How long does it take to play 50,000 games?

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
BrianM
Member
Posts: 25
Joined: Sat Jun 17, 2017 11:34 am
Location: UK

How long does it take to play 50,000 games?

Post by BrianM »

In the thread 'Human vs Machine', Lyudmil Tsvetkov has announced his new book 'Human Versus Machine: How To Beat Stockfish and Komodo Part I ‘. As an occasional player this level of chess is way above my understanding and so the book is probably of little use to me. However what does pique my interest is the statement that the book was written from knowledge gained through playing over 50,000 games. So how long does it take to play 50,000 games? As I like to do ‘back of an envelope’ calculations I thought that I would try and come up with an answer.

The first step is to come up with an estimate of the time needed for one game. For this I assumed at the very least that a game would be 20 moves each for white and black. The time for each move I decided would be 30 secs. This was chosen because: a) it is often used in dedicated machine vs machine contests and b) is is much less than the average 3 minutes per move for tournament play. So 40 moves at 30 secs per move gives 20 minutes per game. Loooking up ‘Fast Chess’ on Wikepedia I found the statement ; ’Time controls for each player in a game of blitz chess are, according to FIDE, 10 minutes or less per player’. So this adds credence to the value of 20 minutes per game or 3 games per hour.

So according to my assumptions, 50,000 games should take 50000/3 = 16666 hours. This does not mean much to me so I converted this to man years. Before I retired my working week was a nominal 39 hours. I had approximately 46 working weeks (6 weeks holiday) per year which give 1 working year as 39*46 =1794 hours. So 16666/1794 gives 9.2 working years to play 50,000 games. That is a considerable effort.

Perhaps someone could come up with a better estimate.

Brian
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Re: How long does it take to play 50,000 games?

Post by spacious_mind »

BrianM wrote:In the thread 'Human vs Machine', Lyudmil Tsvetkov has announced his new book 'Human Versus Machine: How To Beat Stockfish and Komodo Part I ‘. As an occasional player this level of chess is way above my understanding and so the book is probably of little use to me. However what does pique my interest is the statement that the book was written from knowledge gained through playing over 50,000 games. So how long does it take to play 50,000 games? As I like to do ‘back of an envelope’ calculations I thought that I would try and come up with an answer.

The first step is to come up with an estimate of the time needed for one game. For this I assumed at the very least that a game would be 20 moves each for white and black. The time for each move I decided would be 30 secs. This was chosen because: a) it is often used in dedicated machine vs machine contests and b) is is much less than the average 3 minutes per move for tournament play. So 40 moves at 30 secs per move gives 20 minutes per game. Loooking up ‘Fast Chess’ on Wikepedia I found the statement ; ’Time controls for each player in a game of blitz chess are, according to FIDE, 10 minutes or less per player’. So this adds credence to the value of 20 minutes per game or 3 games per hour.

So according to my assumptions, 50,000 games should take 50000/3 = 16666 hours. This does not mean much to me so I converted this to man years. Before I retired my working week was a nominal 39 hours. I had approximately 46 working weeks (6 weeks holiday) per year which give 1 working year as 39*46 =1794 hours. So 16666/1794 gives 9.2 working years to play 50,000 games. That is a considerable effort.

Perhaps someone could come up with a better estimate.

Brian
Well Brian that is quite accurate but I can't imagine someone beating a Komodo or Stockfish in less than 60 moves on average so you could multiply that by 3 and you get 27.6 years. But then again they may all be Bullet games. Hard to imagine though that you can keep your concentration going at a high standard for more than say 3 games a day regardless of what time setting you play. So you could also take 50,000/3 = 15,667 days/365 = 43 years for non tournament level games or 50,000/2 = 25,000 days/365 = 68.5 years ;)

Best regards
Nick
Volodymyr
Member
Posts: 141
Joined: Sat Apr 08, 2017 1:03 pm
Location: Ukraine,Radyvyliv

Post by Volodymyr »

I have not read the book Human vs Machine. But I will answer here.
50,000 games this is a bullet.200-300 games in day its a real.

https://www.youtube.com/watch?v=rGEWzOv5Bc0

What level? My 30 minutes per game, and Carlsen 1 or 2 minutes per game.
I doubt that I will win.

https://lichess.org/@/Chess-Network
20456 bullet games

https://lichess.org/@/jimakos
21267 bullet games

Everything is simple and without mathematics.
User avatar
scandien
Member
Posts: 206
Joined: Mon Sep 12, 2011 1:15 pm
Contact:

Post by scandien »

hello,

i was wondering about the same point...

As bullet game are based on tactical motive, i don't think that bullet can be sonsidered to explain chess, and i don't think too that Stockfish can be defeated in a bullet match (nor in a blitz , Rapid or standard game) . A human can only considered standard game , where he will have time to plan, and can try to outplay the machine in strategy.

Anyway, playing the game is not necessarily enough to explain chess. All game must be analysed and motives have to be extracted. This take time too.

For me this is a fake.

best regards

Nicolas
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

Even if we grant the possibility, playing this way is not chess. It's just ripping through the same obsessive, memorized routines again and again at hyperspeed, with knowledge of a particular vulnerability in particular engines. It's not creative or interesting.

- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Yep I keep visualizing the picture of Fritz 3 beating all the best in the world on a P90 or something at Blitz so what real chance would they have against SF or Komodo at Blitz. All indications are that the faster the setting is, the harder it is to actually beat the comps. Humans make more mistakes. From the analysis I had made playing Blitz causes a strength and quality loss of over 100% for Super GM's when they play each other in live tournaments. So if you were to take it down to bullet they would end up being closer to playing at about 2100-2200 elo player strength because of the drastic increases in mistakes that would be made as soon as they leave their knowledge comfort.

When I tested SF 8 at 18 Ply it performed at over 3050 ELO playing computers programs at ccrl 40/40 (1 minute per move) and took about 2-3 seconds per move on my computer. Put that 18 ply on a super fast computer it might play 18 ply at 1 second per move.

Now pit SF 18 ply at super bullet or bullet against 2100-2200 ELO game quality and see what would happen.

You are always going to get the 1 or 2 game exceptions but that is about it.

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

But having said all that it is possible that you can build this closed wall in games against SF that you work towards repeating every time you play a computer and just grind them to a halt through a 100 or so inhumane chess moves of moving pieces backwards and forwards behind a closed wall hoping that the chess program loses patience first.

But is that really the type of chess that humans should be proud of playing? It sounds to me more like chicken chess. The human being a chicken and the computer being the human who risks first ... what a reversal of roles :)

Best regards
Nick
pennine22
Member
Posts: 74
Joined: Tue Oct 14, 2014 8:01 pm

Post by pennine22 »

Hi

I’m a patzer but enjoy playing engine v. engine games both on hardware and software so can’t offer any insight. But pass on what can be gleaned from an initial browse in case it aids the discussion as bought a paperback copy.

- 15 games won by the author are presented playing both black and white.
- Opponent is Stockfish 4 or DD on a ‘quad core’ with all games played in 2013.
- Opening codes / names declared and ca. 8 different openings played by both sides.
- Author in the intro’ notes they are all ‘blitz’ or ‘rapid’ games as Volodymyr commented above. Author assigned either 5 or 10 minutes plus increment for a game with SF assigned 3 and 5 minutes with increments respectively.
- All games over by move 41 where either SF is mated or imminent (8 games), resigns (6 games) or the end is nigh in a few more moves.
- Clear short comments of key points in the game and for key alternatives from the main line with key board positions shown through games. So hope to follow the games and glean what I can at my level.

Just played through game 11, author white and SF resigns move 26 with mate imminent in a few moves. I’m not equipped to comment on how SF is beaten so decisively and quickly in this and likely the other games and as remarked with short games which should further skew to the SF benefit. So will leave for experts to analyse but certainly an enjoyable game for me.

An intriguing and perhaps enigmatic book for the reasons discussed but all the more reason why I’m pleased I bought one for a modest sum. Kindle version less than a fiver.

David
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

scandien wrote: For me this is a fake.

best regards

Nicolas
Well I really tried this morning to clear all doubts and to be really open minded about Super Bullet and Bullet games. So I clicked on one of the two links that Volodymyr shared. I picket the link that had the most Bullet games.

https://lichess.org/@/jimakos

Once I looked at the link I searched games recently played at Super Bullet and Bullet by this person.

This guy was playing while I was searching. Anyway I picked his most recent Super Bullet game with him being rated at 2391 and his opponent at 1935 ELO and ran an analysis of the game.

WHITE ELO # OF MOV INAC MIST BLUN AV MIST CENTIPAWN LOSS TOTAL SCORE
SUPER BULLET 2391 32 3 6 1 32.81 68 100.81
SUPER BULLET 1935 31 3 3 2 33.87 71 104.87

So anyway based on how I evaluate both the 2391 and 1935 player scored over 100 points. Ok I say to myself this also happens maybe zero point one percent of the time even at GM Level. So I looked at the game a little closer with Chessbase.

[fen]6k1/pp4p1/5Qbp/3P4/2p1pN2/4n1P1/PP4BP/4K3 w - - 0 30[/fen]

Black just played 29. ... Nxe3?? and now white (ELO 2391) leading by a whole Queen just needs to take the Bishop on g6 and push the pawn on d5 for the most basic of wins. (mate in 7).

Anyway White (ELO 2391) must be so panicked because of time counting down plays 30. d6? and misses a forced mate and loses his Queen after 30. ... gxf6.

[fen]6k1/pp6/3P1pbp/8/2p1pN2/4n1P1/PP4BP/4K3 w - - 0 31[/fen]

So anyway after this Blunder we have the above position. Still not bad for White (2391 ELO making a point by showing the ELO!) push pawn forward to d7 or take the Bishop on g6 and the game is won. (trying to forgive the blunder of losing his Queen). Well both these moves are missed as well and White (ELO 2391) plays 31. Bxe4? Nice trick perhaps if white were to take the Bishop and forget that a pawn was on d6 which would Queen! Anyway 1935 ELO Super Bullet player didn't fall for it and moved his Bishop correctly back with 31. ... Be8.

Ok last position for this game:

[fen]4b1k1/pp6/3P1p1p/8/2p1BN2/4n1P1/PP5P/4K3 w - - 0 32[/fen]

White plays Bd5+ which of course lets Black take the Bishop and free himself up with a more than decent chance for a draw.

Well all that is for naught as Black loses on time and all that effort is wasted. White notches up a freebie win by default making it probably yet another in thousands won this way.

But I am amazed how both players even managed to click through 30+ moves each in 15 seconds!! Brilliant hand/eye coordination. Chess Quality however = Absolute beginner.

Anyway still trying hard to be open minded I next looked at the same player's Bullet skills and picked his most recent game played.

Bullet Game

The Bullet game is played at 30 seconds count down. But OMG first thing I noticed was that 60 moves were played! Absolute brilliant hand with eye coordination from both players playing 2 moves per second = light year per second speed. Fantastic!!

Our profiled player in this game has an even higher rating to what he had in Super Bullet = 2593 ELO (Grandmaster class!!) his opponent has 2572 ELO (also Grandmaster Class!!!!)

So this game should be interesting! :twisted: (have to make a note to not be evil).

"GAME
#" WHITE ELO # OF MOV INAC MIST BLUN AV MIST CENTIPAWN LOSS TOTAL SCORE "GAME
#"
BULLET 2572 60 2 6 7 46.67 90 136.67
BULLET 2593 59 4 6 9 59.32 109 168.32

Wow! This is about as bad a combined player score that I have ever evaluated and I have evaluated 3000+ games so far!

Well I don't want to go through all the examples in this game, but needless to say that as soon as what ever knowledge they had early in the game went straight out of the window when the game became more complicated. Move after move it was mistake, blunder, mistake, blunder on and on an on.

Here just one example.

[fen]8/3K4/3Q1Np1/8/8/2P5/2Pq4/4k3 w - - 0 59[/fen]

White plays 59. ... Qd4???????????? I just don't know how many ? marks to place behind this move, but then again what do you expect when your hands and eyes are moving at 2 moves per second!

Conclusion: Absolute Beginner chess or worse!

By the time I finished with this one evaluation, the player completed 3 more games! :P

So yes Volodymyr is absolutely correct you can play 50,000 games. You can do this in a week even! :P

Oh by the way our profiled player lost this game on time forfeit (2 for 2 for time forfeits) So I guess what goes around comes around (sometimes).

Anyway since I lost track of where this game was as a result of the speed this guy is adding new games to his impressive 2593 ELO resume. I decided to scroll down further and there lo and behold I see that he played someone one with (GM) behind his name and rated at 2830 ELO! also at Bullet. So OK still trying to be open minded I looked at this game as well.

Another Bullet Game

"GAME
#" WHITE ELO # OF MOV INAC MIST BLUN AV MIST CENTIPAWN LOSS TOTAL SCORE
3 BULLET 2542 60 4 2 3 21.67 60 81.67
3 BULLET (SUPER GM) 2830 60 1 3 2 15.83 59 74.83

For sure this game scored better, probably at around 1100-1200 ELO level. But its an improvement.

Theory knowledge was much better but when it comes down to time crunch and complications the pattern of blunder, mistake, blunder, mistake, blunder, mistake, continues move by move.

[fen]2r3k1/3R3p/3PNppb/8/2P1p2P/6P1/5rBK/8 w - - 0 36[/fen]

Well our profiled player rated at 2542 ELO (20 minutes later it was 2593 :) ) misses a great chance to probably win this game (not sure what probably means in Bullet chess ? ) against a Bullet Super GM (2830 ELO). He needed to play 36. Rc7! instead of the move he played. Not surprising at this lightning speed it is missed and he played 36. Kg1?? and immediately Super GM has the advantage.

[fen]2r3k1/3R3p/3PNppb/8/2P1p2P/6P1/3r2B1/6K1 w - - 0 37[/fen]

Next move, our profiled player again messes up. He is already in trouble from his previous move and again missed 37. Rc7 as his best continuation. Instead he played 37. Bxe4?? after which he should deservedly lose immediately.

[fen]2r3k1/3R3p/3PNppb/8/2P1B2P/6P1/3r4/6K1 w - - 0 37[/fen]

So now its our Super GM's turn. Our profiled player (2542 ELO) just blundered away a winning game and Super GM (2830 ELO) just needs to take the pawn on c4 with 37. ... Rxc4 threatening checkmate and a massive advantage. Under self inflicted time pressure (nature of the game) Super GM totally misses this and plays 37. ... Be3?? and every bit of advantage is gone!

Well true to form in a game that now looks like being an absolute draw, our Super GM wins because of time forfeit after 38 moves.

Game Quality = @ 1100-1200 ELO player.

So what is the point of all this? I picked 3 any old games all are time forfeits and all are absolute chess garbage.

Help me someone what am I missing? Other than improving hand and eye coordination what else are you learning when you play these games? Is it the chasing after some fictitious GM title through moving faster than your opponent that is the attraction? What is it? I can't figure it out.

I seriously doubt that anyone who spends all day playing bullet or super bullet can actually improve his game through this practice. The habits that he is learning in my opinion is the habit of making mistakes over and over again.

BTW our profiled player at Standard has a downward sliding rating 1970 or something and obviously he doesn't like to play it as he has only played a few games. Standard at Lichess is 10 minutes per game. Perhaps in true over the table games this player with a downward slide from Bullet might only be rated at 1800 ELO face to face against humans? I don't know I am just surmising.

Our SuperGM (2830) has played ONLY! 2 standard (classical games) and shows a rating of 1866.... Yet he has 9788 Bullet games. Kind of proves my 1800 ELO guess for the profiled player.

In my opinion Scandien's comment is kind of correct.

I guess it is to each their own....whatever makes you happy! :)

Best regards
Nick
Post Reply