Reasoning??

[Yvon]

White to roll. The bot plays with Red. Could someone explain me the reasoning behind Bot's decision to take the redoubling to 4?
The score was white 2, Bot 1 at an 11 point match.

[socksey]

Because red has a great board and if red hits, it's almost a given for game.     Red has a nice forward postion in your home, so easy to avoid a gammon if things go wrong.  The match is long.  Time for makeup.

socksey




"Ideas are much like children - your own are wonderful." - Anonymous

[dorbel]

There is no reasoning that justifies Red redoubling in this position. It's an absurd cube, would be a beaver for money. Are you sure that this is the position? Was this a fibs bot?

[diane]

I think some of the newer bots are still learning to play...maybe that is the main reason.

Make the most of it, very soon they wont make mistakes like that 

[Yvon]

There is no reasoning that justifies Red redoubling in this position. It's an absurd cube, would be a beaver for money. Are you sure that this is the position? Was this a fibs bot?

No, no, it was White who redoubled whereas Red (the Bot) accepted!!That was a big surprise for me. The game continued with the cube at 4.  I mean what was the chance for the Bot winning at this position, probably something between 5 and 10%? Yes I am sure about the position in fact is a print screen image from game reviewing by 3Dfibs.
The bot was Gammonbot_IV which currently seems to be the 3rd strongest player on Fibs.
Another 2 FIBs players (both with rating at about 1950) were also watching the match.

[Yvon]

I think some of the newer bots are still learning to play...maybe that is the main reason.

Make the most of it, very soon they wont make mistakes like that 

If you are talking for BlunderBots 1 and II, who make mistakes ON PURPOSE, that would be true, but here we are talking for Gammonbot IV. These bots hardly ever do any mistakes...

[dorbel]

Oh I see, that makes a bit more sense! Well the redouble is a great play, I think over the board I would have blundered and waited a turn but redoubling is clearly correct according to Snowie. 3-ply evaluation shows that this is a nice reference position, a marginal take/pass, while a rollout likes Red's position better and makes this a tough but clear take.

It looks like you way underestimate Red's winning chances. He has several ways to win from here. He can still win the race even if there is no hitting. White is 25 up and on roll, so a big favourite in the race, but not a lock. If this appears very unlikely to you, White will have to play to avoid being hit of course, so will usually have to bury several checkers on the low points on the bear in, thus wasting pips. Red can hit a indirect in the outfield while White is bringing the two stragglers home. They may well take a few turns to get safe. Of course they are quite useful to White as they cover the outfield nicely and make it hard for Red to step out, but they are vulnerable and Red does have a four point board. When they do get safe, White still has to clear the 9, 8 and 6pts, fairly easy, but there's many a slip twixt cup and lip as they say, and Red will be itching to winch an 8 cube into position.

So, great cube action by you, correct take for the bot.

[diane]

No, no, it was White who redoubled whereas Red (the Bot) accepted!!

Ok, that is a different story entirely, between your original post and this one, it was slightly confusing. 

I didn't mean a purposeful blunder, some of the bots are new, and as they learn they make mistakes.  You are right though, Gammonbot IV is past that now.  I don't know what you mean the 3rd strongest on fibs, as far as I know, the Gammonbots are all equal strength, just on different winning and losing cycles, affecting their ratings a bit here or there. 

As for taking, looks like a take to me, lots of chances to hit and a strong board if he does hit.

[Yvon]

Oh I see, that makes a bit more sense! Well the redouble is a great play, I think over the board I would have blundered and waited a turn but redoubling is clearly correct according to Snowie. 3-ply evaluation shows that this is a nice reference position, a marginal take/pass, while a rollout likes Red's position better and makes this a tough but clear take.

It looks like you way underestimate Red's winning chances. He has several ways to win from here. He can still win the race even if there is no hitting. White is 25 up and on roll, so a big favourite in the race, but not a lock. If this appears very unlikely to you, White will have to play to avoid being hit of course, so will usually have to bury several checkers on the low points on the bear in, thus wasting pips. Red can hit a indirect in the outfield while White is bringing the two stragglers home. They may well take a few turns to get safe. Of course they are quite useful to White as they cover the outfield nicely and make it hard for Red to step out, but they are vulnerable and Red does have a four point board. When they do get safe, White still has to clear the 9, 8 and 6pts, fairly easy, but there's many a slip twixt cup and lip as they say, and Red will be itching to winch an 8 cube into position.

So, great cube action by you, correct take for the bot.

Of course nothing is impossible in backgammon even for Red to have won the race as you said, although at current position the pips count is 103-128 and it's whites turn to roll -that would increase the difference even more. I am more interested to know how Snowie decided it was a clear take, in terms of what was the percentage chance of winning for each side.Thanks.

[dorbel]

According to a rollout, Snowie thinks that Red can win 23.4%, losing 5.3% gammons and winning 2.4%. Seems reasonable to me. Over the board I would have thought it was a bit more, hence my reluctance to redouble, but there are some serious flaws in Red's position. The front-loaded home board (i.e. the spares on the low points) is weak and Red's stack on the anchor is a problem. These factors combined allow Red almost no choice in how to play next. He has to run out with fives and sixes and busts his board with 4-4 and 3-3.
A world class human playing a weaker player who assessed this position correctly might even decide to pass, as White's position is very easy to play and most of Red's plays for a turn or two are forced. However, bots assume that their opponent is of equal strength, so can't make this adjustment.

[Yvon]

According to a rollout, Snowie thinks that Red can win 23.4%, losing 5.3% gammons and winning 2.4%. Seems reasonable to me. Over the board I would have thought it was a bit more, hence my reluctance to redouble, but there are some serious flaws in Red's position. The front-loaded home board (i.e. the spares on the low points) is weak and Red's stack on the anchor is a problem. These factors combined allow Red almost no choice in how to play next. He has to run out with fives and sixes and busts his board with 4-4 and 3-3.
A world class human playing a weaker player who assessed this position correctly might even decide to pass, as White's position is very easy to play and most of Red's plays for a turn or two are forced. However, bots assume that their opponent is of equal strength, so can't make this adjustment.

It's a bit of strange to me how Snowie decided it was a marginal case of take at a mere 23.4% chance to win.Is this the rule, I mean does ANYBODY take (even marginally) when the chance is only 23% and the score is just 2-1 in an 11 point match?? God forbit ! Of course everything you said about white's position is correct it ALL however sums up to the percentage chance of winning, isn't it?

[sorrytigger]

Maybe an answer as simple as this can help you:

in a moneygame you want to take with 25% winning chance.

Good reasoning for this astonishing fact can be found here:
http://www.bkgm.com/articles/GOL/Jul99/hank2x.htm

Or take this brief explanation:
Taken you rigorously pass with a winning chance of 25% and do that in 4 games, you will lose 4 points/pounds/dollars/euros/drachmes.
If you take those games, you will lose 3 of them, paying 6 points, and will win one with 2 points, resulting in a loss of 4 points alltogether.
No gain, no loss. Being any better than 25% makes you the winner in the long run.

To me this is the key to understanding cube handling. Gammons and matches make things a little more complicated but don't shake this central position.

Andreas

[dorbel]

Depends how much cube ownership is worth. With so many moves to go it must be worth a lot here. If Red can keep his board intact a hit/fan sequence is a winner owning the cube. In money play it is quite common to take down as deep as the 21-22% range if the cube is live.
If you have difficulty seeing that this is a take, good idea to roll it out yourself, say 50 times. You'll be surprised how long Red stays alive in this game.

[Yvon]

I guess the last 2 posts sum it up.
Thx Dorbel and sorrytiger.

The funny thing about it is that GammonBot IX did win this game, but not in any complex way, in a very simple manner in fact. Here's how:

14)  Takes                       35: 13/8 18/15
15) 22: 20/18 20/18 5/3 4/2      23: 8/6 6/3
16) 13: 6/5 5/2                  33: 15/12 12/9 9/6 4/1
17) 44: 18/14 14/10 10/6 18/14   66: 9/3 9/3 8/2 8/2
18) 32: 20/17 3/1                46:
19)  Doubles => 8                 Drops

It really pissed me off since that 66 was the one and only possible bad roll at the moment (chance 1:36) and the hit that followed a 1:3 chance.

The whole match went on like that in fact i was forced to redouble to 8 in the next game having similar low chances and surprise-surprise i won !!