Topic: What could possibly go wrong? Please, tell me :) (Read 1008 times)

What? Please quote me where i claim i am in profit

In fact, i posted an update stating that i BUSTED. And i also stated my thoughts on it.

Damn, its tiring entertaining sig spammers I am considering locking the thread.

edit: locked, since i shelved the positive EV strat which this was all about

enjoy ur signature payouts

First of all, why would you make a bot of a poker game? 
That is a game of thinking and strategy not just a dice game.
You will only lose more doing this.
Just play the game and enjoy it. That wont work and more likely your opponents will see it.

This kind of strategies will work when we are implementing the first time, and while testing this approach, it will work out. But not always. Once or twice we will win our bet. It maybe the strategy will help us to win our bet or our luck anyone will help us. But I think it will not work for long term. I am not underestimating his work I am just telling the fact in the game. I wish you all the best and get a good result in this project.

It should do work but the main question is for how long? I do see the strategy is pretty interesting and simple to understand but chances of being busted is always there that is why I don't trust any codes which do someone claims that it works. Manual poker would be always be the best.

I think the code will work though I am not a professional coder but I have dabbled a bit with it on my studies on college and I think the code will work with its simulations and its reading of the cards not sure tho whether it might be profitable in the long run.

I would love to let you try the program ! However, it would require a lot of effort from me, and too much (i assume) Python knowledge from you as well.

As tspacepilot accurately states though, i shared the language i used and the essential modules.

If you read the entire thread thoroughly and you have programming skills, you should easily be able to replicate, and perhaps even improve, what I am doing. It is quite simple actually, well, the idea is at least.

To those who might be interested:

After a lot of up and downs, with a peak balance of 0.0425 BTC, my poker balance is now 0.

I am not too worried that it's an indication of failure though.. After all, what happens if you open a casino with insufficient bankroll?

I am tempted to see how a full BTC fares, but I dont have that kind of trust in my own code..

For now, I am using parts of my code to assist in "manual" poker.

I don't think anyone is asking that.  Did you read the thread?  Or even the OP?  The OP author already said he's using Python with the Deuces module and selenium for browser automation.

I think the point of this thread is that the OP is discussing the idea of the program.  Not sure what you're contributing here other than boosting your post-count.

Which site do you play on? Most of the sites I know do not let you leave and the rejoin it with lower stack than what you left with. There is(and should be) always a cool-down period about 5 mins during which, if you plan to re-enter, you must rejoin with the stack you left with or higher.

Well I play poker in computer sometimes and poker requires some skill and luck in able to draw your best game out there. Like everyone's asking here what programming language did you use? We understand that if what you're thinking is " its not yet complete or its not that good ", then let us try the program. Its hard to give feedback if we don't experienced the program, and if you are worried about "what if someone stole your program and idea" then give us more detailed details in your project. Good luck.

If you lost your buy in you don't need to leave the table and can rebuy directly. If you leave the table and there are some players on the waiting list you can not re-enter directly.

If you won an all-in, you can not leave the table an re-enter with same buy in. You have to re-enter with the full balance you had or wait 30 or 60 minutes.

I am not assuming the opponents has worse hands, what led you to think that?

Sure, most people will fold.

Sure, if they dont fold they probably hold qq+/ak/at etc...

If every poker player in the world never went all-in preflop with less than AA in their hand, then sure, I would lose since my bot would calculate the win chances based on any possible hand so if they sit with the strongest one possible I'd be in a bad position.

But people will actually risk their balance on a suited hand, in the hopes of hitting a flush that will win against even my worst-case scenario hand of AA.

Or if they feel like they have a sneaky straight coming, i.e with a 7j hand.

That link is very nice source of information though ;-) Just ran a couple of tests, i am at most ~1% off with my own calcuations, but i could have saved a lot of code by just writing that table into a dictionary.

I know, I was simplifying it a bit for the sake of the argument.

The biggest problem tobi's strategy has is assuming people will call his pre-flop all-in with worse hands. The fact is, people will not generally call a pre-flop all in, unless they hold something like QQ+/AK.

Bascially, you are only getting called by better hands.

Second, calculating the win chance of your hands is not even needed. There are spreadsheets for that based on the outs that you will learn by heart after playing a bit.

http://www.natesholdem.com/pre-flop-odds.php Here you can find the pre-flop odds for all hands against a number of opponents. 

Then prove it.  Show us your graphs and let everyone in the community verify your results.  There are only a very few botters who are good enough to beat the rake.  If it were really that easy then everyone would be doing it.

I don't think your application of "gambling theories" is exactly relevant here.  In contrast to a game like slots, which is pure gambling (and in which case you're right), poker is actually a game between individuals.  There's an element of randomness or chance, but it's not the case that every poker player is guaranteed to approach a loss as their playing goes to infinity (which is the case for slots).

I think the OPs idea is interesting because they try to remove further uncertainty by continually going "all in".  This sorta flattens out the decision tree and makes the modelling simpler.  In any case, you'll observe that in a poker game the house grabs a cut of the pot, and thus guarantees its own win.  But beyond this there's no mathematical certainty for any player at the table that playing more will make them approach a loss.  If you always bet a constant amount and only play on hands that will win 55% of the time, then you ought to see a profit over time.  This, of course, depends on the idea that people will call you with any cards (they won't), and it depends on the idea that you're correctly calculating the a priori odds of wining with that hand.

Anyway, this is why a lot of poker players say that poker isn't gambling.  And I think they're partially correct.  It's definitely not the same animal as dice or slots.

Please test this out for a longer time.
Like a 100 games, if the profit is significant, brand the software, make a website, and cha-ching BTCBTCBTC

Everyone wants easy money, I don't know anybody that doesn't.

Anyway, about your strategy, I would not comment on it's technical value but more on gambling theories. No strategy will work all the time in any gambling game, you will still lose in the end as. From your description of your strategy, you still assume there is a chance to lose (as expected). So there is a possibility that you will end up losing a lot of games in a row and eat all your bankroll up.

As I understand this, you're not doing EV right.  It seems like you're saying that you calculate your chances of winning as if everyone calls, but calculate your payout as if only one person calls, then you adjust the payout according to what actually happens.  I can see why this is seems like a safe miscalculation.  But it's also confusing when you're explaining your approach.

Another thought, and maybe this doesn't matter, but I'll say it anyway.  People can have a short stack and call you, that affects how much you can win.  I t also affects how much you can lose so perhaps it's a wash.


I see, I found that module, I didn't understand before if you were writing your own poker module called Deuces.  I also saw that the "simulation" code is probably also coming from Deuces, the readme talks about simulating hands.    I suppose it's up to me to read the module if I want to know what's being simulated.   My question is sorta like this, let's say we're sampling from the following distribution of values [ 1, 1, 1, 0 ].  You can see right away that you'll have a .75 percent chance of choosing a 1.  Now, we can establish this empirically by writing some code to simulate sampling and then do it 10K times or whatever and look at our results.  However, there's really not a lot of need for the sampling step since we know the distribution to begin with---the sampling step really just tests our ability to write sampling code, we know that if we find that we're approaching anything other than 0.75-0.25 in the long run average, then we're not sampling right.  It seems to me that your situation is similar: you know what cards you have---given that you  already know the distribution of hands you'll have after 5 cards have been drawn.  From there, you also have a joint distribution between your hand and those of N opponents and you can calculate the probability that your hand is better.  As far as I can tell, you don't actually need to do any sampling to learn that, do you?  Don't you just use some multiplication and division to calculate odds?

Anyway, I think that selenium is a nice package for browser automation.  I'm pretty sure you can export a screenshot from selenium if you wanted to try OCR (but I don't know if it'll be fast enough for you.  It should be smart enough if you are able to constraint the size of the screenshot and if you can train your own model.  You only need to recognize a small set of characters.

 

I love that question, it fits so well with your signature ;-)

Yes, it is against ToS/ToU of most poker sites you find.

That is mainly why i havent disclosed the website im playing at.

Although, i assure you that I am looking for fun and experience, rather than malicious intentions and easy money.

@tspacepilot

I calculate the EV with following assumptions:

All players still in the hand when it is my turn are playing.

Only 1 opponent calls.

So basically, the EV is for playing a round i.e against 5 players, while the EV is calculated as if only 1 opponent was playing the hand (since 1 call is assumed)

Therefore, if all 5 enemies decides to go all-in, my winchances are the same (calculations done as if playing against all) but the potential stake is a lot larger than i calculated (my stake*5 + pot, instead of mystake*2 + pot)

About the why of my simulations:

At a table with 1 other opponent, an AA hand is a 100% all-in but with 5-6+ opponents, pairs are harder to achieve a +EV on, while jack/ten and same-suits are easier to get value from.

Since people leave and enter at will, and balances/pots changes non-stop i feel it is necessary to run simulations before every decision.

Regarding the technicalities, I am using the Python module Deuces to handle everything Poker, and the Python module Selenium to handle all interaction with the webpage.

I really wanted to be at another poker site, but tinkering with stand-alone poker clients are way above my league