FIBS Board backgammon forum

the bots get more 66s or 55s to jump primes and take the lead in a race than us non experts ... this is because they take more cubes than an average player would and therefore throw more dice while playing from behind. A double 5 or a double 6 happens once every 18 throws.


taken from boop's pearls of backgammon wisdom published by Calm Down Press ©2022

b p

I am not sure about that logic..I take all kinds of bad cubes...but dont get the 55s and 66s to save me...

Quotethis is because they take more cubes than an average player would

There is no evidence to suggest that this is true that I am aware of. Source?

QuoteI take all kinds of bad cubes...but dont get the 55s and 66s to save me... 

Bots don't take bad cubes. That is the point. If you correctly pass, you won't need saving.

It is a common error to suppose that rolling a prime home against an anchor is a sure win. It isn't. The defender will usually have a substantial amount of equity left. If they do escape a prime with a large set and then win the race, then that is part of the equity that made them take in the first place.

well, if bots only take good cubes and I frequently either pass when i should take, or take when I should pass (as the cube confuses the hell out of me) , then I'm often either

1) not in the game having passed and so i won't get to see the winning sets that I would have thrown ( but I certainly see and REMEMBER when the bot throws them)

or

2) I'm fighting a lost battle where my sets don't work

therefore the bots are seen to get more 66s or 55s and jump primes to take the lead in a race .. and it's all due to their superior backgammon skills


so, the title of this thread stands, at least in my mind ... until you tell me it doesn't

b    p

I didn't challenge the title of your post, but since you raise the matter, where is the evidence for it?
Perhaps you can produce this alongside your justification for "they take more cubes than an average player".
They might take more or less, I don't know, but as you point out yourself, the average player probably takes as many passes as he passes takes.
As for "the bots are seen to get more big sets and jump primes and take the lead in the race", well anecdotal observation of what happens in backgammon games is valueless. Invariably entirely subjective, you "see" whatever your preconception wants you to see. You remember the times that it does happen and forget the rolls when it doesn't. You can test this for yourself. Make up any bug in the rng that you care to think of, say "Every time I'm in the air and my opponent has just the 6 and 4pts made, I dance the first time I roll". You will immediately see this happening over and over again.

Here however is a useful tip to improve your cube action against bots. They rarely lose their market, so you can usually take a bot cube. The exceptions are when it has been winning the game and cubes immediately after a joker sequence.

hehe!!

I didn't challenge the title of your post, but since you raise the matter, where is the evidence for it?

maths .. see below

Perhaps you can produce this alongside your justification for "they take more cubes than an average player".

nope I can't justify that ... I should have said "they make better cube decisions than an average player".


my maths:

let's say that I incorrectly pass X% of my cubes. that means the bot plays X% more games past the cube than me. In some of these extra games, the bot will throw a 66 or 55 to jump my prime. So "the bots get more 66s or 55s to jump primes and take the lead in a race ..."

of course they also throw lots of other annoyingly good dice simply because they're still in the game when they should be (when often I am not) as this is a subset of the rule  "the bots seem to get more luck because they play better."

btw I'm not saying that the bots throw more doubles than me percentage wise.

p.s. how much I should take into account the strength of the opponent in my cube decisions ... I've been following your previous tip and taking more bot cubes but I seem to be  failing miserably. Is there a rule of thumb?

Yes I can follow that. When the bots get a big set to jump the prime and win the race it is of course only part of their equity in the position. They may also win in the more common way of just getting a shot and hitting it and winning from there, usually with a perfectly timed cube.
A "weak" player who regularly passes takes will as you say never get to see these "miracle" turnarouns on his side. However, most weak players make as many bad takes as bad passes!
Altering cube action to maximise equity against bots. First do your best to decide whether it is an easy take or pass, or if it is you ahead, whether the double is theoretically right. A good tip here. Bots rarely lose their market, so if they haven't just thrown a joker, you probably should have a take! Then divide up the games into two types, simple positions that don't need a lot of skill and complex positions likely to need a lot of skill. In the first category, basically pure races and simple holding games, you need to double a lot earlier than usual. The bot can't outplay you by much, if at all, so take advantage of that and hope to get lucky. You can take a bit deeper too, as long as you are prepared to crank it up to the next level if the game swings in your favour. In the second category, more or less everything else, you are very unlikely to be able to play well enough to attain the theoretical equity of the position, so double a bit more conservatively and be prepared to drop more readily.
There is a third category of position that you can exploit, one where one side's play is easy, the other's hard.
When you have had a checker hit, but you have re-entered and the bot still has men back for example, your play is going to be easy, often forced, his will be quite hard, so you can take a bit deeper than usual and if you have the cube, use it aggressively. Switch sides so that you have the hard side and you'll have to pass most of the time.
Hope this helps. Dont be too depressed losing to gammonbots. If you can get anywhere near 40% wins against them you are playing pretty well.
Hope this helps.

Quote from: boop on September 05, 2011, 03:56:00 PM


my maths:

let's say that I incorrectly pass X% of my cubes. that means the bot plays X% more games past the cube than me. In some of these extra games, the bot will throw a 66 or 55 to jump my prime. So "the bots get more 66s or 55s to jump primes and take the lead in a race ..."

But you also incorrectly take Y% of your cubes, so you play Y% more games past the cube than the bot.

If X > Y, then I follow your logic.  But you've shown no evidence that your incorrect passes are more frequent than your incorrect takes.  It may be true, and it might be true of humans in general, but I'd need to see some evidence supporting it.

yes I think X>Y

i did think about the incorrect takes, and my feeling was that they don't even out the incorrect passes although I wasn't certain how to express it in my last post but it was along these lines -

after a bad take -

1) you're less likely to be able to use your 66 55. An example being you're 3 away from the edge of a 4 prime.

2) you can't use your big doubles because you'd still be too far behind in the race.

3)  similar to point 2) above, you're forced to use your doubles to enter an unwinable race when you were actually trying to wait for a blot to hit.

Also, along the lines of "two wrongs don't make a right" I can't imagine how it could be possible that "2 wrongs cancel out"



+ thanks for the above info Dorbel

a qualification to my previous post

Also, along the lines of "two wrongs don't make a right" I can't imagine how it could be possible that "2 wrongs cancel out"  on average

You are right, wrong passes (of takes) and wrong takes (of passes) don't cancel each other out, they add. The cumulative effect is a train wreck. However the bold takes do allow you to see what goes on after the take and you learn something, an opportunity denied to the weak passer. In general though, bad takes are very expensive. If you want to be a gambler, devote yourself to cubing earlier. You may as well make your cube mistakes, if indeed they are mistakes, in the games that you are winning.
Writing 30 years ago, Kent Goulding wrote that the commonest mistake among all players at all levels is not cubing early enough. This is pre-bot don't forget, but the bots and the top players who have grown up with them, cube very early indeed. As with the bold taker, you learn a lot from these. You can tame down your game if you are too aggressive with the cube, but it is very very hard to make the passive player, who passes weakly and doubles too late, into something more predatory, precisely because he denies himself learning opportunities.
Kent's summary of human cube action is still true today. Liven it up out there!