Beyond 0.00: What the Engine Evaluation Does Not Tell Us
Introduction
"0.00" For the modern correspondence chess player, few numbers are more reassuring—and more deceptive—than this one.
Powerful engines have transformed correspondence chess. Stockfish, Leela Chess Zero and other analytical tools can calculate variations to extraordinary depths, while opening databases and tablebases provide resources that earlier generations could not have imagined.
Yet a numerical evaluation is only a summary. It does not tell us everything about the position, and certainly does not tell us everything about the thought process of the player sitting behind the computer.
In three of my own correspondence games, I found myself making decisions for reasons that were not simply "the engine says so". In each case, the position remained objectively playable, but my human objective was different: restrict a piece, choose the right moment for a breakthrough, or transform the position into one that imposed a different problem on my opponent.
These games led me to a question that I find increasingly interesting:
What does the engine evaluation leave out?
Part 1: 24.b5! — Restricting a Piece
Mohanty, Om Prakash – Rodriguez Rodriguez, Mariano Danilo
Tournament No. 1522
The move I want to highlight is 24.b5! (see the game below)
Stockfish preferred 24.a4, and Leela Chess Zero also eventually settled on a4. I nevertheless chose 24.b5 manually. My reason was quite specific: Black's knight on b8 had been out of the game for a long time, and I did not want to make it easy for Black to reorganize and bring that knight back into play.
This is not a claim that the engines "missed" the move. The important point is different. The evaluation remained close to 0.00, while my human evaluation of the position was based on a concrete strategic feature: piece activity.
The parameters behind 24.b5 were therefore not simply space:
- Restriction of piece activity — the knight on b8 was already badly placed, and I wanted to preserve that inconvenience
- Limitation of reorganization — changing the queenside structure reduced the opponent's freedom to find an easy route for the knight
- Reduction of counterplay — Black was forced to react rather than simply improve the misplaced piece
- Strategic economy — I did not need an immediate tactical gain; preventing the opponent's improvement was itself valuable
This illustrates an important limitation of looking only at the displayed engine number. Two moves can both remain near equality while representing different strategic choices. The engine gives an objective evaluation; the human player may be choosing the type of position he wants to reach.
Human parameter: RESTRICTION OF PIECE ACTIVITY
Part 2: 45.f5!! — Recognising the Moment
Mohanty, Om Prakash – Cruz Ramirez, Eliecer
Tournament No. 1522
Here the move that stands out to me is 45.f5! (see the game below)
The interesting point is that 45.f5 cannot be understood in isolation. It was the moment at which a plan that had been developing for many moves became concrete.
The engine can calculate the consequences of a pawn break. But the human player also has to recognise when the position is ripe for that break.
For me, the relevant parameter was timing: not merely "can I attack?", but "is this the moment to open another front?"
Human parameter: TIMING OF THE BREAKTHROUGH
Part 3: 29.c6! — Transforming the Position
Mohanty, Om Prakash – Vyas, Nileshkumar J.
Tournament No. 1520
This game is older than the first two, but it provides a different example of human strategic thinking.
The move 29.c6! changes the nature of the position. The pawn advances deep into Black's territory and becomes a persistent positional factor. Black's pieces must now take the pawn and the resulting restrictions into account, while White can improve the remaining pieces. (see game below)
There is no spectacular combination in the final position. The resignation reflects the cumulative effect of the position: White's advanced c-pawn, active pieces and control of key squares have made Black's defensive task extremely difficult.
For me, this game illustrates a different human parameter from the first two. Sometimes the most important move is the one that changes the problem the opponent has to solve.
Human parameter: TRANSFORMATION OF THE POSITION
What the Number Does Not Tell Us
These three games are not intended to demonstrate that humans can simply “see what engines cannot”. That would be an unfair and technically weak conclusion. Modern engines evaluate piece activity, pawn structures, king safety, space and many other features far more accurately than a human can.
The interesting issue is the interface between the numerical evaluation and human decision-making.
A displayed 0.00 does not tell me why I should prefer one equal position over another. A +0.50 does not tell me whether the advantage is easy to exploit or requires extraordinary precision. And a move that is not the engine's first choice may still be perfectly compatible with the objective evaluation while serving a very clear human plan.
Correspondence chess therefore creates an unusual form of human-machine cooperation.
The machine supplies calculation, comparison and verification. The human supplies questions, priorities and strategic objectives.
The strongest workflow is not necessarily: "Let the engine choose the move."
It may instead be: "What do I want to achieve? Which features of the position matter? What candidate moves realise that idea? Now, what does the engine say about them?"
The Correspondence Player as an Analyst
This changes the meaning of engine assistance.
When I played 24.b5, my question was about the activity of a misplaced knight.
When I played 45.f5, my question was about timing an attacking breakthrough.
When I played 29.c6, my question was whether I could transform a positional advantage into a lasting structural problem.
In each case, the engine remains essential. It can test the idea, expose tactical refutations and tell us whether our plan survives concrete calculation. But the initial human question is still valuable.
Perhaps this is one of the distinctive skills of modern correspondence chess: not calculating more deeply than the machine, but knowing what deserves to be calculated deeply.
Conclusion
The number 0.00 is useful. It is precise within the limits of the model and search, and it is indispensable to modern correspondence analysis.But it is not the whole story.
A correspondence player also sees pieces that are badly placed, squares that may disappear, pawn breaks that are approaching, kings that are becoming vulnerable and positions that are gradually changing character.
In my three games, the key decisions were:
- 24.b5! — restrict the opponent's piece activity
- 45.f5!! — recognise the moment for the breakthrough
- 29.c6! — transform the position into a different strategic problem
Perhaps, then, when the engine shows 0.00, the correct reaction is not to stop thinking.
It is to ask a better question:
"What is the position trying to become?"
The engine can help us calculate the answer.
But the question is still ours.
There is another source of human knowledge that is difficult to express as an engine evaluation: experience from the opposite side of the position. I regularly play the King's Indian, Hedgehog and Pirc structures as Black. Part of the reason is to avoid turning correspondence chess into a simple exercise in reproducing opening-book and engine moves. By playing these systems repeatedly from Black's side, I have learned where Black's pieces can become passive, which pawn breaks are difficult to achieve, which exchanges relieve Black's problems, and which small restrictions can become permanent. When I later encounter similar structures as White, that experience becomes a practical analytical tool. In the first game, 24.b5 was one such decision. I was not trying to out-calculate the engine; I was recognizing a familiar defensive problem from the opposite side.