Every percentile on MindFeats is read from one of these published tables: never converted between feats, never an IQ.
Marks
Verified
Verified: public data, and proof the account is theirs.
Self-reported
Not verified yet: a public rating whose account isn't yet proven to be theirs; or self-reported, typed in (exams and competitions).
Played here
Played here: scored in a MindFeats game, kept in this browser. Reaction, Numbers, Arrows and Symbols are placed on published research norms, estimated; Doomsday has none yet.
Uploaded
Uploaded: read from a score report.
The marker: where the feat sits on its curve; the shaded part is everyone it is ahead of.
Combinations, estimated
A combination (“top 1% in chess and top 3% in cubing: about 1 in 400”) is estimated, until MindFeats has enough players to count it. Skills correlate, so multiplying the percentiles would overstate how rare a combination is; we never do.
Instead each feat's percentile becomes a point on a normal curve, and the feats are taken together as one bell-shaped spread where every pair correlates at 0.5 (a one-factor model). Published correlations between different cognitive skills are mostly lower: chess skill with fluid reasoning is about 0.24 (Burgoyne et al., Intelligence, 2016). Taking 0.5 for every pair errs toward calling a combination more common than it probably is.
Only the best feat of each kind is used: blitz and rapid are one skill and never make a combination. Each percentile comes from its own population (Lichess players, WCA competitors, research volunteers), and the estimate treats them as one; that is the main reason it says estimated. Shares are to two figures.
Competitions, counted
Each competition's own published counts. A percentile only where the counts rank the whole field a result came from; otherwise the count alone.
Percentile against the AMC 10 and 12 entries. AIME I 4,752 and AIME II 672, out of the fall 2025 AMC. A student who sits both the A and B contest counts twice among the entries, so the field of students is smaller and the percentile leans high: estimated. Read from AoPS wiki, AMC historical results (from MAA's reports) on 18 Nov 2026.
2025
92,661AMC 10 and 12 entries
6,208AIME takers (I and II)
Percentile against the AMC 10 and 12 entries. AIME I 4,690 and AIME II 1,518, out of the fall 2024 AMC. A student who sits both the A and B contest counts twice among the entries, so the field of students is smaller and the percentile leans high: estimated. Read from AoPS wiki, AMC historical results (from MAA's reports) on 18 Nov 2026.
USAJMOUSA Junior Mathematical Olympiad
2025
53,917AMC 10 entries
241USAJMO qualifiers
20Top Honors
39Honors
~81honourable mentions
Percentile against the AMC 10 entries. USAJMO qualifiers come from the AMC 10. MAA's list names only awardees who agreed to be listed, so the award counts are at least these. Qualifiers from the AoPS wiki. A student who sits both the A and B contest counts twice among the entries, so the field of students is smaller and the percentile leans high: estimated. Read from MAA, 2025 USAJMO awardees on 18 Nov 2026.
USAMOUSA Mathematical Olympiad
2025
38,744AMC 12 entries
293USAMO qualifiers
20gold awards
36silver awards
54bronze awards
50honourable mentions
Percentile against the AMC 12 entries. USAMO qualifiers come from the AMC 12. Awards counted from MAA's list; the 293 qualifiers from the AoPS wiki. A student who sits both the A and B contest counts twice among the entries, so the field of students is smaller and the percentile leans high: estimated. Read from MAA, 2025 USAMO awardees on 18 Nov 2026.
Percentile against the US Open contestants. Divisions in the 2025 US Open, the season's last contest; a division reached is compared with where its contestants competed, so it is estimated. The 2025–26 season's counts aren't published yet. Read from USACO, 2025 US Open results on 18 Nov 2026.
IOIInternational Olympiad in Informatics
2025
330contestants from 84 countries
28gold medallists
55silver medallists
82bronze medallists
36honourable mentions
A count only, no percentile. Read from IOI statistics on 18 Nov 2026.
2024
362contestants from 91 countries
34gold medallists
62silver medallists
93bronze medallists
42honourable mentions
A count only, no percentile. Read from IOI statistics on 18 Nov 2026.
IPhOInternational Physics Olympiad
2025
~415contestants from 87 countries
37gold medallists
81silver medallists
96bronze medallists
71honourable mentions
A count only, no percentile. No official total is published; press reports give 406 to 440 contestants. Read from IPhO unofficial statistics on 18 Nov 2026.
IChOInternational Chemistry Olympiad
2025
~360students from 96 countries
30gold medallists
73silver medallists
107bronze medallists
30certificates of merit
A count only, no percentile. The host says “over 360 students”; Singapore's ministry says 354 from 92 countries. The medal counts agree. Read from UAE Ministry of Education (host) on 18 Nov 2026.
IBOInternational Biology Olympiad
2025
298students from 78 countries
30gold medallists
60silver medallists
89bronze medallists
30certificates of merit
A count only, no percentile. Read from IBO 2025 ranking on 18 Nov 2026.
Spelling BeeScripps National Spelling Bee
2026
247spellers at nationals
1champion
A count only, no percentile. Counts by round reached aren't published. Read from Scripps on 18 Nov 2026.
2025
243spellers at nationals
1champion
A count only, no percentile. Read from Scripps on 18 Nov 2026.
Tables
Colours group feats by family: blues for strategy games, oranges for speed-solving, greens for contests, reds, a shade per ability for the mind tests here, slate for exams; within a family, darker and lighter shades tell its kinds apart.
Lichess bullet
n = 394,790
1,0001,5002,0002,500
Lichess bullet players active this week (394,790). Read from Lichess on 18 Nov 2026.
Lichess blitz
n = 759,535
5001,0001,5002,0002,500
Lichess blitz players active this week (759,535). Read from Lichess on 18 Nov 2026.
Lichess rapid
n = 579,863
5001,0001,5002,000
Lichess rapid players active this week (579,863). Read from Lichess on 18 Nov 2026.
Lichess classical
n = 47,503
1,0001,5002,000
Lichess classical players active this week (47,503). Read from Lichess on 18 Nov 2026.
Chess.com bullet
n = 12,361,821
5001,0001,5002,0002,500
Chess.com bullet players (12.4 million). Read from Chess.com leaderboard on 18 Nov 2026. Estimated: the leaderboard counts only part of Chess.com's players.
Chess.com blitz
n = 21,564,650
5001,0001,5002,000
Chess.com blitz players (21.6 million). Read from Chess.com leaderboard on 18 Nov 2026. Estimated: the leaderboard counts only part of Chess.com's players.
Chess.com rapid
n = 50,490,415
5001,0001,5002,000
Chess.com rapid players (50.5 million). Read from Chess.com leaderboard on 18 Nov 2026. Estimated: the leaderboard counts only part of Chess.com's players.
Chess.com daily
n = 3,377,201
5001,0001,500
Chess.com daily players (3.4 million). Read from Chess.com leaderboard on 18 Nov 2026. Estimated: the leaderboard counts only part of Chess.com's players.
3x3x3 Cube average
n = 285,541
20 s40 s1:001:20
WCA competitors with an official 3x3x3 Cube average (285,541). Read from WCA results on 18 Nov 2026. Faster to the right.
3x3x3 Cube single
n = 290,926
10 s20 s30 s40 s50 s1:001:10
WCA competitors with an official 3x3x3 Cube single (290,926). Read from WCA results on 18 Nov 2026. Faster to the right.
Codeforces
n = 42,040
5001,0001,5002,0002,500
Codeforces users rated in the last month (42,040). Read from Codeforces API on 18 Nov 2026. New accounts start near 0 and climb over their first contests, hence the pile at the left.
LeetCode contest
n = 892,590
1,2001,4001,6001,8002,0002,2002,400
LeetCode users in its global contest ranking (892,590). Read from LeetCode global ranking on 18 Nov 2026. Drawn from real users at a grid of ranks; a connected account shows LeetCode's own top-percentage for it.
AtCoder
n = 125,538
5001,0001,5002,0002,500
AtCoder users rated in the last two years (125,538). Read from AtCoder ranking on 18 Nov 2026. AtCoder shows lower ratings for players with few contests, hence the pile at the left.
SAT
6008001,0001,2001,400
SAT takers from the past three graduating classes. Read from College Board on 18 Nov 2026.
GRE General Test takers who tested between July 1, 2022 and June 30, 2025 (ETS, all test takers) (788,021). Read from ETS on 18 Nov 2026. ETS counts the share scoring lower than the score.
GRE Quantitative Reasoning
n = 790,864
135140145150155160165
GRE General Test takers who tested between July 1, 2022 and June 30, 2025 (ETS, all test takers) (790,864). Read from ETS on 18 Nov 2026. ETS counts the share scoring lower than the score.
GRE Analytical Writing
n = 785,720
12345
GRE General Test takers who tested between July 1, 2022 and June 30, 2025 (ETS, all test takers) (785,720). Read from ETS on 18 Nov 2026. ETS counts the share scoring lower than the score.
GMAT Focus Edition
300400500600700
GMAT test takers, July 1, 2021 to June 30, 2026. Read from GMAC on 18 Nov 2026. From GMAC's score concordance table, which prints each Focus Edition total's percentile to a tenth. A classic (10th Edition) score is placed through the same table: of the Focus totals GMAC links it to, the lowest percentile, marked estimated.
LSAT
130140150160170
LSAT scores from the 2023-2026 testing years. Read from LSAC on 18 Nov 2026. LSAC counts test scores lower than the score, not people: someone who sat twice counts twice.
MCAT
n = 293,882
480490500510520
MCAT scores from the 2022 to 2024 testing years (293,882). Read from AAMC on 18 Nov 2026. AAMC counts the share scoring at or below the score. This is the table in effect 1 May 2025 to 30 April 2026; AAMC's newer one sits behind a bot check our fetcher can't pass, and replaces it once read.
Forward digit span (Numbers)
n = 9,858
246810
People tested online by TestMyBrain (9,858). Read from Singh et al. 2021 (TestMyBrain) on 18 Nov 2026. Estimated: the paper gives only a mean (6.43 digits) and SD (1.58), so the curve is a normal one. Its test shows digits one at a time and gives two tries at each length; Numbers shows the whole number at once and ends at the first miss, which if anything scores a person lower. Lab norms with spoken digits (French WAIS-R, Grégoire and Van der Linden 1997) put adults near 6 digits too.
Simple reaction time (Reaction)
n = 49,668
150200250300350400450
People tested online by TestMyBrain (49,668). Read from Singh et al. 2021 (TestMyBrain) on 18 Nov 2026. Estimated: the paper gives only a mean (303 ms) and SD (63 ms) of each person's median, so the curve is a normal one (real reaction times lean to the slow side). Its test is the same kind as ours, a tap when a green square appears; Reaction uses five tries, and screen and touch delay vary by device. Faster to the right.
Choice reaction time (Arrows)
n = 17,121
2004006008001,0001,2001,4001,600
People tested online by TestMyBrain (17,121). Read from Singh et al. 2021 (TestMyBrain) on 18 Nov 2026. Estimated: Table 3, Choice RT correct-trial RT, gives only a mean (896 ms) and SD (257 ms) of each person's median time on right answers, so the curve is a normal one. Same rule as ours: name the direction of the arrow whose colour differs. The paper doesn't say how many arrows, trials or keys; Arrows uses three arrows, 20 trials, ← and → or a tap on each side, and counts a run only with 16 or more right. Theirs takes about 2.5 minutes, ours about 30 seconds, and keys answer faster than touch buttons. Faster to the right.
Digit symbol matching (Symbols)
n = 36,000
20304050607080
People tested online by TestMyBrain (36,000). Read from Singh et al. 2021 (TestMyBrain) on 18 Nov 2026. Estimated: Table 3, Digit Symbol Matching score, gives only a mean (51.9 right) and SD (12.2), so the curve is a normal one. Same layout and rule as ours, nine symbols, three for each of 1, 2 and 3, the key always on screen, 90 seconds; our symbols are our own drawings, the pairing is shuffled each run, and a run counts only at 80% right or more. Pressing a key can be faster than tapping.