Feature request: add position-based move results

[nrouzaud38]

Hi 👋

Opening Explorer currently exposes branch-based move results (W/D/L) depending on the move history (play). While useful, under transpositions the same resulting position can show different results for the same move, which can be confusing.

It would be very helpful to include position-based stats in the move results currently branch-based, so that all relevant results are available in a single request.

Example

1. GET /masters?variant=standard&play=d2d4,g7g6,b1c3,f8g7
   → move "e4": branchWhite 29, branchDraws 27, branchBlack 38 → white 30%

2. GET /masters?variant=standard&play=d2d4,g7g6,e2e4,f8g7
   → move "Nc3": branchWhite 8775, branchDraws 7493, branchBlack 6930 → white 38%

3. GET /masters?variant=standard&
   fen=rnbqk1nr/ppppppbp/6p1/8/3PP3/2N5/PPP2PPP/R1BQKBNR b KQkq - 0 3
   → position results: posWhite 8867, posDraws 7581, posBlack 7034 → white 38%

The result of request 3 (position stats) could be included in the details of requests 1 and 2 (branch-based moves).

Proposal

Add position-based W/D/L fields for each move (existing fields remain unchanged)

This ensures consistent, transposition-invariant statistics for all moves in a single request, which will be very useful for users.

Reference
Forum discussion:
https://lichess.org/forum/lichess-feedback/incoherent-opening-statistics#1

Thanks for considering this 🙂

[niklasf]

Hi. The lookup is already by position and includes transpositions. The history from play is only used to obtain an opening name if there is no perfect match.

However the counters on moves intentionally show how often the particular move was played in the position (regardless of history), not how often the resulting position is reached.

Think about a blunder. The implemented approach will likely have the move that punishes the blunder as the top ranking move. Counting transpositions on resulting positions will likely have the move that ignores the blunder and returns to a common position.

[nrouzaud38]

Hello @niklasf,

The current counters (for the moves from this position) are useful for ranking the moves and should not be changed. Additional counters on the resulting positions (including all transpositions) could be used for the results, without having to navigate to each position.

It would be more convenient and logical if the highlighted results were the same (38%). Don't you think so ?

[niklasf]

Ah, I get the idea now. Can at least be considered for a future version, but not easy to retro-fit to the current data model.