Why Engine Accuracy is a Waste of Time (at least for sub-2000 players)

Explore your chess personality across this link and see if you still think accuracy is or is not important.
https://chessiverse.com/chess-personality

The following is just part of what was generate after I typed my user name in the above link:
Embraces unorthodox and creative opening choices, launching early attacks with unusual piece placements and pawn storms that prioritize initiative over soundness. Thrives in chaotic positions with material imbalances and structural complications, combining dynamic piece play with tactical strikes and sacrificial ideas. Prefers to resolve games before reaching technical endgames through sharp middlegame play. When endgames occur, favors unbalanced structures with active piece play. Launches bold, aggressive offensives with sharp tactical strikes and material imbalances. Unafraid to sacrifice, constantly pushing for immediate concrete gains.

"The Mad Tactician" playing style: Chaotic attacker, speculative sacrifices, thrives in complications.
Model Player: Alexei Shirov
89% Match

So I searched for the FIDE player with that name, looked at his ratings and typed those ratings to AI with my ratings and profile including the 89% match to discover to know if I need to be concerned about accuracy at my present rating.

Gemini answered me and I extracted some comments to prove the point of accuracy from an AI point of view. Obviously you have to give real truthful numbers to get this type of response:
"""
In chess, "Accuracy" is the structural integrity.
Here is how you should view it: If your accuracy is... What it means for a "Mad Tactician"
Low (<70%) You are "Bluffing." You are sacrificing pieces and hoping they miss it. This works at 1500, but fails at 1900.
Moderate (75-85%) You are "Calculated Chaos." You are playing moves that might be slightly sub-optimal but are incredibly difficult to defend.
High (>90%) You are "Clinical." You are likely playing a dry game or your opponent collapsed early.
An 1842 FIDE Blitz performance is significant. In the world of OTB chess:
The "Accuracy Trap": At the 1800+ level, opponents aren't just waiting for you to hang a piece; they are looking for small positional concessions.
The Gap: Your 1600 Lichess rating is likely lower because you use those games as a laboratory. You aren't playing for "accuracy" there; you're playing for the "fire."
The Verdict: You don't need to be concerned about "perfect" accuracy, but you should be concerned about "Tactical Waste." With a 2301 peak puzzle rating, any game where you miss a winning tactical sequence (a "Missed Win") is a waste of your natural talent. You are currently in the "Intermediate-Advanced" bracket. To bridge the gap from 1800 to 2000 (Expert/Candidate Master level), you don't need to change your personality. You just need to ensure that your unorthodox openings lead to positions where your 2300-level tactics can actually bite. "Accuracy doesn't win games; making the last mistake lose." Your goal isn't to play the 1st choice of the engine. Your goal is to play a move that keeps your accuracy high enough to stay in the game, while keeping the complexity high enough to drown your opponent.
"""
I think I'm going to be looking at a few Alexei Shirov games.

I think showing inaccuracy is the best feature cuz it shows u lil things u overlooked in ur opening

@Toscani said in #20:

Think of a chess match as a debate. If one player makes sloppy, imprecise moves, the other naturally appears more elegant and convincing by comparison. Accuracy shows who is actually taking control of the conversation and who is merely guessing. Being below 2000 doesn't mean our games are less valuable or that our "arguments" on the board shouldn't be sharp. Growth comes through this pursuit of excellence, and while I used AI to help refine the "vocabulary" of this post, the sentiment remains the same: don’t let anyone tell you that accuracy is only for the elite.

The reason the accuracy evaluation is not a very practical use of time is for 4 reasons.

1. By looking at it it is a terrible blow to confidence to the average player that doesn't have a strong mental capacity for failure
2. The engine is going to have a better move that not even a Super GM will play that makes no sense practically to play for humans because it is impossible to calculate such a complicated move in some situations.
3. The engine doesn't know your play style or your weaknesses, just spits out raw data of the calculation from your games, so it doesn't actually tell you where to improve on
4. On that note of raw calculation, the engine expects every move to be perfect, so it grades a blunder that slips your evaluation from +9 to +1 because of an insane tactic that is impossible to see was missed just as bad of a blunder as a obvious hanging of a queen that changes the evaluation the same amount.

Also, the notion made that I am trying to get players to "stop striving for precision" is completely wrong. I state in my blog post that by getting a coach to help you improve, or using yourself to review your games, you find your weaknesses and learn how to work on them to become more precise. Lastly, when talking about self-review, I state that it is still a good idea to look at the analysis after you have analyzed first yourself to avoid missing something important and adding to overall precision. I never state that being below 2000 means our games are less valuable or that we shouldn't play well, I am stating why I believe there is a better method to reviewing our games to make us play better.

Hopefully this adds some clarification

correct me if im wrong:

what's those games examples for
all you say is how they "felt" but what's it gotta do with engine

the analysis link in the three dots or "X" (2nd diagram]) leads to a study... with no sf evaluation, not even game info

-------------------------~^º^~-------------------

i use sf evaluation to

• find bad blunders
• superficially check if my feeling of being winning in a distinct game phase was right
• to check if my feeling that there were not any blunders is right

you can go in deep from there.

it's an awe inflicting tool.

I don't see how it should be a waste of time if used with a good reason ( not automatically by principle ).

your private personal experience with "waiting a day" ( that's got nothing to do with stockfish ) isn't in any way an argument against the engine

you shouldn't have wrong expectations
in the first place

no one says you would learn or train or improve by requesting sf analysis ...

it's just an unbelievably powerful tool.

of course it's not "human moves" at depth 36+ halfmoves depth.
at 16 halfmoves depth is pretty much sufficient for a quick skipping through a game. ( still then it "sees" differently from humans )
these things are well known.

a review or evaluation of strengths and uselessnesses of stockfish, advantages vs disadvantages ... the whole picture would be more neutral, more helpful to take advice of
( than your distinct & narrow pov on 'accuracy value' only )

{ {
you meant good, invested time and work, you proudly published a blog, ...
but im thinking of all newbies and beginners relying on it, when i say: it's not exactly expert's advice, that one detail 'accuracy' only one single feature among many of sf, maybe not worth a whole long blog.
I by no means want to discourage you ( if it sounded like that ),
and maybe my critique is wrong and you'll get upvotes & followers ( you've got some already ), ...
just ... idk ... maybe talk to others first like what they think of your idea, before you put all the work and time in, and have what you think (any your personal pov) doublechecked.
so, do go on publishing! imo
} }

@TheCommandalorian said in #23:

By looking at it it is a terrible blow to confidence to the average player that doesn't have a strong mental capacity for failure

"The truth hurts!" - GM Finegold

@TheCommandalorian said in #23:

The engine is going to have a better move that not even a Super GM will play that makes no sense practically to play for humans because it is impossible to calculate such a complicated move in some situations.

Indeed. I've more or less given up on watching live chess commentary which doesn't seem to appreciate this. Some streamers do understand, though:
https://youtu.be/FHpavyKu7o0

I agree with one point: it is more important to understand the position (on your level) than just relying on computer analysis (when this is something which is not understandable from your point of view).
But the thing wih "waste of time" is somewhat missing the point since you can get excellent computer analysis in seconds. Nowadays. So the analysis itself is not a waste of time, but you should not just rely on it.

I usually add verbal comments to my computer analysis, it is also important to understand why you made some mistakes (were you nervous? were you short on time? were you overestimating your chances?). For us humans chess is also a psychological thing and the analysis of the computer cannot analyse this. This is something we have to do ourselves.

In books I noticed this also: some authors write excellent verbal comments (plus analysis) like J. Rowson. Others just write endless lines of moves, no text: B. Gelfand (often enough I just don't understand these moves and why he opted for them).

Accuracy is not a measure of playing strength or skill level. In engine analysis, accuracy reflects how closely a move aligns with an engine’s own preferred evaluation—not how difficult, practical, or appropriate that move is for a human player. Because PV lines are judged against an engine’s internal model rather than alternative engines or human decision-making, a move can be labeled “accurate” or “inaccurate” without regard to execution difficulty or practical success.

As a result, engine accuracy can misrepresent human play, particularly when multiple paths lead to the same outcome. Engines often treat the fastest or most precise line as uniquely correct, while slower, simpler, or more robust alternatives are penalized, even when they are strategically sound and more realistic under time pressure. In this sense, accuracy measures alignment with an engine’s chosen path, not the quality of human decision-making or relative skill.

Ultimately, an engine’s analysis is only stating that it would not have chosen that move. A more meaningful assessment requires a second opinion—either from a different engine or by adjusting the parameters of the same engine—to determine whether the so-called inaccuracy persists across evaluation models.

How I play a game is often influenced by the rating I see before starting. In this particular game, there are plenty of inaccuracies because of that. My approach tends to change depending on my opponent's rating. In this case, my main strategy was to capture all the pieces before delivering checkmate. Many players choose to resign when there’s no hope left, which can take the fun out of chess. When I’m clearly winning, time and even material value becomes less of a factor. That's when I tend to opt for more challenging moves rather than the best ones. This doesn’t make my play inaccurate, as it still requires careful calculation to avoid counterplay. Controlling material is enjoyable when ahead. If my opponent didn’t want to continue, they could have resigned.

I also think analyzing accuracy by phase would offer a more detailed assessment. Each phase—early, middle, and endgame—would have its own separate evaluation of accuracy, mistakes, and blunders. In a way, it’s like analyzing the game as if it were three separate games.