Xg calculates luck in standard way: for N-ply luck:
- Get the current equity preroll in N-ply: E
- get the position that 1-ply would play using the dice, make a (N-1)-ply evaluation:F

luck is then L=F-E

All are normalized equities. And luck for all 36 possible dice sums up to 0.

If I'm understanding you correctly, that doesn't make sense.

Get the current equity preroll in N-ply: E (correct, IMO)

get the position that 1-ply would play using the dice, make a (N-1)-ply evaluation:F

I don't follow here. For large N (>=2 if I can think straight), the 1-ply move selection you're using for luck ratings could well not be the move the bot would actually play after looking N-ply deep (which is what's used to calculate E), so your F can be lower than it should be when the move is different.. which would bias to negative luck (for both sides) every time the best move is different at 1 and n ply.

On days when the world is against me and I sit down to play GNU, he cheats like a mother****er.
On days when I am calm, carefree and untroubled, he gives me a fair game.

a 3-ply Analyze is done as follow
for each dice combination:

• Use a 1 ply Analyze to decide what move will be played

• for the resulting position and for each Possible opponent dice, use a 1 ply Analyze to decide what move will be played

• Get the cubeless Equity from the NN

• Convert it to cubeful

Then average the 1296 possible results.

So for the formula i gave, luck averages 0 over the 36 possible rolls.

Hope it clarifies the process

That seriously makes no sense. What is a 1-ply Analyze? How do 2 of them make a 3-ply? Are you not doing any kind of minimax for move selection at >1 ply?

Hey folks...

I'm a total noob here. I found this site and forum from a google search about GNU and dancing and the seemingly miracle rolls GNU gets off the bar. He (funny I say "he" - don't know why) not only comes off the bar but seems to actually improve his board at the same time.

But I know this is simple psychology on my part; remembering the bad and taking the good for granted.

I'm a casual player at best and certainly don't have the math or stat skills to follow calculated luck results. But I have been very frustrated by GNU myself and have read about the apparent cheats online and at FIBS and with GNU.

Man! It seems to cheat. Those freakin perfect beats. But instead of all the math I simply used my own dice for a dozen or so matches. GNU obliterated me as soundly as with "random" digital dice. I then realized I could select different dice sites like random.org and GNU killed me. So I made dice files and used them as a source for GNU rolls. Still obliterated (I win one in five five-game matches and I feel good about a "Casual Player" rating and simply accept that I will always be "Awful!" on the cube.) I've even played with the dice files open in another window and verified GNU is using the file properly. I've even lost to GNU when I was monitoring the rolls beforehand. (Granted - I'm an idiot.)

I've also repeatedly done a match analysis and replayed lost matches from points where I believe I had better positions. I still lose consistently. (I remind that GNU has me at "Casual" and not always "Awful!")

So all this to simply share that I allow myself to HATE GNU and feel that I've been cheated. But I also KNOW I am not being cheated. I think we can all argue til the cows come in the validity of this point but, at least for me, the alternative dice site and manual dice (entered move-by-move so GNU cannot cheat by looking ahead in a file) is all the real proof I need. I would urge the cheat crowd to stop belly-aching about unfairness and simply accept the fact that they are probably weaker players than they think they are.

My greatest achievement in backgammon has been realizing I'm not being "beaten" but instead I'm being shown the rich complexity of such a simple game.

That's all. I'm gonna try and stay with this site. I like it.

Sorry for the ramble-on. But 'been there done that' seemed apropos :-)

Welcome to the site. It's nice to see someone with an experimental frame of mind instead of the folks we get insisting they if they lose to a bot they must have been cheated.

Don't assume you will be a weak player forever. It's not hard to improve at backgammon. Post questions here and people will help you.

playing with manual dice is the best way to prove for yourself bots don't cheat imho. This is not stupid. Many people say 'you should stop whining and grow up and it is psychology and blablablabla' but checking actually things for oneself has more value.

i have seen recently "bots cheat because they use bots".. How do u cope with that

, that's a good one, Love it! Do you know who came up with that one?

PS: I agree with Bill about typoii. It is rare to see people doing the proper home work before complaining about dice. This is the kind of attitude that will surely lead you in the right direction in improving your game.

People can claim GNU backgammon doesnt cheat all they want but no one gets the exact roll they need every single time in real life. GNU does. Also if your knocked off the board and there is three or four open spots, eight out of ten times it will take two or more rolls to get back on the board. Odd how that happens almost every time yet not that often in a real life game.
Also odd is how you seldom gets the roll you need, especially when your doubled 4 times or more.
I seen it where I was winning removing the checkers and hit 6 or more two and one die rolls while the game is able to get the rest of its checkers onto its side and get back to back doubles for the win. I seen it get up to 5 doubles in a row while all my rolls was a two and a one. No one is ever that lucky constantly which is why I say yes, the game does cheat.

Here is an idea from Russ Allbery:

The other thing that you can do is enable the tutor at the same level as GnuBg is playing, and then make sure you always follow its advice. You'll find that you then start playing the computer to a draw.

Source: http://lists.gnu.org/archive/html/bu.../msg00024.html

The key here is never to deviate from GnuBg's recommendation. You also need to keep a record of the points won and lost. If you try this and find that your long-term won/loss ratio evens out, you will know what the problem is!

BTW: To get the most out of this experiment, you will also want to check three GnuBg settings (some of which are set by default).

1. Settings > Options > Tutor: Add checkmarks to Tutor Mode, Cube Decisions, and Checker Play.
2. Settings > Options > Warning Level > Doubtful.
3. Settings > Analysis > Skill thresholds > Doubtful 0.00

If the Warning Level is not set to 0.00, then the Tutor will allow you to make small mistakes without alerting you. The instructions above enable the Tutor, and set the warning threshold to 0.00. Anytime your move is not the same as GnuBg's, the Tutor will warn you.

Write down the old threshold for Doubtful plays so that you can restore that value when you are ready to go back.

Mike

Also a naive player that doesnt scientifically observe luck will fail to recognize that part of playing better backgammon is to allow yourself not only to be able to benefit by slightly more number of future rolls than your opponent (if he were to play same rolls in your shoes) but to also capitalize on luck fluctuation better than opponent. You make it easier for good things to happen to you. This may make you look luckier but it is a result of having better positions that exploit the opponent in more ways than he can exploit you. As a result the opponent gets the feeling he is victimized because he doesnt get to convert rolls to spectacular results as often. The simplistic explanation must be cheating because it takes harder effort to actually see the truth, it requires real work instead that doesnt come naturally unless one is curious and scientifically minded.

Luck is 0 in average for any player. The fact is that 1) against the bot you dance more. 2) It has almost perfect cube skill. That means usually whenever you accept the cube from it,usually you will either a) die roughly 3/4 of the time or b) be gammoned a lot; even if the position seems harmless (in noob eyes it does!! i have a lot of problems with that). The bot WILL get AT LEAST the x% wins he claims and the y% gammons as well (see the small numbbers displayed on the screen).

If you play only DMP against the bot, he won't be as strong. It is not even impressive (although it is stiill stronger).

Advice: If you don't trust the random generator provided by the develoipers, switch to manual dice and play only with them.

They can; except you'll never compile it back to the same hashsum as the released version.

There are three posts in this thread including the OP that are denigrating, in one way or another, backgammon software, all by first time posters! Coincidence. Or not.

What does this even mean?

Are you saying that you need to "uncompile" the program you are playing in order to examine the source code? Surely you can't be saying something so ridiculous.

Are you implying that the program you are playing was not compiled from the freely available source code? Did you reverse engineer to compare? In any case, if you examine the source code and find it satisfactory, then when you compile it, does it matter if it doesn't "compile back to the same hashsum as the released version"? Just play with the new executable.

Can you hear yourself above that woolly jumper you're wearing?

If the "freely available" source code can't be compiled back to the version that's out there as the official release, that says to me that there's something in the .exe that is at odds with the "freely available" source code.

You can see that. Can't you?

And I'm saying "use your newly compiled version." The "official release" (which isn't actually a thing, btw) of an executable version is now irrelevant to you. Doesn't matter if it cheated or not.

Have you been playing against it for money? Is that why you are so hot under the collar about GNU beating up on you.

Firstly, I have no idea how I'd play for money against it? I'm sat here on my computer, I wouldn't have a clue how I play for money against it?

Secondly, I thrash GNU (repeatedly) at World Class level - more than happy to upload the games to a cloud storage system so you can witness for yourself.

Thirdly, and most importantly, the big rebuttal that Alex "look at my collection of fast cars" Choi uses on rec.games.backgammon when defending GNUDung is that "but the source code is freely available, check it if you think GNU is cheating".

I'm saying it *is* cheating and the fact that a piece of code publicised on a webpage that can't be compiled back to the same size and hashsum as the .exe is merely proof of that.

If you think GNU is cheating you, why not stop playing against it? I don't see the value of playing a program which (a) is so weak that you beat it easily and (b) is cheating to boot. Seems like your goal should be to find a stronger program that doesn't cheat.

I think you're right, thank you for your observations. I might just invest 60 bucks on XG and write my own dice rolling dll.

It's just worrying that so many gullible people (who have the vote, by the way) fall for the "but the source code is available, you can check it to see the bot isn't cheating" line when the said source code can't be recompiled back to the same filesize / hashsum as the released version.

Maybe they should be asking "why can't the source code be recompiled back to the same filesize / hashsum as the released version?".....but I guess that's an awkward question they're happier not asking.

not really sure why you continue trolling at this point, no one can be this dumb

well, I have an idea who chuckles1066 is and if I'm right, he complains for dice and claims that he outplays GnuBG for about 2 decades on r.g.b.
Just his language isn't as rude as it was. Maybe he will grow up some time...

Ok, here's your challenge.

Take the "freely available" source code.

That bit's easy, yes? Doesn't it look innocuous when you wade through the dice rolling bit?

Now try to recompile it back to the .exe file size. And hashsum.

That's the challenge.

Hint: you will *not* be able to.

No trolling, just calling it as it is.