No folder selected — select the folder of .rgn files
Chrome or Edge required. Select the output folder from the corpus generator. All .rgn files are read-only.
Analysis depth (plies)12 plies
How many half-moves to record per game. Ply 12 = 6 full moves. Beyond ply 10–12 positions become sparse with 50K games; increase minimum N or reduce depth if coverage is too thin at the tail.
Minimum sample Npositions with fewer games are excluded from analysis
Analysis progress
Files processed
0
of —
Rate / ETA
—
ETA —
Positions / Analyzed
—
analyzed: —
Enumerating files…
Results
Games parsed
—
— errors
Positions analyzed
—
of — total
Unique 1st moves
—
seen for White at ply 1
Game visits per ply — White / Black
Notable positions — click a line to open it in the explorer below
Line
Mover
Games
Tree explorer
Move
N
W
D
B
EV%
No moves meet the minimum sample size at this position — corpus coverage ends here.
Export
The JSON file lists every analyzed position with its qualifying moves and stats, keyed by a plain position string (side to move + board contents) — no engine-specific hashing involved. The Markdown report walks the most-played line move by move, with a full branch table at every step; paste it directly into the manuscript.
How it works: Each .rgn file is replayed move-by-move through a minimal board (no search, no evaluation). At each half-move up to the configured depth, the position is recorded together with the move played and the eventual game result. Positions are keyed by side-to-move plus board contents, so transpositions reached by different move orders are correctly merged into the same node.
Notable Positions: There are three categories of notable positions: (1) the most-visited line (the de facto main line of the corpus), (2) positions where even the best move’s expected value sits closest to 50% (genuinely unresolved theory), and (3) positions with the widest expected-value gap between the best and next-best move (where the choice matters most). Click any line to open it in the explorer.
Recommended settings for 50K games: analysis depth 12–14 plies, minimum N = 20–40. Solid coverage is typically found through ply 8–10; the tail thins quickly beyond that. Reduce minimum N with caution — noisy low-sample entries are not reliable evidence of anything.