WHAT DO YOU KNOW ON
LEARNING ENGINES/FEATURE ?
ranking

Enter your name :

Q1 - When should you train an engine on the "key" positions of an opening ?

1 = when you want to invert the trend of the win percent of an opening
2 = when you don't know the win percent of an opening
3 = when the engine has to learn a long opening
4 = when they are few "key" positions

Q2 - Choose the best settings for learning an opening :

1 = Experience MultiPV = true, Experience Book = true, Experience Read Only = true
2 = Experience MultiPV = false, Experience Book = false, Experience Read Only = false
3 = Experience MultiPV = false, Experience Book = true, Experience Read Only = true
4 = Experience MultiPV = true, Experience Book = false, Experience Read Only = true

Q3 - Choose the best sparring-partner for a learning engine :

1 = an engine using an opening book
2 = an engine without learning feature
3 = the same learning engine
4 = another learning engine

Q4 - What kind of openings can produce this learning curve ?

1 = short openings
2 = unbalanced openings
3 = drawish openings
4 = long openings

Q5 - Choose the best configuration to learn an opening :

1 = TC 60m+60s, 500 games/opening, concurrency = 1
2 = TC 1m+1s, 500 games/opening, concurrency = 1
3 = TC 10s+0.1s, 10k games/opening, concurrency > 1
4 = TC 2m+2s, 2000 games/opening, concurrency = 1

Q6 - When should you reinforce your experience data ?

1 = before your engine starts to learn an opening
2 = after your engine won games
3 = before your engine plays on chess servers
4 = after you analysed the lost games of your engine

Q7 - Choose the best settings for playing a tourney with learning engines :

1 = Experience Book = true, Eval Importance = 1, Min Depth = 4
2 = Experience Book = false
3 = Experience Book = true, Eval Importance = 10, Min Depth = 40
4 = Experience Book = true, Eval Importance = 5, Min Depth = 27

Q8 - After learning 2000 games here, what is the best recommendation ?

1 = learn another 2000 games
2 = use training on "key" positions
3 = do nothing, it's a drawish opening
4 = reinforce experience data

Q9 - Choose the right statement :

1 = it is advisable to train your engine from the starting position
2 = the engines don't learn anything in drawish openings
3 = it is advisable to train your engine with openings which contain 6-8 moves
4 = it is advisable to only train openings where your engine has lost games

Q10 - What does a low conformity with experience data mean ?

1 = the experience data was of poor quality
2 = the learning engine didn't play the most effective moves
3 = the tourney's openings were too long
4 = the learning engine wasn't trained on the tourney's openings

Q11 - Which experience file format hasn't a header ?

1 = SugaR Experience version 2 file format
2 = BrainLearn-like experience file format
3 = none
4 = Experience file format v3.0

Q12 - Choose the right statement :

1 = it was 2 different sparring-partners
2 = there was only one sparring-partner who changed his playing style
3 = it was 3 different sparring-partners
4 = it was only one sparring-partners who does not learn

Q13 - When an engine learns an opening :

1 = the game's average depth increases
2 = the game's average depth varies little
3 = the game's average depth decreases
4 = the game's average depth varies a lot

Q14 - Choose the right statement :

1 = several learning engines can read the same experience file at the same time
2 = opening's books contain engine scores
3 = several learning engines can write on the same experience file at the same time
4 = engines can ignore some opening's books

Q15 - Choose the right statement :

1 = there are 32 bytes/entry with experience file format v1
2 = there are 16 bytes/entry with SugaR experience file format v2
3 = all learning engines have a concurrency experience option
4 = any learning engine can use its experience file as book

Q16 - Among its 4 sparring-partners, how many had a learning feature ?

1 = none
2 = at best, only 2 sparring-partners
3 = at least, 3 sparring-partners
4 = all its sparring-partners

Q17 - When should you train your engine with the Self Q-learning method ?

1 = when you have little time to train your engine
2 = when you have a large experience file
3 = when you want to keep your engine scores on your experience file
4 = when you plan to train endlessly your engine

Q18 - What does the experience rate indicate ?

1 = the amount of openings stored in the experience file
2 = if the experience file has already played an opening
3 = the number of openings learned from an opening's list
4 = if the experience file is old

Q19 - What average percentage of short openings (non-drawish, 1-15 plies) is learned by training 2000 games ?

1 = 20% (32th / 160 plies)
2 = 25% (42th / 167 plies)
3 = 33% (38th / 115 plies)
4 = 40% (39th / 98 plies)

Q20 - What average percentage of medium openings (non-drawish, 16-31 plies) is learned by training 2000 games ?

1 = 28% (40th / 143 plies)
2 = 34% (36th / 106 plies)
3 = 41% (46th / 112 plies)
4 = 50% (53th / 106 plies)

Q21 - What settings force the engine to play Bd3 ?

1 = Experience Book = true, Experience Book Eval Importance = 2, Experience Book Min Depth = 27
2 = Experience Book = true, Experience Book Eval Importance = 9, Experience Book Min Depth = 30
3 = Experience Book = true, Experience Book Eval Importance = 1, Experience Book Min Depth = 27
4 = Experience Book = true, Experience Book Eval importance = 5, Experience Book Min Depth = 29

Q22 - Which experience file format has the smallest entry ?

1 = Experience format v3.0
2 = ShashChess experience bin
3 = SugaR experience v1
4 = BrainLearn experience bin

Q23 - From a drawish opening, how many moves can an engine learn with 2000 games @ TC 1m+1s ?

1 = 30 to 300
2 = 300 to 3'000
3 = 3'000 to 30'000
4 = 30'000 to 300'000

Q24 - From a non-drawish opening, how many moves can an engine learn with 2000 games @ TC 1m+1s ?

1 = 10'000 to 40'000
2 = 40'000 to 70'000
3 = 70'000 to 100'000
4 = 100'000 to 130'000