I love backgammon and can play it endlessly. I am hooked on a particular online version, but I am suspicious about one aspect of it.
The player is able to choose his level against the computer - beginner, advanced, and expert. I started playing at the advanced level, then recently bumped myself up to expert.
Since then I’ve noticed some incredible lucky rolls by the computer and an equal number of unlucky rolls for me. For example, in the game I just finished the computer inexplicably left 3 solo stones on his home board while I was on the bar. I had three shots at hitting one of them and missed all three spots all three rolls. In fact, twice I rolled double sixes (you guessed it, he had the six point covered).
Here is my question: In designing the game, did the creators deliberately tweak the rolls of the dice to affect the player’s chances of winning or losing depending on what skill level he is playing at?
I’d like to think the dice rolls are completely random, as they should be, and I’d be pretty pissed off to learn that they are not.
I know it is probably just confirmation bias, but damn does it look sketchy.
(and if it continues I’ll probably look for another site, but I did want to hear some opinions)
mmm
The only people who know that specific game design are the developers, and I suspect they aren’t telling.
But it’s quite common, in game AIs, for the medium difficulty setting to be the best fair AI the programmers could come up with, and higher difficulties being the same AI but with progressively greater amounts of cheating.
I don’t know how many games you have played on that level. If I played a few and noticed it, I’d be open to the possibility of random bad luck within a small sample size.
If I played 6-10 backgammon games and noticed that pattern of the opponent getting luckier rolls continually happening, I would get suspicious and lose interest.
I’ve probably played 25-30 games at the advanced level. I initially chalked it up to bad luck; I’ve played enough IRL to know that some pretty crazy runs can happen.
But I’m getting more convinced that I’m being jobbed as time goes on.
I had no idea this was a thing. What bullshit.
mmm
If it was cheating, I’d venture to guess that they do something like taking three random rolls, making a best decision for each set, evaluate which of the three ends up in the strongest position, dump the other two, and display the one.
There’s probably a way to determine that it’s cheating, but since “higher is better” isn’t always true in Backgammon, it might take some doing to prove even if you keep track of every roll.
Doubles are almost always better, though, so if the computer was using the above method, then you would expect to see doubles more often than expected.
Maybe I’ll start keeping track of how many doubles the computer rolls vs. how many I roll.
I know it’s not definitive evidence - and it’s not like I could do something about it if it were (aside from not playing) - but it might be fun to know.
mmm
There’s a backgammon game that I have on my linux box that’s scarily good. When I play it on an advanced setting it’s hard to beat (and there are higher settings available), and I would swear it cheats with the rolls. The number of times I’ve failed to hit one of its blots, or that it has come back from way behind by getting high doubles in the endgame, is maddening. But I’ve heard people say that it’s just that good, and it’s only mathematically-sound strategy and confirmation bias that make it look like cheating. There’s apparently a mode where you can roll the dice manually, but I’ve never tried it.
It isn’t necessary for a computer to cheat to play backgammon extremely well unless it a crude homegrown website kind of thing. Since the advent of computers the game of backgammon has changed dramatically and the computer can make some surprising, odd-looking plays that turn out to be correct when analyzed fully.
I played with world class players and the guys who wrote the books in the early 1970s in the Miami area – the modern game is a different animal. Computers have even changed the longtime traditional opening moves.
I suspect confirmation bias. I felt the same way when first playing modern computer backgammon programs. Even though I was way out of practice (haven’t played serious backgammon since 1980 or so), it just seemed the computer was making way too many way too lucky rolls … so I started entering my own rolls and still got whooped. If the site you play has the option to enter your own rolls, do that.
I play Backgammon NJ on the iPhone. A lot of people suspect it of cheating, because its AI is so good. They’ve been accused of it so much, they even have a page set up for it on how to prove it to yourself that it’s not (you can input manual dice rolls and save the RNG seed for the game you’re playing so you can swap sides and get the computer’s dice rolls if you wish.) The backgammon AIs simply are that good.
I personally am an intermediate-to-advanced-intermediate backgammon player, and I have never noticed any fishy about it.
Putting it all together: Runs of good dice rolls will naturally happen every so often in any stream of random rolls. The good computer programs take this fact into account, and play in such a way that enables them to take full advantage of those lucky streaks, when they happen. Less-skilled players, including most humans, consider such runs to be too improbable to bother worrying about, and hence can’t take full advantage of them when they occasionally come up. Thus, when the human gets a lucky streak, it doesn’t stand out, but when the computer gets one, it does, so it looks like the computer is getting more lucky streaks than the human. Similarly, the computer does a better job of contingency planning for the eventual streaks of bad luck, and so weathers the unlucky streaks better than the human, and so it looks like the human has more unlucky streaks than the computer.
Try downloading Backgammon NJ if you can. I think the developers have gotten so sick of being accused of cheating that they’ve included every possible way to show it’s not cheating.
For any given game, you can find the RNG seed number. You can replay the game with the rolls swapped. The game itself has an option to show the source code for the random number generator. You can pick your own RNG seed. It will show you what the next 100 rolls of the dice are based on the RNG seed. You can input the dice manually. And yet, all the negative reviews on the App store are shit like “they say they’re not cheating, but I can prove they are, since the computer always seems to get doubles when it needs it.” All the rolls are predetermined once the seed is generated. You can get a printout of them. The only possible angle for cheating is the computer looking ahead at what rolls are coming and basing its play on that. I’m experienced enough that I’ve never noticed the computer making any whacked out moves that could only be explained by looking ahead in the roll chart. And, if someone is that paranoid, there is the manual dice roll option where you can roll the dice and enter the rolls into the computer. It also keeps dice roll stats for you and the computer, including average pip count, number of doubles, number of 2 doubles in a row, number of 3 doubles in a row, etc. It’s about as honest as it can be. And yet, this doesn’t satisfy the people who thinks it cheats. A lot of people just suck at backgammon and don’t understand more advanced strategy. It’s an easy game to learn, but it has a lot of subtlety and strategy to it, a lot of it not obvious.
Well, it doesn’t answer the question in my OP as to whether some game developers might deliberately tinker with the randomness of rolls at the higher levels of play.
Unless your response is saying that they would not?
mmm
I am an expert Backgammon player and play at the highest AI levels.
I don’t think the computer “cheats” but often it does seem to get lucky rolls whenever you’ve made a mistake or left an opening or yourself vulnerable. I use this suspicion as more of a learning tool.
I usually start with an aggressive blitz/priming strategy with anchor but this may change depending on the opponent’s dice rolls. I’m used to tournament or match play, so I’m not looking to win individual games but to double and/or gammon or backgammon my opponents.
I can’t remember which one it was, but there’s really good backgammon analysis software out there, too, where you can input a board position and enter a roll, and it will show you what it thinks are the best moves.
The question of “Might they?” is easy: Yes, they might. The question of “Do they?” is more difficult, and absent measures like pulykamell described, can only be answered with “I dunno”. But given that even games with all of those measures still get accused of cheating, it seems most likely that, given some other program accused of cheating, it’s most likely not.
