FIBS Board backgammon forum

Hi, back with another bear off question. I'm trying for a gammon (which I do get thanks to a 44 and 66 later) but I can't understand GNUBGs suggestion of 5/off 4/off at this point --- if I roll any 6 or 53 55 next roll I will leave a blot. (9/36 chances)

GNUs second choice of 6/1 4/off leaves a problem with 64 52 53 54 66 55 44 next throw (11/36)

my choice of 6/1 5/1 only leaves a problem with 6/4 4/4 next turn (3/36)

I imagine it may be a mistake not to take 1 or 2 pieces off if I'm going for a gammon but if white isn't going to run for a while to save the gammon why take the above risks?

Maybe she would run quickly as her home board isn't very good ... i dunno, what am i missing?

boop

Taking off checkers makes it so you win more gammons. You're right about gnu's play leaving more shots than yours, but if you get hit you still probably win since white's board is weak. I didn't put it into gnu and edit it, but if you're curious and want to learn it'd be a good idea to strengthen white's board and see what that does to things. A general rule is when gammon chances are really high or really low to play safe, when they're somewhere in the middle (like they are here) then really shoot for a gammon, and the way you do that is to take checkers off.

5/off, 4/off leaves a lot more than 9 shots. All sixes, 5-5, 5-3 is 14/36. However, playing "safe" isn't very safe, as it leaves a shot with 6-4, 4-4 and 4-3 so the difference is 9/36, perhaps that is what he meant.
How bad is an extra 9/36 shots? You only get hit with 11/36 after that, so the chances of being on the bar after the next sequence is 99/1296 or a bit under 8%. If you are hit, Red will have something between 7 and 10 checkers off, be a big favourite to enter immediately and White will have four blots spread around the board. If all else fails, the cube is still available so you can cash once your gammons disappear, but after being hit, you still win a lot of gammons here.
Early in the bearoff, you do need to think about how well the position plays in the future and later in the bearoff, safety considerations can over-ride, largely because by that time the gammon is nailed on, but there is an intermediate section, which is between six and ten checkers off where later jeopardy matters a lot less. Taking two off neatly projects us into this stage.
Roll it out manually, you'll learn a lot.

lol thanks guys .. I'm flaberghasted!!

i've rolled it out many times and have been unable to lose after playing 5/off 4/off.
I even filled up white's 6 and 5 points leaving only the 4 open and still didn't lose if it fell apart because of the cube.

I shall be rolling out many more positions in future.

b p

A Handy Guide. If you get hit and closed out with 9 men off, you are a small favourite to win. If you have 10 off, you can double. With 11 off you can redouble and with 12 off the opponent should pass. This is only a guide. You should take into account the position of the opponent's spares after the close out and the position of your own remaining checkers and of course the score, but those numbers of men off are the basis.

This is a good article http://www.gammonlife.com/writers/07tardieuart1.htm   about win chances after being hit and closed out after taking some checkers off, unfortunately the scope of the article is just for deciding if its worth risking  leaving more shots to take another checker off at DMP, so its not about gammons or cubes. Still its a good site and the chart provided is a great thing to memorize since these situations come up regularly.

limit what you have to lose...maximize what you have already gained...

...maybe? 5/off + 4/off.

IMHO

Quote from: Krazula on July 31, 2011, 11:29:03 AM
This is a good article http://www.gammonlife.com/writers/07tardieuart1.htm   about win chances after being hit and closed out after taking some checkers off, unfortunately the scope of the article is just for deciding if its worth risking  leaving more shots to take another checker off at DMP, so its not about gammons or cubes. Still its a good site and the chart provided is a great thing to memorize since these situations come up regularly.

The way I remember it is that if you're closed out on the bar with 8 checkers already borne off your game winning chance is about 50%. If you have more than 8, add ten percent for each of the first two additional checkers and seven percent for each checker thereafter.  If you have less than 8, do the same calculation, but subtract instead of add.

You'll get the following numbers:

4   5   6   7   8   9   10  11  12    checkers borne off                
16  23  30  40  50  60  70  77  84    game winning percentage      


It's off by a bit from the actual chart, but good enough for most doubling decisions.  11 off or 5 off are close take/pass decisions (you'll need to look at the actual position of the checkers) , the others should be pretty clear.  The main thing for me is that it's easy to remember - I'm awful at memorizing stuff.

related to this thread, in this game i actually passed but used gnu to play on as apparently it was a slight mistake  

I actually then won but not without some very lucky dice ... then again gnu thinks only my 66 was lucky.



as far as i can see there's a very good chance that my red on the 17point will get hit if i dance. even if i get on and move that red or better still throw 35 to cover it there's still a minefield to get through and in all likelihood a worse chance of entering if I get hit. + if i get on the 1 or 3 white can throw recyclable grenades at me ad infinitum ... plus my home is abysmal and I only have 6 off.

My question then is about the cube ... umm ... what is the cube telling me? - i mean (apart from the fact that i didn't) i'm about to lose 4 points instead of 2! .. if it was -3 -3 to 7  (4 points makes the match for either player) would it be a pass?

boop

It would absolutely be a pass at -4/-4. You can drop and be and -4/-2 where you should win about 33% (didn't check match equity table for this number, just used rick janowski's formula 85D/(T+6)+.5 for the leader's MWC, janowski's formula is usually very accurate and is a lot easier than memorizing a table, at least for me) or you could take and play on for the match, and there is no way you win more than 1 in 3 times here, so a pass would definately be right at that score. Another thing is according to your gnu's evaluation it really doesn't matter much if you take or drop this cube, so against gnu or a strong player at -7/-7 I think its clear to take, bigger cube means fewer rolls thus fewer chances for you to make mistakes. Against a weaker player at -7/-7 I think its an easy pass because then there will be more rolls in the match and more opportunities for your weak opponent to make more mistakes than you. When the choice is close its best to go for high cubes against the stronger players, and gnu is one of the strongest.

a quick thanks krazula, i'm just about to jump on a train to france for a well needed break. i'll digest your post over the next 2 weeks 
but ... i'm actually still not getting this cube equity stuff. how can nearly always being -7 -3 points behind after the take be almost the same as being -7 -5 behind?

Some points. Gnu 0-ply evaluation is not very accurate, particularly as it happens in this type of position. Gnu 2-ply will show that this is a clear pass and eXtremegammon (which is stronger) thinks that it is a big pass. This is one of those positions where White's position will improve while Black's can only deteriorate. Even though Black can't lose a gammon, it's time to say goodbye to this one.
I must disagree with Krazula that you should take this against a strong player and pass against a weak one. IMO the reverse is true! If Gnu 2-ply (a gammonbot) is playing White, then it will play the position very skilfully. Black's plays won't be too hard with the checkers, so the general effect is a wash there, but knowing when to recube to 8 is going to be crucial. You do need to have a clear idea of when it is right to turn it and very few players below expert will have that. If this was a marginal take/pass in theory, in practice weak player v. bot should pass IMO.
If on the other hand White is the weak player and it's a marginal take/pass, then it's an easy take for Black. White will quite certainly butcher this position with the checkers and may well mishandle the cube if Black gets a chance to ship it in.
Krazula's theory works well in very simple positions, races and holding games for example, very badly in complex positions where Mr Weak will get outplayed. You do need to take that into account when deciding whether to shorten the match or not.

Quotehow can nearly always being -7 -3 points behind after the take be almost the same as being -7 -5 behind?

When you pass a 2 cube you always lose two points. When you take a 2 cube that is a marginal take/pass your equity will be the same, but made up of a mixture of losses at 4 and wins at 4. If the cube is dead you will need to win one game in 4 (25%)to come out with an equity of -2, the same as if you had passed. If the cube is live, you can take with slightly less than 1 in 4 wins, typically with about 21-22% wins. This is because once you own the cube, your opponent has to win the game to get the points. He has to reach 100% in the game. Owning the cube, you don't. You can win the game with the cube, at this score, once you reach about 76% to win the game.
Think of it as a ladder, where he has to climb 100 rungs, where you only have to climb 76. Granted you have started lower on the ladder than he has, but it is still an easier target to reach!

Quote from: dorbel on August 19, 2011, 10:25:17 AM
Some points. Gnu 0-ply evaluation is not very accurate, particularly as it happens in this type of position.

sorry dorbel ... I did change my settings to 2ply last time you mentioned it but they somehow reverted without me noticing.

+ thanks again everone for the responses ... I played a lot on my holiday with someone who was better than me last summer (just before I discovered fibs).
This time I came out on top 4 out of 5 days so the improvement is clear