Scrabbleforumet

Diskussionsforum om Scrabble
Aktuellt datum och tid: ons 13 dec, 2017 07:23

Alla tidsangivelser är UTC + 1 timme [ Sommartid ]




Ny tråd Svara på tråd  [ 13 inlägg ] 
Författare Meddelande
 Inläggsrubrik: English Scrabble Puzzles
InläggPostat: sön 02 apr, 2017 18:55 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
I created two puzzles, which you can find in the linked document, which itself has links to the .gcg files (readable by Quackle).
https://docs.google.com/document/d/1pRSultWgO0_PzRoDnIuZleX2ubXiiZDeROs_ECAj9Uw/edit?usp=sharing
I would be grateful if someone could say whether or not you can access these puzzles; I am a forum noob :)
Please share your thoughts about these puzzles in this thread.
I will add more puzzles as I create them. Please add puzzles of your own if you wish.


Upp
 Profil  
 
InläggPostat: sön 02 apr, 2017 20:25 
Offline
Ordlistekommittén

Blev medlem: tis 05 okt, 2004 21:01
Inlägg: 960
Ort: Solna
Without using Quackle or such:

1. 12J M(I)D looks promising as I hold the only remaining hook tiles (I, S). The alternative is L10 MIND. The number of immediate points doesn´t seem to matter - unless the opponent plays a bonus and I don´t, I will win.

2. 14J ZEAtI(N) for 70 leaving OS seems strong. O1 ZA for 35 would be OK on a more open board.

_________________
Det är skönare lyss till en sträng, som brast, än att aldrig spänna en båge.


Upp
 Profil  
 
InläggPostat: sön 02 apr, 2017 21:43 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
1. I guess the probability that our opponent could find a bingo depends on their skill level, so there is a distinction between what is theoretically best and practically best and we can view these problems either way (a similar choice is whether we choose to use computer-assistance or not).
Quackle gives 0.00 % win rates from both 12J M(I)D and L10 MIND, (however, these would both be winning plays if played against me (for the reasons you gave) ).
How could we get a higher win rate if our opponent was Quackle?

2. Quackle 1.0.3 "5 min player choices list" gives these win rates:
4.17 % for 14J ZEAtI(N).
33.33 % for O1 ZA.
I interpret these win rates as corresponding to 1.5 (where 0.5 means a pick-up that results in a tie) and 12 respectively out of the 36 possible pick-ups.
So, Quackle sees the board as more bingo friendly than it seems to us. Where could the hidden bingo lane be? (I saw the game play out live; I would not know the answer otherwise).


Upp
 Profil  
 
InläggPostat: sön 02 apr, 2017 22:14 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
I have just noticed that the board is incorrect in these two problems; specifically, I forgot to add a double letter score bonus at C7 (and the other 7 symmetry-related spots).
A joy of problem creation is that I could choose to use different board layouts; however, I intend to fix this awful error and use correct board layouts from now on.
I suggest we continue to discuss these first two problems as written.


Upp
 Profil  
 
InläggPostat: sön 02 apr, 2017 22:50 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
I have added a third problem, which I hope is both challenging and possible to solve without computer assistance.
But... how can we discuss it without revealing the answer. Creating problems is so problematic :)
If you find the correct answer, please tell us without posting the answer on this thread.
If you would like hints, I will try to do this without killing the problem.


Upp
 Profil  
 
InläggPostat: sön 02 apr, 2017 23:48 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
Problem 4 has been added (access via link at top of thread).
In this game Marc played a move with a win rate of 0.00 % despite the fact that there were over 40 possible plays with a higher win rate (see below left). What would you play and why?
(Disclaimer: Most of my games against Quackle are one-sided losses; the game positions selected for these problems are not representative at all).


Upp
 Profil  
 
InläggPostat: mån 03 apr, 2017 00:26 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
Problems 5, 6 and 7 have been added (access via link at top of thread to see board position and download .gcg file) .
Problem 5 (board position not needed): With the 10 unseen tiles (ACDEEFGNOT). How many possible pick-ups are there?
Problem 6: Our opponent has just played UN(REALISE); should we deduce anything from this?
Problem 7: With so many possibilities, let’s let our sixth sense speak. What does yours say?


Upp
 Profil  
 
InläggPostat: mån 03 apr, 2017 02:27 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
I have now added Problems 8-15, which mostly re-live key moments in some of my games against Quackle.
I don't have any more problems to add; I haven't had that many key moments ;-)


Upp
 Profil  
 
InläggPostat: mån 03 apr, 2017 20:04 
Offline
Ordlistekommittén

Blev medlem: tis 05 okt, 2004 21:01
Inlägg: 960
Ort: Solna
4. I wouldn´t bother blocking. Maybe G9 (R)UE for 16 threatening ANANA+A(PE).

8. Scary. A bonus through E---E looks like it might win. Opp has a bonus too, though.

10. I'm torn between 10B J(E)T for 26 and 16L TYG. The former move gives reasonable bonus chances, the latter threatens a huge score for the J.

_________________
Det är skönare lyss till en sträng, som brast, än att aldrig spänna en båge.


Upp
 Profil  
 
InläggPostat: tis 04 apr, 2017 19:42 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
response to Radagast:
4. G9 (R)UE is Quackle's top choice; 75% win rate. I believe you can deduce which two of the eight pick-ups are losing and why.
Incredibly, picking up a fourth A to get the rack AAAANN is winning; if this does not lose, what does?

8. There is at least one bonus, but not through E---E.

10. Quackle's win rates are
41% for 10B J(E)T, and
0 % for 15L TYG,
which is a surprisingly big difference in my view. Can we explain this?


Upp
 Profil  
 
InläggPostat: tor 06 apr, 2017 20:04 
Offline
Ordlistekommittén

Blev medlem: tis 05 okt, 2004 21:01
Inlägg: 960
Ort: Solna
marc roddis skrev:
response to Radagast:
4. G9 (R)UE is Quackle's top choice; 75% win rate. I believe you can deduce which two of the eight pick-ups are losing and why.
Incredibly, picking up a fourth A to get the rack AAAANN is winning; if this does not lose, what does?

8. There is at least one bonus, but not through E---E.

10. Quackle's win rates are
41% for 10B J(E)T, and
0 % for 15L TYG,
which is a surprisingly big difference in my view. Can we explain this?


4. I did a quick scan of the eight different pickups and found that it was surprisingly hard to find a bonus ending with E. The board was blocked enough to make it hard to score with the strong-looking racks.

10. I´m not surprised about JET´s win rates, but it´s strange that TYG fails so completely.


Now for some new analyses:

12. My bonus kung fu doesn´t extend to seeing the bonuses in AABIMY?. 15A rIMY scores 64, so I settle for that.

13. The bag is bonus-prone. I want to play off at most two tiles, definitely the J. Hence 3L JE(U) for 29 looks reasonable.

14. 10A (O)M for 20. I don´t want to destroy a strong rack, and there are several bonus lanes (B and E).

_________________
Det är skönare lyss till en sträng, som brast, än att aldrig spänna en båge.


Upp
 Profil  
 
InläggPostat: tor 06 apr, 2017 21:12 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
4. Yes. In the actual game I made the lazy assumption that I should block bonuses ending in E. I now realise there was little danger. I found it really surprising how weak the strong-looking racks really were.

10. I tried some playouts with TYG and found it really surprising how ineffective it is. I failed to see TYG in the actual game; I might have played if I had seen it.

12. 15A rIMY for 60 looks great to me but Quackle doesn't like it all that much.

13. I inadvertently played off three tiles in the game; a disastrous error.

14. In the game I played off six tiles, hoping to pick up a J, H or Z for the hotspots at J6, O8 and G9 respectively. I find it hard to judge whether your bingo-approach is better; holding AEINTT also seems risky to me.


Upp
 Profil  
 
InläggPostat: tis 11 apr, 2017 15:04 
Offline

Blev medlem: sön 02 apr, 2017 17:50
Inlägg: 12
I have added 18 more pages of puzzles; use the link at the top of the thread to access all 31 pages of puzzles (you will also find links to the games themselves therein).
Puzzle 16 shows twenty alternative plays of varying quality; why are the best ones best and the worst ones worst?
Puzzles 17-20 show a total of 17 board positions where you can try your luck at making fast predictions. Good luck :)


Upp
 Profil  
 
Visa inlägg nyare än:  Sortera efter  
Ny tråd Svara på tråd  [ 13 inlägg ] 

Alla tidsangivelser är UTC + 1 timme [ Sommartid ]


Vilka är online

Användare som besöker denna kategori: Inga registrerade användare och 1 gäst


Du kan inte skapa nya trådar i denna kategori
Du kan inte svara på trådar i denna kategori
Du kan inte redigera dina inlägg i denna kategori
Du kan inte ta bort dina inlägg i denna kategori
Du kan inte bifoga filer i denna kategori

Sök efter:
Hoppa till:  
cron
Powered by phpBB® Forum Software © phpBB Group
Swedish translation by Peetra & phpBB Sweden © 2006-2010