Where does the decisive point of most chess games occur?

[Jason Manley]

The title to this thread is a question that I've been considering while looking through some of my past games over the past few days. What, in your experience is the most typical point of the game where you can point and say "this is why White/Black lost." From my own games, I've come to the tenuous ranking of the following four situations (including time trouble errors as a situation, because I wasn't sure where else to put them), ranked in order from most frequently to least frequently containing the decisive point:

1. Middlegame mistakes
2. Time management induced errors
3. Endgame errors
4. Opening slips

How does everyone else feel about this ranking? What kind of experience have you had with decisive points in your own games?

[John Torrie]

Re: Where does the decisive point of most chess games occur?

Hello Mr. Manley,

You might be hoping to catch some bigger fish than me with this thread, but here goes:
What I've found most engaging in looking at some of my own and other player's games over the last few years
is the late middlegame transition to the ending where often a series of accurate moves are needed to bring
about the win. See the Franzoni(sp) - Day game I mentioned below. Alongside pure endings, I like this technical
aspect of chess. Perhaps when I've gained some proficency in this area I'll turn my attention to the middlegame,
of which I only have a basic understanding. By nature I am a gambit player and like to solve things tactically.
I've almost never been in time pressure because I am impatient and prone to take risks. Whether I get in a good
spot or bad it livens the game for me. The results often depend on the caliper of my opponent.

[Roger Patterson]

Re: Where does the decisive point of most chess games occur?

The title to this thread is a question that I've been considering while looking through some of my past games over the past few days. What, in your experience is the most typical point of the game where you can point and say "this is why White/Black lost." From my own games, I've come to the tenuous ranking of the following four situations (including time trouble errors as a situation, because I wasn't sure where else to put them), ranked in order from most frequently to least frequently containing the decisive point:

1. Middlegame mistakes
2. Time management induced errors
3. Endgame errors
4. Opening slips

How does everyone else feel about this ranking? What kind of experience have you had with decisive points in your own games?

At one time, I was interested in the question of when one could definitively state the game was won or lost. i.e. not white is up a piece and should win but the game is winning and will win barring earthquakes and heart attacks against anybody - Carlsen, super computer, what ever. I ultimately concluded that in most of my games, such a definitive statement was only possible usually only a few moves before the actual end of the game.

As to your question, the point at which you can say why one lost is probably well before such a point - which is to say, after making that mistake, there were likely opportunities to fight harder or confuse the issue after making the mistake. Particulary at low levels of play, there are usually multiple opportunities to change the course of the game. Thinking of your losses being due to one particular mistake is probably not that good a way of viewing the progress of your games.

[Matthew Nicholson]

Re: Where does the decisive point of most chess games occur?

Interesting topic. I think the standard at the tournament level for a 'lost' game is a position with a +2.00 computer evaluation or better. It's been my experience that most players (even masters) have to be down 5+ to resign, but generally most people are willing to accept that they had a "very bad" position at the 2.00 point, especially if that has already been converted into a material advantage. Those who continue to analyze in the post mortem beyond that point are destined to be perennially weak players.

With that as my definition of a decisive advantage, I'd say against somebody considerably weaker I would often achieve that advantage in the opening. For example as a junior with a rating of 1800 I played somebody around 700 and the game went 1.e4 e5 2.f4 f6? 3.fxe5 fxe5, at which point my advantage has reached the decisive level.

Against somebody considerably stronger I find myself hanging around until the late middlegame. This is likely because as players get stronger their opening repertoires tighten up. As... Shirov? or Nakamura? or Ivanchuk? or maybe all of them... said... 'A 2200 can be as strong as a super GM for the first 20 moves', and that's a direct result of preparation.

Now the real question likely is: "Against a player of absolutely equal skill, at what point is the game likely to become decisive?". I'd answer the middlegame, again. Often the complexity of the game is such that both sides miss the decisive point, but that doesn't change my definition of decisiveness. In last year's Ontario Open I lost twice (disappointing) to roughly equal players. One had about 80 points on me and the other had 2. In both cases I was under a small bit of pressure and made an error in a critical position where I could have brought about equality. In both cases it was a calculation error quite deep into a non-forcing variation. But the key is that in both cases there was already pressure.

Now, everything changes if you're asking "When does the last decisive mistake happen." In that case it's very often the endgame. Modern chess players very rarely put themselves into a position where they are at risk of getting mated, and so a lot of games make it to an endgame. Often these endgames are decisive because one player has an extra pawn or two, but in many cases the technique to close out the game is completely lacking because studying endgame books isn't nearly as fun as playing bullet on chess.com.

My thoughts, anyway :).

[Mathieu Cloutier]

Re: Where does the decisive point of most chess games occur?

Interesting topic. I think the standard at the tournament level for a 'lost' game is a position with a +2.00 computer evaluation or better. It's been my experience that most players (even masters) have to be down 5+ to resign, but generally most people are willing to accept that they had a "very bad" position at the 2.00 point, especially if that has already been converted into a material advantage. Those who continue to analyze in the post mortem beyond that point are destined to be perennially weak players.

I don't disagree with your thresholds. +2 is usually over, barring any major mistake.

However, there's +2 and then there's +2! Sometimes, you have a winning position, but the position is critical and only one very difficult move wins it all. If you miss it, your oppenent comes back into the game. And then you have these games where you simply have a piece for a pawn (+2) in a quiet position and any reasonable move/plan will win the game.

Even though the objective evaluation might be +2 for both positions, the probability of winning is not the same, from a practical point of view.

[Paul Bonham]

Re: Where does the decisive point of most chess games occur?

I don't disagree with your thresholds. +2 is usually over, barring any major mistake.

However, there's +2 and then there's +2! Sometimes, you have a winning position, but the position is critical and only one very difficult move wins it all. If you miss it, your oppenent comes back into the game. And then you have these games where you simply have a piece for a pawn (+2) in a quiet position and any reasonable move/plan will win the game.

Even though the objective evaluation might be +2 for both positions, the probability of winning is not the same, from a practical point of view.

+2 can be losing if you allowed your opponent to give up the exchange in return for an attack on your King. The attack either must be mating or else must get significant material back.

+3 is sometimes not enough to win a game, as in K+N vs. K or K+B vs. K.

If you are down a whole Knight or a whole Bishop during a game, there is hope if you can eliminate all opposing Pawns. Therefore one possible answer to the question of the OP is the moment when one side (and only one side) has a Pawn promotion that cannot be stopped or can only be stopped by giving up significant material. In endgame play, if a draw is all you are after, you can even give up a Knight or Bishop for a single Pawn if that Pawn is the last one your opponent has and (preferably) Queens and Rooks are off the board.

To make chess more exciting (i.e. less draws), I thought of a rule change whereby if you get a Pawn to your 6th rank, you have the option of promoting that Pawn to Knight or Bishop. If you keep it as a Pawn, and get it to your 7th rank, you can choose to promote to Knight, Bishop or Rook. Or you can leave it a Pawn and try and promote to Queen or Rook or Knight or Bishop on the 8th rank. I call this VP Chess (for "Variable Promotion").

Mathieu's point about critical position actually came up in the last WC match. Can't remember which game, but Carlsen won with a flourish that everyone thought was a brilliancy. But in fact, Karjakin had a way to stop the brilliant finish and turn the game into a less certain ending in which Karjakin would likely lose but might have drawing chances. I can't recall the exact point deficiency that Stockfish gave Karjakin for the safer line, but it might have been around -2.

Instead, Karjakin was drawn into the psychological trap that (I believe) Carlsen planned: Karjakin tried for a back-rank mate that looked promising but was stoppable at the last moment and allowed Carlsen's brilliancy (which by the way was the only path to victory and wasn't obvious -- any other move and Carlsen loses).

Carlsen to my knowledge has never been asked whether he foresaw all that when he made the critical move. If he did see it, that means he predicted that Karjakin would try for the back-rank mate, which would make it one of the great psychological traps of all time imo. The point is that Carlsen had, at the critical juncture, a much more obvious and safer move that gave him almost +4 advantage and Karjakin had no real chance. But the move that Carlsen did play gave rise to very complicated double-edged play in which one small slip was disaster.

I think Carlsen visualized and calculated it all, including that Karjakin would try for the back-rank mate. The other possibility is that Carlsen actually slipped up and got lucky, at least in the sense that if Karjakin played more safely the game would have dragged on much longer.

[Andy Shaw]

Re: Where does the decisive point of most chess games occur?

Chess is a game of absolutes and objective evaluation. In any given position, assuming each player plays the best moves (obviously a huge assumption) either one player will win and the other lose or the game will be drawn. While I realize that you cannot assume that each player will indefinitely make the best moves, it doesn't make sense to look at chess any other way. This is the same as when we say that you should 'play the board' and not the player.

If you look at a position where one player has a minor piece over his opponent and the other has no attacking combination, opportunity to queen a pawn, etc. the other player has a 0% 'chance' of winning using this model. Unless you can calculate that even though you are down material you will be either to a) gain back more material than you lose or b) mate your opponent and there is no combination of moves to prevent this, you have lost the game. Of course if this was the case, you weren't really 'down' to begin with. There is no 'chance' or probability involved here unless you are playing with faulty tactics and hoping that your opponent won't find the correct defense. The attack either works or it doesn't... purely objective.

I believe the same can be argued even when there is no decisive material deficit for one side. Either one player has a convertable positional advantage or they don't and the game should be drawn. For example in a rook and pawn end game where material is equal, however one player has double isolated pawns and the other has both rooks doubled on the only open file. We would likely be able to look at that position and very quickly evaluate that the player with the doubled rooks and no doubled pawns had a positional advantage in that position. However this is only useful if the positional advantage can be converted into a mating attack or a material advantage which will eventually lead to a mating attack. In the original position with the rooks on the open file there is a finite number of possible moves and responses from those moves (though the finite number is very large). Either there is a possible set(s) of moves which will result in one side gaining a mating attack or decisive material advantage (again assuming best play from each side) or there isn't.

Middlegame mistakes are most often the decisive factor.

[Mathieu Cloutier]

Re: Where does the decisive point of most chess games occur?

+2 can be losing if you allowed your opponent to give up the exchange in return for an attack on your King. The attack either must be mating or else must get significant material back.

+3 is sometimes not enough to win a game, as in K+N vs. K or K+B vs. K.

In the first case, it means the evaluation is not +2 in the first place. You can be the exchange up and the evaluation is losing for you because of a specific variation.

As for a piece being +3, most engines understand a lot of the exceptions. K+B vs. K on an empty board is not +3. Lots of engine correctly evaluate it as 0.00.

There's more than material in the evaluation of a chess position.

[Roger Patterson]

Re: Where does the decisive point of most chess games occur?

I don't disagree with your thresholds. +2 is usually over, barring any major mistake.

However, there's +2 and then there's +2! Sometimes, you have a winning position, but the position is critical and only one very difficult move wins it all. If you miss it, your oppenent comes back into the game. And then you have these games where you simply have a piece for a pawn (+2) in a quiet position and any reasonable move/plan will win the game.

Even though the objective evaluation might be +2 for both positions, the probability of winning is not the same, from a practical point of view.

One flaw with stating a threshold for this problem is that different engines produce different results and even the same engine produces different results depending on move depth. Correspondingly, what is an easy win for Carlsen is not an easy win for me.

Too much significance is put on the size of the numerical value an engine puts out. It is fairly easy to construct positions where an engine has a large value but does not correlate very well into actual results. You might dismiss constructions such as the recently discussed Penrose problem as not being relevant to normal play but even actual game positions can show this type of behaviour.

For example, Stockfish gives this position from Patterson-Fuentabella as +2 (and considers almost any plausible move by white as winning - it's evaluation does not depend particulary on concrete lines) but in fact there is a long way to go before one can say this position is easily winning for most players. I can agree that white's position is very good but it's not clearly winning. Stockfish's analysis does show any actual path to victory - it just likes white's position.
(the game was eventually drawn).

white to move


or as another example, from an analysis position from Patterson-Cao, again Stockfish considers it to be +2 (if you give it a bit of time). Again it's not particulary obvious that this can be described as a win. The variations are more tactical so it's closer to be forced and one might see how it is winning with a fair amount of thought. Personally, in practical play, I would be (and was) very happy with white's position but would not describe it as an easy or forced win although after studying the position I willing to believe it. [stockfish's best line of play is predicted to be 34.....Ba8 35 Kh2 - not a very likely line of play for a forced win.]

In short, an engine's numerical evaluation, while interesting and certainly a good guide as to what move to take, is not a fully trustworthy indicator of the truth of a position. Using it as a sole indicator of when a game turns from being simply bad to being lost is an oversimplification

black to move

[Paul Bonham]

Re: Where does the decisive point of most chess games occur?

In the first case, it means the evaluation is not +2 in the first place. You can be the exchange up and the evaluation is losing for you because of a specific variation.

As for a piece being +3, most engines understand a lot of the exceptions. K+B vs. K on an empty board is not +3. Lots of engine correctly evaluate it as 0.00.

There's more than material in the evaluation of a chess position.

I was referring to cases where Player A is up the exchange but there's an attack against Player A's King, and I should have mentioned that the culmination of the attack is beyond the horizon of the engine. Or it could be just an overall positional advantage that doesn't come to anything in the search horizon of the engine but will come to something beyond that (passed pawns being a prime example).

In the variant I mentioned where Pawns can promote on the 6th, 7th or 8th ranks, new tactics emerge where pushing a Pawn to its 6th rank and promoting to a Knight or Bishop is done while at the same time the Pawn push discovers an attack on the opposing King or Queen or Rook. I've seen this because I have an engine that plays this variant. I mention this because the engine probably changes its evaluations based on the positional characteristics of how far Pawns have advanced, more so than a standard chess engine would do where the Pawns have further to go to promote.

In the case of +3, I was meaning where one is down a whole minor piece but the board is still well populated. The engine will evaluate it as roughly +3 against you, but if you can manage to trade off all pawns and trade off pieces evenly, you can still work towards a draw. In other words, +3 middlegames are not necessarily won if we are talking about players below a certain rating threshold, such as maybe 2200 or 2000.

It would be interesting to have a stat on that: the total number of all games between non-titled players (maybe excluding junior players) in a major database where one player was at some point down a whole minor piece in material and didn't resign and managed to get a draw versus the total of all such games where the player down the minor piece played on let's say at least a dozen moves after being down the piece and did lose. The stat maybe could be broken into rating categories, to see how the likelihood goes down as the ratings go up.

One could also tabulate this stat for games between titled players, but how often would one expect a titled player to be up a minor piece in the middlegame and end up having to settle for a draw?

One could even do a study of this using only a chess-playing engine. Devise a set of middlegame start positions that are completely benign, i.e. there are no immediate tactics. Remove one of the Knights or Bishops of one side, preferably a Knight or Bishop that is not critical to the position (ideally none of them would be) for the side that is to move. Then run a monte carlo simulation of maybe 1000 games run to completion from that position (engines could be set to play at only 4 ply search depth for speed, with high randomness, sometimes choosing the 2nd or 3rd best move rather than the best, and will play until the game is won or lost by mate or drawn by 50 move rule or 3 time repetition or stalemate). Ideally you can remove each Knight and Bishop, and get 4 separate monte carlo runs from each such start position. Tabulate each game result.

This could be a "lesson" for young chess players at a club. Pair them off (as evenly matched ratings-wise as possible) and have each pair of players set up the start position (as described above). It would be better NOT to just use the standard chess starting position, because then too much opening knowledge might impact results. Perhaps just take an opening line out of ECO, play the first N moves for each side until everything is pretty much developed, and have Stockfish evaluate the position. If it's between -0.5 to +0.5 evaluation, then it should be suitable. Have one of the players play being down the minor piece. Resignation is not allowed, game must be played to completion. Then have the two players set it up again, but with roles reversed, and play it again. Tabulate all results.

The lesson could be repeated but this time where:
- one side is down (up) an exchange
- one side has 3 minors for a Queen
- one side has 2 minors for a Rook
- one side has Rook and minor for a Queen
- 2 Knights versus 2 Bishops

There was a thread here maybe 3 or 4 months ago where someone posted problem positions where the engine evaluated one side being up significantly in material despite there being no winning play possible. The point was that the engine wasn't considering the 50 move rule until the rule actually got within its search depth.

[Matthew Nicholson]

Re: Where does the decisive point of most chess games occur?

Lol kinda figured I'd get replies like this.

Computers are good now. Computers are amazing at chess. Computers are even good enough to recognize some positional fortresses, which is insane. Stating that occasionally a computer improperly evaluates some obscure position that isn't likely to ever show up OTB does not invalidate the truth of what I said... if a computer says one side is winning, then 99.999%(~) of the time they are. Whether the person actually manages to win is another story, and that can vary (as Cloutier said) with the complexity of the position. And if you're the type of person to go for a terrible evaluation because you think the computer is mistaken, more power to you. OTB it may not matter... the Latvian gambit no longer exists in correspondence chess, but OTB it can be fun even among masters.

Funny that people always need to be the devil's advocate around here :).

[Mathieu Cloutier]

Re: Where does the decisive point of most chess games occur?

The devil's advocate? Sometimes it feels more like Hitler's lobbyist or something worse.

[Paul Bonham]

Re: Where does the decisive point of most chess games occur?

Lol kinda figured I'd get replies like this.

Computers are good now. Computers are amazing at chess. Computers are even good enough to recognize some positional fortresses, which is insane. Stating that occasionally a computer improperly evaluates some obscure position that isn't likely to ever show up OTB does not invalidate the truth of what I said... if a computer says one side is winning, then 99.999%(~) of the time they are. Whether the person actually manages to win is another story, and that can vary (as Cloutier said) with the complexity of the position. And if you're the type of person to go for a terrible evaluation because you think the computer is mistaken, more power to you. OTB it may not matter... the Latvian gambit no longer exists in correspondence chess, but OTB it can be fun even among masters.

Funny that people always need to be the devil's advocate around here :).

Two points about that:

(1) wasn't replying to you

(2) I wasn't playing devil's advocate against engine evaluations. In fact, I was saying what you are saying: "Whether the person actually manages to win is another story, and that can vary (as Cloutier said) with the complexity of the position." I would also add it can vary with the rating levels of the players (assuming each player is rated about the same).

The thing about a computer evaluation of +3.00 or +2.00 is that it is derived from a minimax search where each side is the engine itself, and the engine does all the evaluating of positions. So if the engine is Stockfish 8 and it has a rating of say 3400, you only know that from this position forward, if each player is rated 3400, White has an advantage of 3.0. That doesn't say anything about if each player is rated below 2400... below 2200.... below 2000.... below 1800.... below 1600...

[Edit: also, what if the person down 3.00 is much higher rated than the opponent, and just happened to drop a minor piece? Should s/he resign or give up? The number 3.00 doesn't mean much just by itself.]

My earlier idea of chess lessons came about because I got thinking that at these lower rating levels, players may not be very good at turning such material advantages into absolute wins. Conversely, they may be equally poor at being down that material and fighting for a draw against an equally-rated player when they actually should have good chances of achieving such a feat.

I'm going to run some monte carlo simulations and see what I can find. I will report results back here with full explanation. It could be very interesting.

[John Torrie]

Re: Where does the decisive point of most chess games occur?

I did not follow the recent world championship, but did have a look at the queen sacrifice that decided (?) the match. If this is the game you refer to, you may want to add the possibility of Carlsen’s knowledge of Bernstein vs. Kotov, Groningen 1946 to the speculation to why Carlsen played as he did. In that game black resigned a move short of the queen sacrifice on h6 which also forced a model mate with the rooks.

[John Torrie]

Re: Where does the decisive point of most chess games occur?

I like your post, especially the note on the ending. However, at my level I have to disagree with any absolute evaluation such as a +2.00. Please note that I am not disputing the computer’s efficiency to win from a +2.00 advantage. This advantage may be enough for computers, but I don’t think it applies to the people I usually play. For that matter, consider Gelfand vs. Svidler 2001 world championship. Svidler had Q against R and was unable to pull it off in 50 moves. Or on the opposite end, Polgar vs Kasparov 1996, R vs R & N. The draw was in hand several times over, but at the time the defence showed no understanding of this ending. Also, aside from these mishaps, there are numerous examples where extra material does not win in the ending. Lasker vs Lasker 1924, R&P vs N. Closer to home, at the local club, a master player of 40 years experience admitted he could not finish a K&B&N vs K ending in the required number of moves. Others know nothing of Q vs R. It just does not interest them. It goes on. Way back when, Morphy and company often began games with -3.00 or even -5.00. All this is great news for gambit players like myself, for even if your attempt to gain development and space and seize the initiative falter, and you up with a material disadvantage going into the ending, if you have an inkling of what you are doing there, compared to your opponent who has less, the game is not over till it’s over. Someone should compile an ‘Endgames for Gamiteer’s’ book. It would be a best seller.

Consider the apparently well known Savvedra Position: Black to play; W- Kb6 Pc7/ B- Ka1 Rd5. If someone is Black what should they do? Do they resign on the spot? Do they check, check, check and hope their opponent does not know about the under-promotion win? Do they instantly resort to a R vs Q ending? Do they check once just to see? What gives them the best chance to delay or influence the outcome of the game?