5 REVO · v1.10 · Region: North America

Pai

SWAT1107

Career matches4,902
Career win rate49.6%
Round win rate50.7%
Peak rank30
Distinct opponents93
Period2026/03/26 – 2026/10/18

Elo rating

Overall, and on each character played 10+ times

1547 #1,686 of 3,686 on the Elo list →

The overall rating moves on every match, whatever was picked; a character's rating adds this player's own adjustment on it. Both run from every recorded match since v1.10; how they are worked out is on the Elo list.

Characters played — VF5's own record

Pai303,2571,6311,62650.1%
Jean—1,64580084548.6%

These are the totals VF5 keeps for each character, read from the newest replay stored for it. They cover the player's whole ranked career and take no period.

Inputs per second

1,000 of 1,471 rounds counted

4.7 inputs a second (282 a minute) this season. Faster than 1% of the 4,069 regulars with enough rounds counted to compare.

By character played

Pai4.72821,000

By opponent's character

Sarah4.8285144
Akira4.9292103
Jacky4.426665
Shun4.627491
Brad4.426668
Jean4.627661
Taka4.527150
Pai4.829034
Jeffry4.627465
Lei Fei4.728148
El Blaze4.728138
Lau5.030045
Goh4.828554
Wolf5.130826
Aoi4.728239
Eileen5.130611
Kage4.728429
Vanessa4.828529

An input is any change on the controller (a new stick direction, a button pressed or released) from the bell to the moment a round is decided, over the live time of the round. Only rounds whose inputs have been decoded are counted. Who you face changes the pace too: see the inputs-per-second article under News.

Against each character

Sarah46252154.4%+3.3
Akira43281565.1%+13.7
Jacky37162143.2%-7.0
Shun28131546.4%-3.3
Brad2662023.1%-25.9
Jean2518772.0%+16.4
Taka2491537.5%-9.7
Pai22111150.0%-0.5
Jeffry1910952.6%-2.2
Lei Fei189950.0%-2.1
El Blaze146842.9%-11.0
Lau147750.0%-3.3
Goh134930.8%+0.6
Wolf106460.0%+6.3
Aoi97277.8%+39.1
Eileen83537.5%-17.7
Kage85362.5%+19.2
Vanessa86275.0%+17.7
Lion42250.0%-14.5
Dural2020.0%-50.6

0 of 1 rows differ from what this player's level and their opponents' ranks predict, once the test allows for asking 1 questions at once; about 0.1 would by chance alone.

A matchup row is the particular people played on that character, not the character in general: one strong regular moves it as much as the matchup does.

By ring shape

Low Fence48291960.4%+4.9
Open44192543.2%-7.3
Half Fence40172342.5%-4.3
Octagon40192147.5%+3.6
Rectangle37201754.0%+0.2
Single Wall34122235.3%-18.9
Breakable Full Fence33181554.5%+2.2
Breakable Half Fence32181456.2%+10.8
Full Fence31191261.3%+4.5
Full Fence and Open2213959.1%+6.2

0 of 9 rows differ from what this player's level and their opponents' ranks predict, once the test allows for asking 9 questions at once; about 0.5 would by chance alone.

With or without walls

StagesMatchesWin %Expectedvs expected
Walls21352.1%49.8%+2.3
No walls12647.6%52.9%-5.3

Walls against open: +7.6 points (95% range -3.9 to +19.1), p 0.198 — no measurable difference.

The fence-and-open stages and Training Room are on neither side, as on the stage report.

How these are tested
  • Each match gets an expected result from the two players' ranks, read off the ladder's rank-against-rank win rates, and shifted by one number per player so the expected wins add up to the actual ones. A row's gap is how far above that expectation it sits, against the same measure over the player's other matches.
  • Matches against the same opponent are not independent, so each test is widened by how unevenly this player's results fall across the people they played. Rows under 30 decided matches or 10 different opponents on either side carry no test, and each table is corrected for the number of rows tested in it.
  • Not adjusted for: how strong an opponent is for their rank. Rank carries most of it; the rest needs a rating model.

Most played opponents

Most played opponents, most recent first

samtheseed (Rank 32)2026/09/1351420.0%
rich.tapestry (Rank 29)2026/08/30660100.0%
Da-Goat-Pawn (Rank 32)2026/08/301431121.4%
DDRMONKEY (Rank 26)2026/08/0351420.0%
ulilb1ckard (Rank 30)2026/08/02125741.7%
Exodus6162 (Rank 27)2026/08/0252340.0%
Based_Ganja (Rank 30)2026/07/2764266.7%
TayoTayoTaka (Rank 29)2026/07/2752340.0%
seedash (Rank 32)2026/07/2782625.0%
DatDemo (Rank 28)2026/07/20104640.0%
Jiikae (Rank 32)2026/07/19107370.0%
LuBuMBX (Rank 29)2026/07/1963350.0%
UltraShock5 (Rank 29)2026/07/1864266.7%
gainkiller (Rank 27)2026/07/1364266.7%
own1jef123 (Rank 29)2026/07/031431121.4%
Kingbison (Rank 28)2026/06/22129375.0%
sideswipe75x (Rank 25)2026/06/2286275.0%
leefung._ (Rank 30)2026/06/1951420.0%
RetroTobyCEO (Rank 29)2026/05/2671614.3%
bighorny (Rank 29)2026/05/251910952.6%
ChrmnMa0 (Rank 28)2026/05/2475271.4%
SolidSnek (Rank 39)2026/05/1852340.0%
fivey70 (Rank 27)2026/05/17117463.6%
TomSavage (Rank 30)2026/05/1062433.3%
MisterDingle (Rank 28)2026/05/0586275.0%

Round openings

Pai

Career
1,631–1,626 · 50.1%
This season
378 matches · 50.5%
Peak rank
30

First hit in 51.2% of 1,467 rounds. Won 54.5% of the rounds they struck first, 47.1% of the rounds they were struck first.

Significant Tells

  • In round 1: backs off first in 10.8% of rounds, against 6.7% in their other rounds.
  • Against Akira: evades first in 40.1% of rounds, against 22.9% against everyone else.
  • In round 2: opens with “P punch into crouch” in 15.4% of rounds, against 8.6% in their other rounds.
  • After opening the round before with “P punch”: opens with “evade up” in 13.0% of rounds, against 5.3% after anything else.
  • After winning a round: opens with “backdash” in 5.2% of rounds, against 1.9% after losing one.

How is this tested?

punchP into crouch15310.4%
punchP15110.2%
evade down1077.2%
evade up1047.0%
backdash674.5%
Souanshou3P+K664.5%
Ensei Haijinkyaku7K+G583.9%
evade up into Enbikyaku3K+G533.6%
Meishouho4P+K+G into punchP513.5%
evade down into Enbikyaku3K+G432.9%
guardG392.6%
crouch into Renka Sentai2KK362.4%
crouch into Souanshou3P+K292.0%
Kochoukyaku4K+G231.6%
Hi'en Tankyaku9K into Hi'en Rekkyaku9KK211.4%

142 further openings account for the other 475 rounds.

By stage type

Record against each character on walled and open stages, as Pai

OpponentWallsNo wallsWalls vs open, rank-adjusted, pts
Sarah11–8 · 58%11–10 · 52%too few to test
Akira18–8 · 69%8–7 · 53%too few to test
Jacky6–13 · 32%7–4 · 64%too few to test
Shun8–12 · 40%3–1 · 75%too few to test
Brad3–14 · 18%3–6 · 33%too few to test
Jean12–4 · 75%5–1 · 83%too few to test
Taka7–6 · 54%2–9 · 18%too few to test
Pai8–5 · 62%2–4 · 33%too few to test

Walls: every fence, rectangle and octagon stage. No walls: open, low fence and single-wall stages. Snow Mountain and Training Room are left out.

How is this tested?

How these are tested

A tell is a habit in how this player starts a round: something they do clearly more often in one situation than in the others — backing off more in round 2, say, or opening with one move more against one character. Each situation is compared with how they actually opened their rounds, and a tell is listed only if it is too strong to be a fluke even after allowing for how many situations were checked on this page. Anything weaker is left out, so a character with no tells listed may still have habits, just none the data can show for certain.

  • An opener is filed by what the player did first: attack (a named move or any button), move in (forward dash, forward movement or crouch dash forward), wait (guard, crouch or no input), back off (backdash, crouch dash back or backward movement) or evade. Each is also tested as the exact opener: the character's five most used, and the rest together.
  • Up to ten things known before a round starts are tested against the opener: which round it is, the opponent's character, the ring shape, how the player opened the round before and how the opponent did, whether they won or lost the round before, who landed the first hit of the round before, whether they go into it ahead, level or behind, and -- where a Shun is playing -- the drink points he starts it on: the player's own if they are Shun, the opponent's if they face one. Each is a chi-square test, divided by how much more alike the rounds of one match are than separate rounds would be, and never made more generous by it. Checked by shuffling the data so that nothing is there: the tests then find something no more often than chance allows.
  • Round results and drink points come from each match's round record, matched to its openers by the two players and the date. A match whose record is missing, lacks rounds the openers have, or does not add up to the match's result sits out of those four tests only. These four are compared only between rounds of the same number -- the score and the drinks follow from which round it is (every match starts level and on no drinks), so a pattern that is only “the first round is different” is not counted as one of theirs. Drawn rounds count toward neither side of the score, and the round after one has no previous result. Drink points are grouped as on the Shun report (none, 1–4, 5–9 and so on), with neighbouring groups joined where they are thin.
  • A character is tested with at least 300 rounds from 30 matches. Rare cases are pooled before testing, and a pattern is only named where it holds up after correcting for every test on this page (15 here; about 0.8 would come up by chance alone before the correction).
  • Walls against open is tested per opponent with 30+ matches and 10+ different players on each side, allowing for both players' ranks, and corrected across every player page at once: 0 of 2,552 hold up. With stages shuffled at random: 0 of 2,651–2,673.

Tests behind each character's list

  • Pai: 5 of 15 tests hold up