On average, how many pawns are there at the start of the endgame?

Question: On average, how many pawns are there at the start of the endgame?

1. You can choose your definition of the endgame as long as you can show a credible reference that supports your choice of definition. What I have in mind is the same as lichess which I understand to be 6 or fewer non-pawn and non-king pieces.

2. You may answer for both sides total (e.g., 10 pawns total on average) or for just 1 side (e.g., 5 pawns per side on average).

3. As to the sample...maybe the same/similar sample from which 38 moves average was obtained.

• 3.1. You can pick your own sample like, say, all standard world championships except 2021. Or 21st century standard world championships except 2021

What I tried:

• Checking out some of my recent games on lichess (that reached the endgame), I see a lot of my games have at least 10 pawns. A few of my games have 6 or lower though. I estimate the average is at least 10 pawns.

Little context:

• I think choker has too few pawns, relative to start of the endgame of regular chess/9LX. I want to verify/disprove this conjecture by seeing the average in actual games. In particular, choker seems to have, on average, 5 pawns total for both players. This is about half what I think the average for start of endgame in regular chess/9LX is.

• Alternative question: Is the average number of pawns probably significantly greater than 5 (total, not per player)?

• I find this question interesting. It's best put to people interested in standard chess statistics, not variants. Commented Nov 20, 2021 at 17:09
• you do not do yourself any favours by undoing edits, that people make (often with considerable effort) to make your questions presentable, in order to restore unreadably casual language and large volumes of irrelevant extraneous details. Commented Nov 20, 2021 at 17:49
• I find it interesting too (*figures*/types can be easily googled) and strongly suggest replacing "however" by "officers at most 9 P~=1Q~=2RR~=3NNN, any pawns, for both sides". Whereas this can be questioned, it's quite realistic aaaaand, it can be easily queried on a megabase. Commented Nov 21, 2021 at 18:35
• @BCLC: If it's OK for you (see above ;-) I try to edit the first paragraph. Or do you mean "how to query"? I never work with online stuff, including megabases, but if this can't be done with "internal" features of them, it can be done with Costeff's CQL system. Commented Nov 22, 2021 at 8:11
• @BCLC: I haven't the software to chase after the question, so my definition is rather moot. I just wanted to suggest a sensible definition for you to adapt which is also computer applicable (I'm certain Chessbase or whatnot allows a search for this material, maybe split by type), such that you question doesn't get closed on "too nebulous". Commented Dec 3, 2021 at 12:21

for all games in magnus carlsen world standard and world fischer random chess championship matches (basically all world championship matches except world rapid and world blitz)

Average throughout all championships:

for all: (13+10+11+12+10+12+8+14+11+14+14+13+12+12+14+10+12+12+9+9+10+14+12+12+14+11+12+13+12+11+6+16+15+14+12+14+8+10+9+12+11+10+10+10+14+7+5+12+14+14+14+12+9+6+12+5+11+10+8+6+11+9)/(4+2+4+11+11+12+10+8)=11.1129032258... ~ 11.11

for all classical (and thus excluding 9LX huhu): (13+10+11+12+10+12+8+14+11+14+14+13+12+12+14+10+12+12+9+9+10+14+12+12+14+11+12+13+12+11+6+16+15+14+12+14+8+10+9+12+11+10+10+10+14+7+5+12+14+14+14+12)/(11+11+12+10+8)=11.5769230769... ~ 11.58

for all rapid including 9LX: (9+6+12+5+11+10+8+6+11+9)/(4+2+4)=8.7

for all rapid excluding 9LX: (11+10+8+6+11+9)/6 = 9.1666666666... ~ 9.17

Average by championship, rounded to 2 decimal places:

Championship Average
2021 vs nepo 11.73
2019 vs wesley (which was all rapid) 08.00
2018 vs fabi (classical only) 11.55
2018 vs fabi (classical and rapid) 11.38
2018 vs fabi (rapid only) 10.50
2016 vs sergey (classical only) 12.33
2016 vs sergey (classical and rapid) 11.38
2016 vs sergey (rapid only) 08.50
2014 vs vishy 10.60
2013 vs vishy 11.50

Remarks:

1. the lowest average in all these (8 pawns) is the 1 that magnus lost namely to wesley so in 2019 fischer random world championship. of course all those games were rapid.

2. Lowest in classical only is 10.60 in 2014 vs vishy. the 3 rapid only entries are all below 10.60: they are 08.00, 10.50 and 08.50. I guess pawns have a higher mortality rate in rapid compared to classical.

2021 vs nepo

https://lichess.org/study/1NIB7sSp

1 - 13

2 - 10

3 - 11

4 - 12

5 - 10

6 - 12

7 - 8

8 - 14

9 - 11

10 - 14

11 - 14

average: (13+10+11+12+10+12+8+14+11+14+14)/11=11.727272...

2019 vs wesley

1 - 9

2 - 6

3 - 12

4 - 5

5 - N/A

6 - N/A

average (all rapid): (9+6+12+5)/4=8

2018 vs fabi

https://lichess.org/study/fCum15PU

1 - 13

2 - 12

3 - 12

4 - 14

5 - 10

6 - 12

7 - 12

8 - 9

9 - 9

10 - 10

11 - 14

12 - N/A

T1 - 11

T2 - N/A

T3 - 10

average (classical only): (13+12+12+14+10+12+12+9+9+10+14)/11=11.545454...

average (classical and rapid): (13+12+12+14+10+12+12+9+9+10+14+11+10)/13=11.3846153846...

average (rapid only): (11+10)/2=10.5

2016 vs sergey

https://lichess.org/study/vENPCM25

1 - 12

2 - 12

3 - 14

4 - 11

5 - 12

6 - 13

7 - 12

8 - 11

9 - 6

10 - 16

11 - 15

12 - 14

T1 - 8

T2 - 6

T3 - 11

T4 - 9

average (classical only): (12+12+14+11+12+13+12+11+6+16+15+14)/12=12.3333333333

average (classical and rapid): (12+12+14+11+12+13+12+11+6+16+15+14+8+6+11+9)/16=11.375

average (rapid only): (8+6+11+9)/4=8.5

2014 vs vishy

https://lichess.org/study/nurfx2OL

1 - 12

2 - 14

3 - 8

4 - 10

5 - 9

6 - 12

7 - 11

8 - 10

9 - N/A

10 - 10

11 - 10

average: (12+14+8+10+9+12+11+10+10+10)/10=10.6

2013 vs vishy

https://lichess.org/study/9xUbdEqD

1 - N/A

2 - 14

3 - 7

4 - 5

5 - 12

6 - 14

7 - 14

8 - 14

9 - N/A

10 - 12

average: (14+7+5+12+14+14+14+12)/8=11.5

• This answer could be greatly improved with formatting for future visitors in mind. Commented Dec 20, 2021 at 21:17
• @SecretAgentMan probably. i just didn't think anyone cared about my dumb questions or answers. hehe. thanks. what do you suggest?
– BCLC
Commented Jan 23, 2022 at 6:42
• Not sure...maybe WCC across the columns (see here). Commented Jan 23, 2022 at 13:14
• @SecretAgentMan oh thanks. i added something!
– BCLC
Commented Jan 23, 2022 at 14:39

It is impossible to predict the number because there are more possible outcomes in chess than in any other game I can think of. When you play in the first 15 moves "most" games will lose 2-4 pawns and at the end of a game with Grandmasters or Masters, the number can range from 4-2 pawns on the board because they tend to sacrifice the pawns for an advantage. But for 1 advanced player and one beginner player you could end up having 6 pawns on both sides because the beginner player will most likely lose in fewer moves than 2 advanced players. If 2 beginner players are playing it could go from 5-8 because they tend to play quicker and not spend time analyzing the board.

Some people define the endgame as the last 5 pieces while come define it as the last 10-20 moves or when one player has a major advantage. When you say average you have to count all if not most of the possible outcomes so you should the number depends on the type of player (defensive or attacking), Strength of player (Rating), Time control (5 min, 30 min, 1 hr). Many GMs play defensively, and many beginner players prefer attacking their opponent's pieces. So there is no real definite answer but it depends on what kind of player you play.

• 1 - 'Some people define it as the last 5 pieces ' --> ok let's do 6 like lichess assuming 'it' is start of endgame ? 2 - why can't we apply what you're saying to average number of moves per game? 3 - in my own games i found the average is at least 10. was what i did somehow invalid?
– BCLC
Commented Dec 3, 2021 at 1:58
• 3 - Of course not. I mean number of pawns per side (Dark and light). Commented Dec 3, 2021 at 3:20
• I really don't think the answer is impossible to find. Just get a set of games and look! Yes, it will vary by player, but there's still going to be an average.
– D M
Commented Dec 5, 2021 at 14:19
• Oh could you please update the question to just world championships. Commented Dec 6, 2021 at 0:03
• Also, could you please define your definition of endgame like as a recordable way not opinion based. Commented Dec 6, 2021 at 0:12

Not on average, but in general, there are 1 to 6 pawns in the endgame (7 and 8 excluded because that is statistically impossible.

• I don't think 7 and 8 pawns are statistically impossible at all. Players can just exchange all their minor and major pieces if they wanted to. Also, you have now included all possibilities of pawns... e.g. 1-6 and not 7 or 8, which doesn't really answer the question! Commented Dec 7, 2021 at 23:01
• any references? statistics? and is it 1 to 6 per player? or total? no offense downvoting until this is clarified
– BCLC
Commented Dec 8, 2021 at 0:48
• What is the basis for your answer? Commented Dec 20, 2021 at 21:17