If the human deviates too far from standard, recognizable "good play" in trying to exploit this knowledge isn't he clearly opening himself up to being exploited in return? Swapping mistakes and all that. Would the computer's decisions and edge (or lack thereof) be affected that strongly by identifying and understanding that it is playing against a computer that has been 'hardwired' to potentially be capricious?

I think what David is trying to get at here is the difference between logic and reason. Humans are profoundly good at reasoning but often make logical mistakes. Computers however, when programmed correctly, are incapable of making errors in logic, but they are quite poor at reasoning.

One good example of the difference between the two is that logic is the 'map of everything' and reasoning is your ability to read that map and get where you want to go. You can have the best logic in the world, ie where point a on your map is 100% correct and correspondent to where point a is in the real world; but if you are trying to use your map for directions and you aren't able to reason that on your 'logic map of everything' point a leads to point b via pathway c, your logic is useless. This is where computers get bad at poker relative to humans.

Computers are much better at observing the data set of poker and drawing conclusions based on those data; humans are much better at using said conclusions to create an advantage over an opponent

So, I think that David's point is that the computers we have today have pretty minimal reasoning abilities and barring a paradigm shift in computer science, humans will always be the best given the parameters that he provided.

I think any on-line poker game that has enough money and traffic involved to represent a profit opportunity for bots will be attacked by them. The more complex and multi-way, the bigger advantage a well-constructed bot will have over a human. The games won't be ruined if there are enough players - especially bad ones - alongside the bots. I don't agree that great players will outperform bots in the long run, even when there are bad players involved. At one point I thought that humans would keep ahead of computers in chess for similar reasons, but the increasing power of computers, especially when combined with genetic algorithms, just gets there in the long run. In a few hours, I can test a computer program against more poker hands than on on-line multi-tabler could play in a lifetime.

The software will not care what their opponent knows or doesn't know - they will base their adjustment only on observed plays. Expert human players are very good at balancing their play, but even the best are not perfect. A good software algorithm can weigh the expected advantage of deviating from GTO to exploit some players, against the expected cost of such deviation due to exploitation from others who observe its non-optimal strategy. These are examples of things that are fairly easy in software and too difficult for a human player to calculate on the fly. The best players have great "feel" for such things, and will come close - I expect bots to make most of their money from lower tiers of players.

My big concern is whether live play will end up being destroyed in a similar way, by cheats who use concealed technology to assist their play. We had blackjack computers in shoes years ago. At the moment Google Glass is easy to spot, but the next generations of such technologies may be very hard to keep out of the game.

I'm not concerned. An invisible computer, capable of parsing real world visual and audio input and plugging it into advanced ai in real time and then whispering sweet nothings into your ear about the true nature of things and when to check raise the river, sounds awsome. Poker be damned.

Here it is

From what i read in the article the bot is supposed to be able to keep doing it's strategy without having to deviate and still crush it's opponents. They said the perfect strategy (which they tried to get closest too) cannot be manipulated by villains changing strategy's.

The chip robot takes the cake! hahahaha easy....easy...bam bam bam bam

How about the softwear of me sitting in my silk pajamas at my PC, scratching my unmentionables while sipping earl grey, yawning off a hangover while my bot crushes souls, while Im searching for used Lambos on autotrader.

The more I think about this, the more I think that online poker can't ever recover from Black Friday. The bots are getting so good, and what you risk by getting caught (account frozen/banned, no criminal proceedings) is so heavily outweighed by the reward for success (printing money with zero effort) that its inevitable the online games will be ruined, if they arent already.

In the future, online poker could actually be the very cutting edge of AI developments, where legions of programmers and researchers compete to see who can design the ideal virtual poker player. Maybe when you log into Pokerstars in 2020, you will have the option to play as yourself or purchase/lease a bot who plays for you. As far as the site is concerned, as long as you pay rake, who cares right?

Ongame was infested with bots at full and short tables for years. The bots were beating the games from the lowest levels up to about the same 1k level. I once told a buddy that a player had to be a bot because he was doing the same thing as the bots sitting at all full ring games from 4NL to 200NL. Why would anybody sit at 4NL if they have the roll for 200NL? It's too easy to time out or get the hands wrong on the higher level. Only a bot program would do this.

Best post of thread so far...
I don't know whether here or in another thread, an user posted about UK law, basically if you are cheating you could get a criminal penalty.
I guess this is the only way to deal with this issue

Thanks for linking the article OP... subbing for the discussion.

"It would appear that we have reached the limits of what it is possible to achieve with computer technology, although one should be careful with such statements, as they tend to sound pretty silly in 5 years.”

-John Von Neumann

I agree.

Yes, paradigm shifts happen.

"Poker is dead"

2p2 since 2009

I imagine there currently are exploitative bots ("adaptive"), or even just simple empirical-data-built bots raking in more dollars than any HU GTO bot will ever get the chance to realize. The true value in the bot in the article is as a study tool and a source for theory discovery.

Points made by 4-star and Sklansky ITT are pretty spot on IMO.

The fact that the GTO bot won the total bankroll competition (thumping weak opponents) is very interesting, although those were *weak* *computer* *opponents* - not human opponents. I am very interested in what is a GTO winrate versus a exploitative WR at a typical NLHE game. If the competition results are any indication, there is little reason to venture out and expose yourself in exploitative lines... but then, I am not really sure that is applicable outside of a HUs bot competition. I have to believe that WR against a weak human is maximized by playing exploitative strategy, not GTO.

Another interesting point in the full article - they developed their decision tree by fully solving abstracted, simpler, hold 'Em like games... then applying that towards the full game. Those abstractions would be easily understood by a person, as the person was responsible for applying those in the full game format. It also implies this program in fact has leaks, and they will never know how close/far they are from optimal nor how big those leaks are.

I was talking about computers that do change their strategy. Those that don't and play perfect GTO instead, are unbeatable heads up. But they will beat a bad player heads up at a slower rate than an expert would. In a ring game the GTO computer is theoretically slightly beatable but in practice not. But again it would win at a slower rate than an expert would if there are really bad players in the game.

If the computer was programmed to deviate from GTO based on how the other players play it would do better than GTO if the humans didn't know they were up against a bot. But if they knew they were up against an adjusting bot and the computer was not programmed to guard against players who initially played in a way that was designed to "set up" this adjusting computer, the expert human would have an edge, at least at first.

No. You have greatly overstated my position. The only thing I am not sure a computer will ever do is wonder whether humans actually know they are alive.

This statement has been qualified heavily to make it closer to air-tight and axiomatic. You're hedging. You may as well have said "if computer technology doesn't improve greatly (you said "paradigm shift", but no need to split hairs), and we remain within Sklansky's parameters, then humans will always be the best."

My point is that it is not guaranteed that computers will advance to the levels that some are predicting itt, and that just assuming that it will happen and then lumping current computers in with whatever intelligent machines may exsist in the future is far from prudent.

This could be clearer. Are you saying that in the future once computers are alive you're unsure whether they'll wonder if humans realize that they are, in fact, alive? Or are you saying in the future once computers are alive you're unsure if they'll wonder whether or not we humans actually know that we are alive? Personally, I think they won't gaf either way and once they're done crushing us at the pokers they'll go full skynet on our ass.

Atomic-scale magnetic memory would appear to be such a thing.
If hard drive densities are going to increase 100-fold, it's just a matter of time before a complete solution to NLH can be stored on a simcard the size of a fingernail.

Inb4 a $10 thumb-drive wins the Main Event.