Opening 4-3 and alike
Peter Backgren
While we are on the topic of openings...
Lately I have noticed that I've developed a preference for playing the
opening 4-3 like 24-21/13-9. Now the opening 4-1 seems to get the same
treatment from me. Could anyone provide me with the opening rollouts for
those two?
The main reason I've started to dislike splitting to my opponents 5 is
that any piece I have there is such a prime target. Having split in front
of his 5 discourages him from slotting it, prevents him from hitting me
AND slotting/covering it, still presents me with a decent anchor
possibility and finally lets me play some of my awkward low rolls on my
side of the board creating a decent home/prime.
Playing like this has increased my wins over the Fish with 5% or so (3.5
Light, lvl7, 60-40 in my favour, however only thirty 7 pts played since I
started realizing I was opening differently). Apart from the obvious,
"training makes you better" I can think of several other reasons:
1) Back-and-forth hitting around the 5 results in backgames and timing
where the Fish is better than me.
2) An anchor on the 4 makes me play a better priming and timing game.
3) The equity loss from my new preference is so small it doesn't matter.
Comments?
Eskimo
Simon Howard
usually use for 43 is 13/10, 13/9 because that is exremely agressive and is
a brilliant basis for building strong boards. However, its success is all
dependent on whether my opponent returns with an 8 roll or 9 roll.
As for 41, I agree with you. 13/9, 24/23 helps build a board with little
danger and it splits the back men. I would say, the only other possibility
is 13/8, which I occasionally use if I want to play safe.
So I would say that 43 - 13/9, 13/10
41 - 13/9, 24/21 are the best moves.
This is what JellyFish Rollout said after 100 games:
43 - 13/10, 13/9:
WINS G/BG BG
Opponent 41.7 9.5 0.7
You 58.3 17.3 1.2
Your equity: 0.247
43 - 13/9, 24/21 (your way)
WINS G/BG BG
Opponent 42.8 9.7 0.3
You 57.2 17.2 0.8
Your equity: 0.224
This concludes that 13/10, 13/9 is definitely the best move for 43. 43 turns
out to be a very good roll as the match equity (on average) is 0.224 which
is very high.
41 - 13/9, 24/23
WINS G/BG BG
Opponent 43.7 10.6 0.9
You 56.3 16.4 0.7
Your equity: 0.182
41 - 13/8
WINS G/BG BG
Opponent 47.3 11.1 0.4
You 52.7 14.4 0.7
Your equity: 0.090
This clearly shows that we were right. You are far more likely to win the
game and even more likely to win the match.
--
Simon
Gregg Cattanach
I'm not sure we can say from 100 Jellyfish rollouts that 13/9 13/10 is
'best'.
Doing a recursive rollout set, (rolling out all of the 2nd rolls, then
accumulating the value of the best responses to determine the best first
move), and using Snowie variance reduction rollouts, 3-ply, with equivalent
~150,000-200,000 games per 2nd move I have these results:
For money: 13/9 13/10 .002
24/21 13/9 .007
24/20 13/10 .005
24/20 24/21 -.007
At Gammon Go, however: 13/9 13/10 .270
24/21 13/9 .260
24/10 13/10 .251
24/20 24/21 .224
so for money, the 3 most popular moves are really a tie, (.005 is
statistically insignificant). However, when gammons count for you, and not
for your opponent, bringing two builders down is best, (which is usually the
case when two down is an option.)
My preference is 24/21 13/9 because the builder on the 9 point is more
useful, (making the 9 point is good if you don't roll an innerboard point
making combo, and it is less provocative than slotting the 20 point, and my
opponent is a bit less likely to slash away at his 4 point than his 5 point.
Gregg C.
"Simon Howard" <si...@howard666.freeserve.co.uk> wrote in message
news:8mh1d3$mj4$1...@news8.svr.pol.co.uk...
Simon Howard
100 games in the Level 6 (1-ply) rollout which is the EQUIVALENT OF 2041
GAMES.
And our results seem to confirm each other anyway.
Simon
Robert-Jan Veldhuizen
<si...@howard666.freeserve.co.uk> wrote:
>Okay then
>100 games in the Level 6 (1-ply) rollout which is the EQUIVALENT OF 2041
>GAMES.
Not trying to nit-pick or something, but that's still not very much for
plays that are likely to be very close. I beleive JF has some kind of
standard deviation value where you can see what margin of (statistical)
error such a roll-out gives in this position.
>And our results seem to confirm each other anyway.
Well, I think it's very important to note that you rolled out the plays
*for a certain matchscore* (presumably something like 2-away 1-away
Crawford or something similar) and not for 0-0 or for money.
At 'normal' matchscores and for money, your equity is definitely not
>0.2 after an opening 4-3...
What's more, at normal matchscores and for money, 13/10 13/9 is not so
much better than either 24/21 13/9, 24/20 13/10 and even 24/20 24/21. I
think the bots like the split better actually, than bringing two
builders down like you suggest "is much better".
Except of course...when you want gammons more than your opponent, like
Gregg already pointed out.
--
Robert-Jan/Zorba
dmg
For what it's worth, here are some rollout results for the opening 4-3
from other sources:
Michael Manolios shared his rollout results with the group a while
back. He used JellyFish 3.0, level 6, and rolled out 9076 games
(equivalent to > 32000 games):
Move Equity Wins Gammons Opp. Gammons Std. Dev.
=========== ====== ==== ======= ============ =========
24/20 13/10 0.006 50.2 13.0 12.7 3 3
13/9 13/10 0.005 49.7 14.8 13.6 3 3
13/9 24/21 0.003 49.7 13.6 12.8 3 3
Achim Müller rolled out 12960 games with JellyFish, level 5, producing
the following:
Move Equity Wins Gammons Opp. Gammons Std. Dev.
============ ====== ==== ======= ============ =========
24/20, 13/10 0,007 50,1 12,0 11,7 0,003
24/21, 13/9 0,015 50,3 12.5 11,5 0,003
13/9, 13/10 0,003 49,5 13.4 12,3 0,003
Torsten Schoop doesn't provide the data on his pages, but his
JellyFish and Snowie rollouts rank the plays in this order:
13/9, 24/21
13/9, 13/10
13/10, 24/20
Midas (Sorry, don't know his real name) rolled out 1296 games (SD=8)
using JellyFish 3.0: He also didn't provide complete stats, but he
shared the following with the group:
Move Equity
=========== ======
13/9 24/21 -.003
13/9 13/10 -.004
24/20 13/10 +.009
24/20 24/21 +.003
Dean
("Chase" on FIBS and GamesGrid)
_______
To respond via email, replace "USERNAME" with "demiga" in my address.
Simon Howard
41. And I told him! OK the figures might be different for the different
match positions but quite frankly, I don't have time to do all those
rollouts so I just did it for a 1 point match. The best move doesn't change
no matter what the score so all I am saying is that the best move for 43 is
13/9, 13/10 and the best move for 41 is 13/9, 24/23. Ok?
--
Simon
Robert-Jan Veldhuizen
<si...@howard666.freeserve.co.uk> wrote:
>For heaven's sake, all Peter asked was which move was the best for 43 and
>41. And I told him! OK the figures might be different for the different
>match positions but quite frankly, I don't have time to do all those
>rollouts so I just did it for a 1 point match. The best move doesn't change
>no matter what the score
That's absolutely not true (for heaven's sake?) :-) Different
matchscores give different values to gammon wins/losses and therefore
the best move DOES change.
> so all I am saying is that the best move for 43 is
>13/9, 13/10
If you read the other articles carefully, you can see that this is just
not right, or at least not the whole truth.
>and the best move for 41 is 13/9, 24/23. Ok?
Your enthusiasm is encouraged, but your figures just can't be right and
some of your conclusions (therefore) aren't right either. I guess you've
made some error in your roll-out somewhere. I'm not trying to pick on
you (and I'm sure neither are the others here), I'm just trying to
correct some possible errors that you probably have made, hopefully
it'll help you (and others).
Perhaps you set up the board positions *after* playing the opening roll
and then gave the roll to the opener *again* during the roll-out? That
could explain your equity figures I think:
>43 - 13/10, 13/9:
> WINS G/BG BG
>Opponent 41.7 9.5 0.7
>You 58.3 17.3 1.2
>Your equity: 0.247
>[...]
>This concludes that 13/10, 13/9 is definitely the best move for 43. 43 turns
>out to be a very good roll as the match equity (on average) is 0.224 which
>is very high.
Something just must have gone wrong here, 43 is definitely NOT a very
good roll for the opening; rather a mediocre one. The real equity after
a 43 opening roll in a 1pt match is very close to 0.000 (as you can see
in the other articles), whether played 13/10 13/9, 24/20 13/10, 24/21
13/9 or 24/21 24/20.
The best move is not always 13/10 13/9, according to the roll outs; you
can see that in Dean's and Gregg's articles.
For the 4-1 roll opening moves, your figures just can't be right either.
While I agree 24/23 13/9 is the best move, the equity values you gave
are out of the question. For what I know, 4-1 is in fact a bad opening
roll which leaves you with a negative equity, although also very close
to 0.000. Definitely not the 0.180 your rollout came up with; I suppose
a similar thing like the wrong player on roll happened there? Just
trying to sort things out :-)
--
Robert-Jan/Zorba
dmg
<si...@howard666.freeserve.co.uk> wrote:
|For heaven's sake, all Peter asked was which move was the best for 43 and
|41. And I told him! OK the figures might be different for the different
|match positions but quite frankly, I don't have time to do all those
|rollouts so I just did it for a 1 point match. The best move doesn't change
|no matter what the score so all I am saying is that the best move for 43 is
|13/9, 13/10 and the best move for 41 is 13/9, 24/23. Ok?
For several of the opening rolls, rollouts indicate that the best move
DOES change depending on the match score. 43 and 41 happen to be two
of those rolls. For an opening 43, at an even match score, splitting
the back men is considered slightly better than bringing two down from
the midpoint. At 2-away; 1-away, however, 13/9 13/10 generally comes
out slightly better, presumably because it leads to more gammonish
positions (for both players), but a gammon doesn't help your opponent.
For an opening 41, 13/9 24/23 is generally considered best, but at
2-away; 1-away 24/20 24/23 often slips up into the top spot. Other
rolls whose best move is dependant on the match score are 64 (this one
probably changes the most), 21, and possibly 52.
Peter Backgren
says...
Thank you very much everyone. First of all I wanted the confirmation that
I'm not throwing away a lot of equity here. Secondly, while I thought
about gammons I didn't consider match scores closely enough.
--
//------------------------------
//Remove nospam to mail directly.
//Ascended:WVPT
//In progress:K