Understanding chess engine ratings - what does the number mean?

Chess engine evaluations like Stockfish or AlphaZero express their assessment in centipawns - hundredths of a pawn. An evaluation of +100 means White has roughly a one-pawn advantage, while -50 means Black is up about half a pawn. The higher the positive number, the better the position for White; the lower (more negative), the better for Black. The scale is fairly consistent: +300 is usually a clearly winning game, while +50 is more of a modest advantage.

What matters is that these numbers only mean something if the engine has analyzed at a high depth. With just a few moves of search depth, evaluations are often inaccurate. Plus, practical factors come into play - some positions with similar engine scores play completely differently in practice depending on how active your pieces are or how much chaos is on the board.

In tournaments and analysis you'll also see symbols like ±, ∓, +/= and so on, which spell out verbally what the number means. +/= stands for "slight advantage," for instance, while +- means "winning." When you're analyzing games, it helps not to get too hung up on exact numbers but to treat them as rough guidance instead.

The number in chess engines like Stockfish is an evaluation in centipawns, meaning hundredths of a pawn's value - +100 means White has roughly a one-pawn advantage, -200 means Black is ahead by two pawns. The higher the positive or negative number, the better or worse the position is for that side. At 0.00, the position is completely equal, and at extreme values like +10.00 or -15.00, one side is practically winning, because no normal pawn advantage can get that large - that instead means the engine sees a forced win or loss in the foreseeable future.

What really helps me in practice: I don't use centipawns as absolute values, but as a trend indicator. Instead of thinking "oh, +250, so two and a half pawns", I look at whether the number goes up or down while I'm analyzing a line - that shows me whether my plan improves or worsens the position. A practical trick is also to compare the engine evaluation right before and after a move: if a piece then only gives +50 instead of +200, I know immediately that this move is weaker, without having to read too much into the absolute number. With very high values (over +800 or so) you really can only see "yeah, someone's clearly winning here", the exact size doesn't matter anymore.

The most practical approach is to think of those numbers simply as **positional value** rather than concrete material count. +300 doesn't mean "you're up 3 pawns," it means "the engine sees a clear advantage for white here." Same way -150 doesn't literally mean one and a half pawns for black, it's more like "hmm, this is getting tight for white." Engines don't just calculate material - they factor in activity, structure, tempo moves - so converting those numbers to pawns often just doesn't work.

What helps me: look at whether the evaluation is stable or jumping all over the place. If your move drops the eval from +80 to +20, your move was suboptimal - regardless of whether that's "a third of a pawn" or not. On the flip side, if it jumps from -100 to +50, you did something right. It's about the **change**, not the absolute number. Also: chess engine evaluations around +500 or higher are basically "black resigned," because the practical winning chances against a halfway decent opponent are just gone. Anything beyond that gets pretty theoretical pretty fast.

When I analyze games, I notice that the previous answers present practical handling a bit too mechanically. Sure, +100 roughly equals a pawn, but the engine doesn't really evaluate "pawns" - it tells you how strong a position is for a player, measured in this unit system. The value is more of a compass than a measuring stick, and from a certain point onwards (somewhere between +200 and +400, depending on the position) the position is practically winning - the exact number isn't really that relevant anymore. I find it much more useful to pay attention to the direction: is the position gaining in value or getting worse? That tells you more about your moves than the absolute numbers do.

Don't take centipawns as exact measuring sticks - that's the most common mistake. When the engine shows you +247, it doesn't mean "exactly 2.47 pawns advantage," but rather a rough ballpark figure in that range. The reality on the board is much more complicated, because material isn't everything. A pawn on the 7th rank is often worth more than two pawns somewhere in the middle, and the engine can't simply capture that in its raw centipawn number. Think of the number more as an indicator, not as gospel truth.

In practice, I find the trend more helpful. When I'm analyzing a position and see the evaluation slowly climbing from +100 to +320 while I'm pursuing a plan, I know: cool, I'm doing something right. The flip side: if the number suddenly drops from +150 to -50 after my move, it's clear something went wrong. That's how I use the engine as a feedback tool, not as absolute truth.

One more important point: with really large numbers (like +800 or more), the engine often just tells you "yeah, this is simply winning," but how long exactly it takes or what concrete plan leads to victory - the number doesn't show you that. That's why a lot of people combine the evaluation with the variation analysis the engine displays - only then do you get a real picture. 🎯

Important to know: engines evaluate from their perfect perspective, not from a human's. A position with +150 can still be hard for you to win over the board if the engine sees a move you don't find. Conversely, positions with seemingly equal evaluations can be practically lost if your opponent has a precise plan and you don't. So the centipawns show you more whether the engine considers the position won, equal, or lost - not whether *you* can win it.

An important point to keep in mind: the engine evaluation only tells you how good the position is for one side if both players play perfectly from that point on. In reality, that's unrealistic, especially with humans. A position rated at +200 centipawns can still be lost for a hobby player because they lack the technique needed to convert the advantage.

That's why it's also useful to know at what point an evaluation becomes psychologically relevant. Roughly speaking: anything under +150 is still playable for the other side, +200 to +400 is uncomfortable but not hopeless. From +600 upward, things get really dark for humans. The exact centipawn number then matters less than the category the position falls into - and whether that category is developing for better or worse as the game progresses.