I have been playing backgammon for a few years now, and I would say I'm somewhere between an advanced beginner and an intermediate player. I prefer to play in person on a real board, but I also use software to sharpen my game.

My question is, do any of you play human players different than "by the book" based on reading them, as in poker?

human players tend to have clear and readable tendencies that vary from the "correct" plays. You may notice a player getting rattled and over aggressive, or you may notice and opponent getting hesitant and turning down a lot of doubles they shouldn't. In this case, the doubling decision may technically call for "no double" when in fact doubling could yield more points, assuming your read on your opponent is correct.

Just wondering how many other players read their opponents and adjust their checker and cube strategy accordingly, even if it is different than what the odds say you should do.

I'm not really able at this point to distinguish between the correct play, and a play that may be "superior" based on my opponent, certainly with the checkers. With the cube it's a little different, in that I know some of the guys I play are even worse than me with the cube, and I can consequently delay offering it until I have a very winnable position, and they will still take.

One example.

In checker play, one of my primary opponents tends to play overly conservative, not leaving blots. This allows me to play more already out and aggressive.

I would say it is very rarely correct to make a play that you think is not best, because of some factor related to your current opponent.

I do think that it is OK to tailor your play to your opponent, when choosing among plays that you think are essentially equal. This is still not very common though, and your mental energy is almost always better spent finding the best plays.

I would agree to a point with checker play.

I would disagree on cube decisions. A players tendency to take or decline heavily weighs into my decision to double, if inplay that person often and am confident in my read.

Given your level of play, you'd do a lot better by concentrating on your play, rather than trying to analyze opponent's errors and tendencies. Too many intermediates justify their technical errors by claiming they adjusted to their opponent. Don't fall into this trap, otherwise your game won't progress.

That is a fair statement.

However, the nueral net models used to find the optimized moves are making some basic assumptions. Hi including, projecting out several moves ahead vs another expert player.

If your opponent has known tendencies, especially opponents you play against frequently, I would argue it makes sense to account for that. Especially in cube decisions.

Take a parallel to Texas Hold Em. You can become a great player by knowing all of the percentages, making strong value bets, etc. But you can also become a great player by becoming good at reading your opponent.

Now, this is less of a factor in backgammon, but it is still a factor. Sometimes if I have an opponent on the ropes, I will do an unconventional or surprising move just to throw them off their game.

Just like in poker, I think being unpredictable is a valuable asset when playing a human opponent.

You can't make any meaningful comparisons with poker since backgammon is a complete information game and poker is not. If this hasn't occurred to you up till now, think about it.

You are hurting your winning chances drastically by making "inventive" plays. You will do much better if you simply make the best move in every position. But this is much harder to do than to "invent" plays and justify them as adjusting for an opponent. The former takes a lot of dedicated systematic study and makes players experts of the game; the latter is the reason why most intermediates remain intermediates for their entire life. I see that it looks like you already made your choice and will not change your ways. To each his own.

I understand the sentiment, and I agree that studying best play will give greater returns in the long term.

However, there definitely are cases where the best play against a human is not the same as the best play against a bot. For example: money play, you hold the cube, and reach a position that you know is too good to redouble. But your opponent is steaming badly, and you just know he will take. By all means, send it over.

This speaks to my point of exactly. Anyone who routinely plays vs. Humans has seen a point where there opponent is "on tilt". Overly aggressive checker moves, taking too many cubes, etc. When I see that happening, I do make adjustments to my play. More-so in my cube decisons than my checker play.

I'm not speaking to off the wall "inventive" plays. In may situations, there are offensive and defensive plays that have similar expected return points. In those cases, I would rather factor in my opponents tendencies than the mathematical model output that increases my expected return by .01.

I also think very few human players can separate emotion from cube decisions. Reading your opponent's cube tendencies can provide a massive advantage.

Backgammon is a game of complete information ... I always disagree with that statement. You can see all the pieces but looking at the board in a race can you tell me what percentage you will win? I use Arthur Benjamin's Modified Kleinman count which lets me calculate the win % very accurately. I therefore have that information but my opponent does not, so really he does not have "Complete information". In the Australian champs in November 2015 my first opponent doubled me in a race. I calculated he was 67% to win. I took then redoubled him 2 moves later and he passed. I asked him what formula he used to calculate the double and he said "none, it looked like a double". If I know from memory Match Equity and Dead and Live Take points then I have more information than my opponent so we do not have equal information because it is not complete information for both of us.

Sounds like a correct double/take. This time you got a big roll or two and were able to cash. That happens sometimes, and is not really relevant to your point. (Or are you saying that his pass was wrong? That would be different.)

Some people are pretty good at eyeballing a race. You don't always need the exact count to know it is a double.

Also, you seem to be misunderstanding what "complete information" means as a term in gaming. It means that the entire current game state is available to both players; nothing is hidden. It does not mean that the players are equally good at evaluating that information. I can't analyze a chess game as far ahead as Magnus Carlson, but it is still a complete information game. All you are saying is that you are better at a particular aspect of backgammon than some of your opponents. If this was not possible, it wouldn't be much of a game!

"Complete information game" is a term that is rigorously defined in game theory. It's not what you would like this term to mean.

'Complete information" doesn't mean that both players understand the game equally well (or perfectly). It just means that no relevant information is unavailable to them. (Like your opponent's cards.)

And the way you're using it, the rigorous definition is meaningless for your argument.

I stand corrected, I did misunderstand "complete information".

One of the replies stated that most players are pretty good at eyeballing a race. I keep a running count of the difference in pips when I play. I played a lot of matches with really good players (NZ and Australian players) and asked them who they though was ahead in the race. I asked this question about several times with different good players. In all cases except one I was told one side was clearly ahead. In one case Steve Clarry (2011 Australian Champion) said "I think it is even". They could use a pip count if they wanted but most said it was obvious and they would not bother here. The kicker here is, I only asked the question when the pipcount was exactly even.

What I said was that *some* players are pretty good at eyeballing a race.

Still your experiment is interesting. Even allowing for the possibility that a few of them were trying to hustle you, the result seems meaningful. Maybe fewer have this skill than I thought.