Originally posted by Pkrpains
Originally posted by belthazorrrr
gto meaning anyone?Game Theory Optimal.
That is the original name, but Game Theory Optimization is probably more correct
Notifications
Clear all
General Poker Discussion
71
Posts
24
Users
0
Reactions
12.9 K
Views
The thought came:
few decades ago when computers were not popular or there were no software accountants were doing lot of work by hand. And I believe some could be thinking - oh, those computers will do lot of work and there will be no money in accounting.
And in lot of areas computers, machines are replacing humans. So maybe we should accept that with poker as well.
But of course its not fun to accept that.
Originally posted by SPeedFANat1c
The thought came:few decades ago when computers were not popular or there were no software accountants were doing lot of work by hand. And I believe some could be thinking - oh, those computers will do lot of work and there will be no money in accounting.
And in lot of areas computers, machines are replacing humans. So maybe we should accept that with poker as well.
But of course its not fun to accept that.
I see your point but at the same time I think you can't really compare poker to accounting, since accounting is something that needs to get done. It's not like if everyone stopped playing poker everyone would be like "omg who is gonna play these cards!?"
There is a huge misconception that computers are replacing people in their jobs. Similar misconception that big stores like Walmart are bad for the small communities.
With these things coming in there is a short term problem but then people find better jobs and overall systems become more efficient. A software might do accounting but you still need a human to take decisions the important account decisions that a computer wont be smart enough to take.
Decision making is something that computers cant do at the same level as humans where emotions are involved which makes humans superior, their emotional intelligence.
There is a huge emotional concept to poker just like many other things where humans will always be needed for decision making until humans can create superior computers with emotional intelligence more advanced than humans.
Basically what I'm trying to say is, a computer cant replace a human to run a company, make economic policies or similar things that need the so called complex decision making.
Something that makes humans perform so well at poker are positive emotions that they have like motivation. Stuff that Tendler talks about in his book. We get better at poker cause of our emotion that drives us to improve. We evolved faster than other animals cause we had the emotion that wanted us to improve and get better, we eat cooked food cause we wanted to eat good food, something better than raw food.
A computer can't evolve on its own cause it lacks the emotion that is supposed to drive it to evolve, it doesn't feel the need basically and we all know that poker over the last few years in itself has evolved cause people have had an emotion of wanting to make money doing this and get a better lifestyle and we have been willing to go to extents to study the game and understand our opponents cause of that emotion.
Don't someone dare tldr me.
A software might do accounting but you still need a human to take decisions the important account decisions that a computer wont be smart enough to take.
I think they dont invest enough in this. I really sometimes wonder why people are entering data by hand to that acounting software. Why isn't everything done electronically.
People go to shops and they get incoices in paper where acountants has to enter to the database. While shops could easily use electronic invoices and give that data to the company who buys things. So less and less could be done by hand.
The problem is probably the money required to create everything. Otherwise I don't know. Also there are some difficult rules which can be applied depending on how some interprets them. But thats also a leak in the system I think. Ones say the rules are broken, the one who 'broke' say it is not broken, etc. Computer would not understand that way. Rule either applies or it does not.
But each country would need those rules to be entered to software, because they are different. By rules I meant law.
Decision making is something that computers cant do at the same level as humans where emotions are involved which makes humans superior, their emotional intelligence.
emotions at least in poker does not add profit, they only decrease like tilt.
Yes, emotion makes us improving, computeres will not improve by themselves but those who create the soft will improve the software and computer will just do the task.
Disclaimer: What are you going to read is from a standpoint of a programmer and a guy who is currently studying computer science, so I might be a little biased as I love computers but still I think I have a point in what I'm saying
Guys, machine or a "bot" have no feeling, can't feel tilt, doesn't get bored, it's like ultimate idea for a money making at poker.. 
Yeah, we do have emotions, and after seeing similar situation occurs a lot of times we can get a gut feeling about it, and stuff like that..
BUT seriously, bot can be made to do whatever you're doing now and plus to do it even faster and more accurately..
Just think about it, when you play poker, be it any stake, even up to nosebleeds, you have a though process.. You LEARNED it.. You don't play by your EMOTIONS (which is what humans have that computers lack so badly).. Someone taught you that.. You see things, you use deduction, logic (which also can be programmed), you make educated guesses, you see numbers, deduct things, build ranges in the back of your hand, think about your hand vs range, think about pot odds, outs, everything..
I'm pretty sure, almost 100% sure if someone really puts in a LOTS of work into building a really good bot, it can really be devastating..
I really can't think of any situation a real player can perform better than a bot.. Give me any seriously..
Ie, someone folds 2 times in a row to a 3bet and you make an educated guess that he folds alot to a 3bets and start to 3bet him lighter until he re-adjust? Piece of cake to implement that to a bot..
======Example of a pretty complicated and accurate system of probability distribution of an opponent's range into a poker bot======
Take this for example..
HERO: K :club:Q 
Villain: AQ+, TT
Board: Q :diamond:T :heart:4 :spade:6 :club:4
Villain bet's the pot.. HERO?
*Just as easily as we could use deductive process to trim villain's range on the turn based on his pf play and betting the pot on the turn, we can program our bot to do so as well..
In order for a bot to make a pretty complicated but accurate play in this spot, he needs to do the following:
-Take the pot odds into an account: bot needs to be good >33% of the time to make a call
-Bot losses to AQ and TT but beats AK.
BOT then starts analyzing the no. of combinations of AQ and TT..
It simply figures out that it has an Queen and there is another Queen on the board as well..
There are 2 Queens and 4 Aces(s) to make AQ..
Combos of AQ: 8
There is one T on the board, that makes 3 combos of TT..
Combos of TT: 3
Sum of combos that beats the bot: 11
Now bots simply looks at the hands it beats.. How many combos of AK are there? Bot has King, so there are 3 Kings(s) left and 4 Aces(s) left..
Combos of AK: 12
Bot figures there are 11 combinations that beats the bot and 12 combinations that bot beats.. That's a total of 23 combinations.. Bot can simply divide the combinations of AK by the total num of combinations to see how often will it win if it calls..
12/23 = 0.52
Bot wins 52% of the time and so make as profitable call, as he only needs to win 33% to break even..
Conclusion:
Example looks tough, right? Imagine doing it at the tables while grinding.. I'm sure some guys that plays really high stakes go through such a routine, but I guess the majority of poker players that are playing NL2-NL50 have only seen this stuff in books and they rely on making notes/reads and using hand-reading, etc..
For human, this is super hard to do in a 10-30s period of time.. For a computer? It can do this in a nanosecond and then chill for a bit.. ♦
However, I doubt bots like these fill really hit the market, I think they're priceless and who ever builds this monster would never sell it, as it's really a money making machine.. So I don't really see how bots will ruin the poker.. Simpler bots are easy to beat, and you should be happy if you suspect a player at your table is a bot since they rarely adjust to you and you can hammer them and exploit the sh*t out of it.. ♦
Just my 2 cents..
-eXtremeACE
Well I play microstakes, never played a stake higher than NL 50 and the hand reading process you described is like inherently built in my system and every other player's that I interact with.
You didn't mention the action that happened all through the hand as in who had initiative etc cause before reaching the river I might throw the bot off the hand. I might not even give him a reason to put me on that range based on the actions I took.
Now if I was against a bot here's what I would do. Overbet the shit out of the bot everytime in such a spot, not give him the odds to call. I can do that because I have the emotional intelligence to understand how the bot works and thinks and what I can do to mislead him. His hand reading skill goes out the window and BTW this isn't just something that happens against bots, this happens against some very good players too.
Last Dec I remember a headsup match I think between IFS and KTU. IFS knew how KTU played and knew KTU would not call oversized bets when not getting the odds to call. IFS overbet the shit out of KTU and took down a ton of pots. Not saying KTU plays like a bot but in that match KTU seemed not to deviate from his plan. KTU did not seem to adjust and IFS seemed to exploit it.
A bot will never adjust it doesn't have the sense to. When you know how the opponent plays, no matter how tight or loose he is, he is always exploitable.
Against a TAG if you really wanna play him you need to open up your range. 3bet/4bet wider, put more pressure post flop, take certain lines he might not be able to understand or exploit. That makes my range wider and how does a bot adjust to that? if he looks at my stats to make a decision then I can exploit him by making my stats look a certain way. I could c/r weird sizes just to make my c/r stat look different. A human here might be able to adjust, he might say, you know what this guy is trying to exploit me by widening his range a lot and playing aggressive I can either nit it up, start playing lose myeslf, start bluff catching, start c/r to put pressure on him, OMG so many options how will your bot make decisions here?
A programmer can program a bot to a certain level but can't do much in real time, I find weakness in real time and exploit, the programmer will only know what adjustments to make after the game if he is good enough to even realize and by that time the damage is already done, next time I see the bot I'll make a few plays to see if he adjusted and make my plan accordingly to thrash him again.
In a nut shell what I said is very similar to what so many good players say all the time, the person to adjust first wins. A bot can never make that adjustment on his own which gives me a huge edge.
If you come back with the argument that a programmer can update the bot in real time then well that bot might be exploitable by other opponents. A programmer will have to make a bot that is more or less generic in its play, what makes us better is that we play totally different against different players which is again cause of our emotional intelligence, we play against humans and we figure out what the other human could be thinking or how he thinks based on his stats, we look at what happened in the previous hand, did he get sucked out and does that affect him. Heck we even look at what country he's from what avatar he uses.
Like I said earlier, I won't be scared of bots to me they are like standard ABC players that are easy to exploit.
PS: I remember sweating with speed one day and asked him to raise a cbet based purely on the fact that the opponent was a PS member and like most ABC players he would think that we had to be super strong to raise there. How do you program a bot to react to that when I don't give him the odds. The next adjustment you make, I'll be waiting ;).
Regarding the river overbet, river decisions are easiest to solve from GTO perspective and bot would absolutely know what to do vs river overbet, good bot may even change his calling frequency if for example in same scenario instead of 35.75$ you'd bet 35.55$, he'd know exactly how much more he'd have to call and could randomize it perfectly, what no human would be capable of doing.
And that's just in most easiest scenario facing river bet w no further actions possible except call or fold. The more complex the situation becomes the more of an edge well made bot should have.
And why would bot never adjust? You can absolutely program it to play good exploitative poker same way you do and it would be superior in that too.
Like rly its not even a discussion whether or not a good bot would win.
And the most obvious thing.. How many of you are capable of beating NL400+? I can't, bots can. Saying you would beat a bot just because he's a bot is pretty much like saying you are better than the regular line up at NL400+.
Originally posted by SPeedFANat1c
A software might do accounting but you still need a human to take decisions the important account decisions that a computer wont be smart enough to take.
I think they dont invest enough in this. I really sometimes wonder why people are entering data by hand to that acounting software. Why isn't everything done electronically.
As an IT student I confirm that it is done now electronically? ICT has been much more evolved from what you are speaking was used to be done 20 years ago.
Originally posted by NightFrostaSS
And the most obvious thing.. How many of you are capable of beating NL400+? I can't, bots can. Saying you would beat a bot just because he's a bot is pretty much like saying you are better than the regular line up at NL400+.
I haven't read about this ever, can you post a thread or source for this?
Originally posted by maheepsangari
Originally posted by NightFrostaSS
And the most obvious thing.. How many of you are capable of beating NL400+? I can't, bots can. Saying you would beat a bot just because he's a bot is pretty much like saying you are better than the regular line up at NL400+.I haven't read about this ever, can you post a thread or source for this?
One thread posted by lnternet himself. http://forumserver.twoplustwo.com/56/medium-stakes-pl-nl/alarming-news-about-botting-ongame-1112918/
Originally posted by eXtremeACE
Guys, machine or a "bot" have no feeling, can't feel tilt, doesn't get bored, it's like ultimate idea for a money making at poker..
Yeah, we do have emotions, and after seeing similar situation occurs a lot of times we can get a gut feeling about it, and stuff like that..
BUT seriously, bot can be made to do whatever you're doing now and plus to do it even faster and more accurately..Just think about it, when you play poker, be it any stake, even up to nosebleeds, you have a though process.. You LEARNED it.. You don't play by your EMOTIONS (which is what humans have that computers lack so badly).. Someone taught you that.. You see things, you use deduction, logic (which also can be programmed), you make educated guesses, you see numbers, deduct things, build ranges in the back of your hand, think about your hand vs range, think about pot odds, outs, everything..
I'm pretty sure, almost 100% sure if someone really puts in a LOTS of work into building a really good bot, it can really be devastating..
I really can't think of any situation a real player can perform better than a bot.. Give me any seriously..Ie, someone folds 2 times in a row to a 3bet and you make an educated guess that he folds alot to a 3bets and start to 3bet him lighter until he re-adjust? Piece of cake to implement that to a bot..
======Example of a pretty complicated and accurate system of probability distribution of an opponent's range into a poker bot======
Take this for example..
HERO: K :club:Q
Villain: AQ+, TT
Board: Q :diamond:T :heart:4 :spade:6 :club:4
Villain bet's the pot.. HERO?*Just as easily as we could use deductive process to trim villain's range on the turn based on his pf play and betting the pot on the turn, we can program our bot to do so as well..
In order for a bot to make a pretty complicated but accurate play in this spot, he needs to do the following:
-Take the pot odds into an account: bot needs to be good >33% of the time to make a call
-Bot losses to AQ and TT but beats AK.BOT then starts analyzing the no. of combinations of AQ and TT..
It simply figures out that it has an Queen and there is another Queen on the board as well..
There are 2 Queens and 4 Aces(s) to make AQ..Combos of AQ: 8
There is one T on the board, that makes 3 combos of TT..
Combos of TT: 3Sum of combos that beats the bot: 11
Now bots simply looks at the hands it beats.. How many combos of AK are there? Bot has King, so there are 3 Kings(s) left and 4 Aces(s) left..
Combos of AK: 12
Bot figures there are 11 combinations that beats the bot and 12 combinations that bot beats.. That's a total of 23 combinations.. Bot can simply divide the combinations of AK by the total num of combinations to see how often will it win if it calls..
12/23 = 0.52
Bot wins 52% of the time and so make as profitable call, as he only needs to win 33% to break even..
Conclusion:
Example looks tough, right? Imagine doing it at the tables while grinding.. I'm sure some guys that plays really high stakes go through such a routine, but I guess the majority of poker players that are playing NL2-NL50 have only seen this stuff in books and they rely on making notes/reads and using hand-reading, etc..For human, this is super hard to do in a 10-30s period of time.. For a computer? It can do this in a nanosecond and then chill for a bit.. ♦
However, I doubt bots like these fill really hit the market, I think they're priceless and who ever builds this monster would never sell it, as it's really a money making machine.. So I don't really see how bots will ruin the poker.. Simpler bots are easy to beat, and you should be happy if you suspect a player at your table is a bot since they rarely adjust to you and you can hammer them and exploit the sh*t out of it.. ♦
Just my 2 cents..
-eXtremeACE
What you presented is a final decision more or less. You forget every hand has equity and you need EQUITY not to have the BETTER HAND given the pot odds. Now your bot needs to run equity sims which also take a little time. Apart from that, it has to know to construct it's own range which is the real key, not playing one situation and one hand.
It's a lot more complex than you make it seems. Given the format of No Limit, the number of possibilities has hundreds of zeros of which a bot has to correctly analyse. I will link this thread to someone that can explain it a bit better with actual numbers.
AMaA: The Good Life
:diamond:Follow me on Twitter @emanuelc16
The Education of a Modern Poker Player by TwiceT, jbpatzer & EmanuelC16
Originally posted by EmanuelC16
What you presented is a final decision more or less. You forget every hand has equity and you need EQUITY not to have the BETTER HAND given the pot odds. Now your bot needs to run equity sims which also take a little time. Apart from that, it has to know to construct it's own range which is the real key, not playing one situation and one hand.
Yeah, totally agree with you. I just wanted to give an example, to show atleast how one aspect of the whole NL poker can be programmed..
EQUITY and the BOT
Alright, so there are 2 basic ways to make money at NL and they are to either make a good hand a go to showdown and win, or to win before the showdown with whatever hand, hence the showdown equity and the steal equity.
Bot would do the same thing, it would assign a value to its hand in terms of showdown and steal equities.. It would then weight all the factors in to decide what's the propel play given the situation..
Example: Bot has AA in MP...
Obviously, AA has a great showdown equity, thus bot cares little to his position and previous action and action to comes as well.. It doesn't care whether players left to act behind him are aggressive/passive or tight/loose... So its line is bet/bet/raise blah..
Example: Bot has 72o in MP...
Obviously, 72o has poor showdown equity thus bot then looks at its position, players to act after him, previous action etc and calculates whether it's a profitable steal spot and what's the steal equity of given hand in given situation.. It realizes the hand has poor showdown equity and poor steal equity in this particular situations and it folds..
Example: Bot has K7s in BTN..
K7s has a not that great of a showdown equity, so bot then looks previous action, players left to act behind him etc.. It realizes it is in one of the profitable spot to steal from "CO, BTN, SB", so it then further analysis the situation and now it checks the tendencies of a players in the big blind.. It takes into an account how often they fold to steal, call the steal or 3bet the steal, and from that he build their ranges and sees how K7s comapres to their calling/3betting ranges as well as how often do they folds to see whether or not it has a profitable spot to steal..
I mean, yes I'm giving a simple examples, but I just want to show you what I'm talking about.. I'm not talking about a bot that checks if the hand is AK, and then bet.. ie if (hand == hands.AK) player.bet(betSizes.OR_size);
I'm talking about a bot that thinks like a player, that evaluates situations, not just blindly follow a guideline and cares little to not at all about table dynamics..
It's a lot more complex than you make it seems.
Certainly, but it's doable given a lot of time and dedication I think..
Now your bot needs to run equity sims which also take a little time.
It needs to run the equity sims the whole time, but it can be simplified really.. I mean, even if it has to run a whole simulation it would be doable, even tho operation of comparing is really costly in programming world..
Let's do a little test, some of you great players, can you think of a spot in where you had to think really hard and you doubt bot would be able to do it as well? Just take into an account "I had history with the villain and knew his tendencies to do X stuff after Y" is not an excuse, because bot would have access to the same hands in the past you've played with that villain, and could compare all the times villain did X sized bet on a particular street to see how often it yielded a bluff or a value hand and such..
Now if I was against a bot here's what I would do. Overbet the shit out of the bot everytime in such a spot, not give him the odds to call. I can do that because I have the emotional intelligence to understand how the bot works and thinks and what I can do to mislead him. His hand reading skill goes out the window and BTW this isn't just something that happens against bots, this happens against some very good players too.
Yeah, when you start to overbet the shit out of the bot, bot would simply realize that you can't be having that tight range when you do it, and would start to increase its calling or even shoving range as you start to overbet him more and more..
Again, how does his hand reading skills go out of window? I'm not talking about those primitive bots, and I guess no one in this discussion is, because we're talking about a bots that could ruin the NLHE, and those bots can only really help you print even more monies..
It uses deduction just like you do and put you on a range of hands, trimming it more and more as you do more actions in front of the bot, and the more hands it has on your play..
There's no doubt a WELL PROGRAMMED bot would destroy regs at lower limits when it starts to have sufficient hand history, but at higher limits, there's no way to make the bot adapt to a human player that adapts to the bot, because a good human player would always adapt to the bot's adaptation until the bot runs out of code, and then it starts losing money.
Of course, in such a situation the bot could be programmed to default to abc style balanced play.
Point is, if online poker ever became infested with WELL PROGRAMMED bots, it would make online poker 50x harder than it currently is. Especially when those bots could run 24/7 24 tables at a time and gather massive hand histories against every opponent.
to anyone denying the effect and playing potential of bots, you should really just read the (pretty old!) 2+2 thread posted by lnternet.
this is a quote from there:
Originally posted by skier_5
- there's a guy who developed a neural network based bot and created a web based api that people can query with a situation and get an action back. He was soliciting partners and as of January had 22-27 of them. His goal was to cover all the tables on all (perhaps not stars?) networks up to 5/10 every hour of the day. The partners develop their own front ends and set up their own accounts while he takes a cut (read 50% after a certain threshold). A guy this sophisticated is going to be able to improve his bots faster and faster with millions of dollars and the tens of thousands of hand histories he's accumulating every day.- it doesn't sound like it is necessary, because from what I gather running a virtual machine is undetectable, but some people seem to be intercepting video monitor feeds to screen scrape and using hardware to control the mouse.
and that dates back to 2011! who knows what they are doing right now in 2013.
this is a super huge $$$$$$$$ business, no wonder there are people who put in a lot of effort into it.
this is fucking serious for the future of online pokers, they are definitely already taking out millions every year.
i prefer to be pessimistic and think that online pokers don't have that long of a future life span, but i'll enjoy it while it lasts. 
Clearly its possible for bots to be crushers. A good amount of poker skill is mathematical, and when it comes to math computers > humans obviously.
It's really on the poker rooms to stop the bots. The one I saw recently on pokerstars is good, where the verification word comes up and you have to type it in, its simple but the bot will stand no chance to see it.
But the problem with that is that if the bots can beat midstakes+, it is still worthwhile for someone to sit in front of the computer and watch the bots crush, maybe surf the web in the meantime...so its not a great defense, and its hard to understand how to combat a running bot with a human at the screen to help it get through the tests.