Is it normal to have much worse ACPL in the middlegame?

So, I was checking the statistics in my insights (A thing Kramnik taught me, and I like numbers and statistics now), and I sorted my ACPL by game phase, and it shows a huge difference between phases: 44.7 in the opening, then up to 75.33 in the middlegame, and then drops to 45.71 in the endgame ...

I understand the logic that it should be lower in the opening phase, since we memorize the first few opening moves, but the difference is still 30 points, which is huge ....

So my question is: is this normal? this big difference in ACPL between different phases? do you have the same situation, or my insights are "bugged" or something?

When I switch to average accuracy, it shows: 80% - 60% - 75% .... which is still a big difference
Is this common?

Or am I simply uniquely suck at middlegames??

The middlegame is the most complex phase. In the opening, I could repeat book moves and my APCL will be similar to Magnus's. Same for the endgame- I may already be completely winning or reach a theoretical position and once again have a low APCL. But in between... that's what separates the men from the boys.

If you're stressing out over accuracy...
Don't.

Your accuracy being worse in the middlegame is normal, and accuracy can be higher or lower based on how complex the game is.

For example, if a game has a queen trade move 9, is very open, not a lot of attacking going on, the accuracy should be higher. On the other hand, if the game is very complex, Stockfish may be more disapproving of your moves due to there being "only one move that really works."

How to get better at the middlegame?

First, the opening is important. Having good development and castling should usually get the task done (don't castle into an attack, though. And don't use too much time.

Then...

1. Start rerouting your pieces to better squares.
2. Start an attack.
   2a. Even if your opponent starts an attack, think of counterattacks! Sometimes, counterattacks are the best attacks.
3. Make sure your king is defended well.
4. Look for tactics.

Hope this helps.

You are an intermediate player. Most games in your level are decided by tactics and positions are more open and sharp. Of course it is normal to have high ACPL for your level in all stages of the game. It will reduce as you get better. You should measure your progress on less number of mistakes per game instead of ACPL.

Yes, most players have higher ACPL in the middlegame.

My Game phase Average centipawn loss:
Opening 17.29
Middlegame 55.35
Endgame 44.17
Overall, my ACPL is about 40, with a rating about 1950.

ACPL by Rating might look like (guesses at Lichess Blitz ranges):

| Rating........| Opening | Middlegame |..Endgame..|
| <1600........ |..60-80.....|...120-150...... |....90-120.... |
| 1600–2000 |..40-60.....|...80-120.........|....60-90.......|
| 2000–2400 |..20-40....|...50-80........... |....30-60.......|
| 2400+........ |..10-20.....|...30-50.......... |....20-30.......|

A rule of thumb is Opening ACPL about half of Middlegame ACPL, and Endgame ACPL about 75% of Middlegame ACPL. For me, Opening ACPL ratio is 31% and Endgame ACPL ratio is 80%. I make less mistakes in the opening compared to average players of my rating.

Accuracy might be a better metric.

My Game phase Accuracy:
Opening 93%
Middlegame 77%
Endgame 77%

Accuracy by Rating might look like (guesses at Lichess Blitz ranges):

|.. Rating....... |..Opening.. | Middlegame |..Endgame.. |
| below 1600 | 60% to 80% | 60% to 80% | 65% to 85% |
|..1600-2000 | 75% to 90% | 70% to 85% | 75% to 90% |
| above 2000 |..above 90% | 85% to 95% | 90% to 98% |

A rule of thumb is equal Accuracy in each game phase. For players around 2000 rating, Accuracy by game phase is 90%/85%/90%. I need to work on my Endgame.

Of course, the guesses in the tables are probably wrong, and are just for discussion.

=== links ===
"Note that there is some criticism about centipawn loss since it doesn't differentiate from going from a 0.0 position to a -2 position or from +7 to +5. The second scenario obviously isn't a mistake, but in both cases, there is a CPL of 200. So many chess websites started using accuracy instead of ACPL"
https://lichess.org/@/jk_182/blog/centipawn-loss-distribution/If1ed2rj

relationship between aCPL and rating through a scatterplot:
https://web.chessdigits.com/articles/predicting-rating-from-centipawn-loss
https://www.reddit.com/r/chess/comments/7ba050/there_is_no_relationship_between_average/?rdt=41678

"I've heard this to convert ACPL > ELO
3100 x e^(-0.01 x ACPL)
Seems accurate."
https://www.chess.com/blog/raync910/average-centipawn-loss-chess-acpl

Using the formula:
rating = 3100 x e^(-0.01 x ACPL)

I am inclined to repeat @Lizani: "But in between... that's what separates the men from the boys."

but with the caveat, that middlegame mastery depends on knowing which endgames you want, and which you don't. And how to get there. Yeah, it's the how to get there that makes it middlegame mastery, it's the where am I going that gets you in the right direction. Chess is never simple nor easy.

https://lichess.org/insights/Toscani/accuracy/phase
Game phase Accuracy Number of moves
Opening 75.8% 42,939
Middlegame 61.2% 52,608
Endgame 69.5% 43,693

https://lichess.org/insights/Aborigen100500/accuracy/phase
Game phase Accuracy Number of moves
Opening 92.9% 46,225
Middlegame 84.7% 69,643
Endgame 81.1% 84,132

https://lichess.org/insights/penguingim1/accuracy/phase
Game phase Accuracy Number of moves
Opening 85% 47,371
Middlegame 68.5% 75,825
Endgame 61.4% 66,367

https://lichess.org/insights/YoBot_v2/accuracy/phase
Game phase Accuracy Number of moves
Opening 99% 2,330
Middlegame 98.7% 5,114
Endgame 98.6% 10,804

Intermediate category (Blitz Rating: 1720)
https://lichess.org/insights/Toscani/acpl/phase
Game phase Average centipawn loss Number of moves
Opening 54.47 42,940
Middlegame 83.34 52,611
Endgame 57.41 43,704

National Master (Blitz Rating: 2764)
https://lichess.org/insights/Aborigen100500/acpl/phase
Game phase Average centipawn loss Number of moves
Opening 18.62 46,220
Middlegame 33.72 69,626
Endgame 36.00 84,154

Grand Master (Bullet Rating: 3238)
https://lichess.org/insights/penguingim1/acpl/phase
Game phase Average centipawn loss Number of moves
Opening 28.83 47,374
Middlegame 58.89 75,839
Endgame 59.96 66,469

Bot (Stockfish Dev) (Rapid Rating: 3115)
https://lichess.org/insights/YoBot_v2/acpl/phase
Game phase Average centipawn loss Number of moves
Opening 3.56 2,330
Middlegame 4.21 5,114
Endgame 6.09 11,024

@Enlightenmania2022 I don't want to offend you, but I think you are wasting your time with such analyses.

Measuring the quality of your games against the ACPL when you don't even have ‘average’ chess skills is like a child who is just learning to ride a bike and is trying to decide which clipless pedals to buy next.

Free yourself from these numbers and learn to find ‘logical’ moves according to the position. After that, your real chess journey really begins and you can start thinking about more ‘complex’ topics.

In your last two games, you disregarded all opening rules by moving your queen back and forth several times. But your statistics say that you have problems in the middlegame? shakes head

Good luck on your LIFELONG chess journey!

There is limited study time. Can Chess Insights be used to determine a player's weaknesses, and so guide a study plan? Choices are openings, tactics, and endgames.

Accuracy is a better metric than ACPL, but is measuring a player's performance against the AI ideal. Boring games will have high accuracy, while wild games will have low accuracy, for both players. I prefer wild games, but don't like the low accuracy. Relative Accuracy (advantage or performance) would be interesting.
Relative Accuracy = my Accuracy minus my opponent's Accuracy

I looked at my Accuracy and ACPL on Chess Insights, and couldn't find any meaningful trends over time.