The Viveka Method · v3.0
Where to look. Then what to play.
Viveka (विवेक) is Sanskrit for discernment. Split the board into four quadrants. One small equation tells you where the action is. Your intuition proposes two or three moves there. A second small equation picks one. Each equation has three variables and one or two constants, all fitted to Stockfish 19 with the best move judged at depth 22.
1 · Find the quadrant
Φ = U + 1.5·ln(1 + N)
- U, material at stake: enemy pieces there you can win, plus your pieces there they can win (in pawns, 1/3/3/5/9).
- N, your options: legal moves that land there. 1.5·ln(1 + N) is about one point per doubling: N = 1, 3, 7, 15 → 1, 2, 3, 4.
Quadrants are seen from your side: your half or theirs, queenside (a–d) or kingside (e–h). Examine the top two.
2 · Let intuition propose
Look first at moves that land in those two quadrants (on average 24 of 32 legal moves; the top quadrant alone has 13). Pick the two or three you like. This is the step the method does not replace: the better your candidates, the better the result (table below).
3 · Score the candidates
S = M + 1.25·K + 0.25·C
- M: pawns the move wins after the exchange on its square, minus the most they can then win from you.
- K: 1 if it gives check. C: rings the piece moves towards the centre (−3 to +3).
Play the highest S. Material first; a check is worth about a pawn; centralising breaks ties.
Your learning path
Understand it on worked examples, try it on any position, then make it automatic.
- 1
Learn
14 worked examples, nine steps each. The board shows every number as you meet it, and you predict before each reveal.
Start lesson 1 → - 2
Explore
Load any position in the calculator below. Switch the board layers on and off and score your own candidates.
Open the calculator → - 3
Practise
17,000+ positions with every move rated by Stockfish. Guided mode checks each term; Independent mode only the result.
Start practising →
Worked examples: one idea each
An enemy piece is loose. Watch how it lights up its quadrant, and how M pays for taking it.
One of your pieces can be won. Danger counts towards U exactly like opportunity does.
Nothing can be won by either side. The equations still give a clear answer.
One candidate gives check. How much is a check worth?
An opening position. Most best moves in the opening land in your own half.
An endgame with eight or more pieces. The same two equations apply.
An honest example: here the method chooses wrongly. Find out why.
Work a position
Pick a test position or paste any FEN. The calculator ranks the quadrants, lets you choose candidates, and scores them term by term. For the test positions you can reveal Stockfish's verdict. Everything runs in your browser.
Black to move. Play a move on the board to add it as a candidate (the position does not change). Lettered arrows: your candidates; the solid one scores highest.
U = enemy material you can win there + your material they can win there (pawns). N = your legal moves landing there. 1.5·ln(1 + N) is about one point per doubling: N = 1 → 1, 3 → 2, 7 → 3, 15 → 4, 31 → 5.
| Quadrant | Loose | Hanging | U | N | Φ |
|---|---|---|---|---|---|
| Q1 · Own half · queenside (a–d) | 0 | 0 | 0 | 16 | 4.2 |
| Q2 · Own half · kingside (e–h) | 0 | 0 | 0 | 11 | 3.7 |
| Q3 · Enemy half · queenside (a–d) | 0 | 0 | 0 | 3 | 2.1 |
| Q4 · Enemy half · kingside (e–h) | 0 | 0 | 0 | 2 | 1.6 |
Pick two or three moves you like that land in the shaded quadrants: tap them below or play them on the board. 27 legal moves land there.
M = pawns the move wins after the exchange on its square, minus the most they can then win from you. K = 1 if it gives check. C = rings the piece moves towards the centre (castling counts 0). Play the highest S.
No candidates yet.
How good is it? It depends on your candidates
Average loss against Stockfish's depth-22 best move on 5,587 positions from held-out openings, choosing among three candidates with S. Lower is better; our target was 30 cp or less.
| Intuition | Top quadrant | Top 2 | Top 3 | All 4 |
|---|---|---|---|---|
| Strong (your three are the region's three best moves) | 71 | 36 | 26 | 22 |
| Good (the region's best move plus two other sensible ones) | 83 | 47 | 37 | 33 |
| Club (three of the region's eight best, best move not guaranteed) | 137 | 103 | 87 | 81 |
For scale: a random legal move loses 359 cp, and Stockfish searching four plies 28 cp. With the same strong candidates in the top two quadrants, a one-ply Stockfish search would lose 28 cp and the best of the three 14 cp.
Live validation from the trainer
Every move played in the trainer is checked against what the method would have played, and both against Stockfish. These are the running results, next to the study's figures.
Collecting: 20 positions recorded so far. Figures appear here once 30 distinct positions have been played.
What it is
- Two equations you can do at the board, with every constant measured against Stockfish rather than asserted.
- A way to narrow the board: the top two quadrants hold the engine's best move 78% of the time.
- A tie-breaker for good candidates: with strong candidates in three quadrants it reaches 26 cp, inside the 30 cp target.
What it is not
- Not a substitute for candidate-finding: with club-level candidates the loss is 103 cp in two quadrants.
- Not uniform: middlegames are the hardest (62 cp against 29 cp in openings, two quadrants, good intuition).
- Not a promise of rating points. We publish what it does on test positions.
Make it automatic
Practice positions: find the quadrant, choose your candidates, score them, and see Stockfish's verdict on every one. 10 positions a day free, unlimited with Pro, plus live Stockfish, game reviews and the AI coach.
Earlier versions: Viveka v2 (scan, score, check; archived) · Viveka v1 (archived) · the book (2nd edition, v2 method)