|
drtofu66
|
05-09-2007, 04:49 PM
Post subject: NL Bots on Full Tilt
|
#1 (permalink)
|
|
Flush
Join Date: Apr 2005
Posts: 595
|
|
There's a thread on 2p2 (with this subject title-- I wasn't sure if direct linking conformed to board rules so you can search for youself) which is pretty interesting. The author has some pretty damning evidence posted and it's worth a read, even if it's a long post.
|
|
|
Play for FREE and practice your game at...
Join the FTR Poker Forum to disable these banners and start posting!
|
|
zook
|
|
4-of-a-Kind
Join Date: Mar 2006
Posts: 3,676
|
|
There's no problem with linking to it:
http://forumserver.twoplustwo.com/sh...t=1&vc=1&nt=18
It's interesting, because I played 30k hands at 200NL FR on Full Tilt. I played a ton of hands against these "bots" and never suspected. I just thought they were bad, unimaginative, taggs 
Really curious to see how this turns out.
|
|
|
|
Sheetah
|
|
Flush
Join Date: Oct 2006
Posts: 278
|
|
This is huge. And it seems that bot owner/programmer joined the conversation pretending he's sceptic (maybe not but that guy is lol).
ZOMG that thread is monstrous!
|
|
|
|
xyu
|
|
Join Date: Oct 2006
Posts: 49
|
|
that thread really is a monster!
its pretty scary stuff. i'm amazed there is anyone in this universe that can possibly believe this isn't a bot from the post flop stats. 4 people in a room playing 15 hours a day, that have "learned" to play exactly the same way in thousands of different situations, consistently and without even a slight variation from player to player on any given stat. crazy.
nobody in here seems to care much though, do we just not see bots as a threat?
|
|
|
|
mrhappy333
|
|
Full House
Join Date: Jan 2006
Location: Hartford, CT
Posts: 1,078
|
|
a Bot once beat me at holdem, but he was no match against my kick boxing!
|
|
3 3 3 I'm only half evil.
|
|
zook
|
|
4-of-a-Kind
Join Date: Mar 2006
Posts: 3,676
|
|
Quote:
|
Originally Posted by xyu
nobody in here seems to care much though, do we just not see bots as a threat?
|
Obviously I don't think it's good for online poker, but they didn't kill the game, b/c I was a regular at 1/2 FR at FT for a few months while these accounts were active and it was really soft. I guess it would have been softer if these "bots" weren't taking the fishies money, but I'm not convinced they weren't just seriously dedicated grinders.
|
|
|
|
xyu
|
|
Join Date: Oct 2006
Posts: 49
|
|
firstly, i hope i don't sound insulting at all in anything i say here.
do you honestly believe that 4 guys sat in a room, playing for 15 hours at a time (serious, mentally draining sessions) could possibly play identical post flop poker based on a memorised system?
My first thoughts when i read the post, were that it could just be a few guys with a program that prompts them as to their next move, just in the same way i read a blackjack chart that told me exactly what to do when i casino whored. this could easily be done, but realistically, it has all kinds of other issues. They would effectively be data entry clerks, which is a horrible job to say the least (whether the pay would be enough to make them do it idk).
But the fact is, FTP didn't do anything about it, and with stats like this, how can they not believe they are bots? AND, if they can do this and not be stopped from playing, these guys (this guy) might aswell make their prompt program go the next step and get the software to do the clicking too, eradicating the immense boredom that this setup would bring about.
I understand they didnt kill the game, but this looks a pretty crude model from what i can tell. I don't doubt for a minute their are tons of people out there that could do better, and build a much better bot with a better basic strategy, aswell as a more reactive bot. and if this ever became big news, why would casual gamblers ever deposit and play at any site, not just FTP, but any online poker site. I really don't wanna play against bots and good players only.
|
|
|
|
zook
|
|
4-of-a-Kind
Join Date: Mar 2006
Posts: 3,676
|
|
I hear what you're saying and I think it's quite possible they were bots. As someone who's played ~8k hands total with the 3 accounts and seen them chat and occasionally make very non-standard pre-flop plays I know that humans controlled the accounts at least some of the time. As for the post-flop numbers, I actually find it hard to believe that a bot could be programmed to have identical numbers on the turn and river, even over a 100k sample size. (FWIW, in my mined database of 20-28k hands on each of the 3 accounts their post-flop numbers diverge significantly.) A bot that was programmed to bet the turn a certain percentage of the time regardless of flop action or the turn card would be a horrible player. Much worse than a 1ptbb/100 winner. I have hh's where they flat-call turns and rivers with AA b/c the board is scary and some where they get it AI vs. bad players. The problem is, individual hands prove nothing. But unfortunately the stats don't either.
I don't want to get in a big argument with you b/c I mainly agree. The stats make it look like they're bots, Full Tilt handled it poorly and if the story gets out it will be a big blow to their business and all of online poker. They should have returned their money and banned the accounts.
|
|
|
|
biondino
|
|
4-of-a-Kind
Join Date: Apr 2005
Location: Putney, UK; Full Tilt,Mansion; $50 NL and PL; $13 and $16 SNGs at Stars
Posts: 3,170
|
|
Here's a thought. Might the bots be so unsophisticated that they are simply programmed with 1) a simple, tight pre-flop range and 2) a cbet on 100% of flops. Should the hand still be active post-flop, the owner takes over?
If this were the case, the bot would be ultra-simple, with set actions both pre-flop and flop, and presumably a built-in "fold to a flop bet with less than TP" (say) rule for when it has to be reactive rather than proactive.
|
|
|
|
zook
|
|
4-of-a-Kind
Join Date: Mar 2006
Posts: 3,676
|
|
Quote:
|
Originally Posted by biondino
Here's a thought. Might the bots be so unsophisticated that they are simply programmed with 1) a simple, tight pre-flop range and 2) a cbet on 100% of flops. Should the hand still be active post-flop, the owner takes over?
If this were the case, the bot would be ultra-simple, with set actions both pre-flop and flop, and presumably a built-in "fold to a flop bet with less than TP" (say) rule for when it has to be reactive rather than proactive.
|
I think this is quite possible.
|
|
|
|
xyu
|
|
Join Date: Oct 2006
Posts: 49
|
|
it suggests to me that bots aren't going to ever be detected, and even if they are, they will just be let off if the owner says he has a system. i dont think the stats would be this close for any one of these guys over 2 different 100k samples, let alone the same for 4 of them.
but this is what scares me. they were only caught(or flagged) because of the 4 accounts, and even then it wasn't FTP that stepped in. 4 accounts, playing on the same IP, with identical stats? if they don't flag these accounts, do they even have a proactive security department, because they clearly should be spotting this one straight away. clearly, even if they don't have proof, 4 accounts on the same ip with identical stats, playing crazy shifts at exactly the same times is fishy enough to suggest they should be banned.
and if this is a bot, clearly, at some point the guy will actually play poker on the account, he'd be an idiot if he didn't, unless he was experimenting with how long it'd take to get suspended, or to see if he had managed to bypass the ftp checks.
there are a million reasons to suggest this is a bot ( are we really supposed to believe that 3 guys all leave their pc's and the 8 tables they have running to discuss a hand with the 4th player, what happened so far, the many variables, theorise and agree and disagree together AND never ever improve and thus change their identical stats without timing out and misplaying one of the other 24 tables they have open between them?), but clearly, if FTP can't spot this and does nothing about it, then the future has some serious doubts hanging over it.
|
|
|
|
jyms
|
|
Tilting Mod
Join Date: Feb 2006
Posts: 4,837
|
|
Quote:
|
Originally Posted by xyu
that thread really is a monster!
its pretty scary stuff. i'm amazed there is anyone in this universe that can possibly believe this isn't a bot from the post flop stats. 4 people in a room playing 15 hours a day, that have "learned" to play exactly the same way in thousands of different situations, consistently and without even a slight variation from player to player on any given stat. crazy.
nobody in here seems to care much though, do we just not see bots as a threat?
|
Actually we do, this is the mini thread, most people are posting in this one http://www.flopturnriver.com/phpBB2/...707&highlight=
|
|
|
|
FlyingSaucy
|
|
4-of-a-Kind
Join Date: Oct 2004
Location: Watching the kids
Posts: 1,603
|
|
I read the whole first post on 2+2 but none of the replies. a couple of thoughts.
1. I'm asserting that a significant population of bots exist on every major poker site, not just full tilt, and he only found these because they were programmed with a very unsophisticated algorithm.
2. no poker site has incentive to kick bots because the more players are being dealt in, regardless of human or bot, the more money is being made for the site. The only incentive to kick bots is for PR and to not scare off too many humans.
A good programmer/team could make a very sophisticated program using artificial neural networks that could kick the crap out of 95% of players out there. A great programmer and research team could do much much better.
There are no theoretical barriers that separate bots from humans when it comes to a decision making process that has finite variables and finite results. A few years and we'll be looking at the winningest NL holdem player in the world, and it won't have pores to sweat in a tough decision.
|
|
|
|
biondino
|
|
4-of-a-Kind
Join Date: Apr 2005
Location: Putney, UK; Full Tilt,Mansion; $50 NL and PL; $13 and $16 SNGs at Stars
Posts: 3,170
|
|
A bot could take advantage of game theory more reliably than any human player. A bot *could* feasibly develop reads on a regular by making a statistical analysis of his past play. But a bot will always struggle with on-the-fly reads because it's literally unprogammable in situations where the opponents' small sample size makes useful analysis impossible.
|
|
|
|
gingerwizard
|
|
4-of-a-Kind
Join Date: Sep 2006
Posts: 1,815
|
|
Quote:
|
Originally Posted by biondino
A bot could take advantage of game theory more reliably than any human player. A bot *could* feasibly develop reads on a regular by making a statistical analysis of his past play. But a bot will always struggle with on-the-fly reads because it's literally unprogammable in situations where the opponents' small sample size makes useful analysis impossible.
|
Incorrect. Even I could write a program that adapted a reasonable default prior linearly to small data after around 20-30 hands. Small samples are only a problem if you want to always use the central limit theorem. If not there are loads of techniques you can use.
A human can take advantage of a bot because it's actions are deterministic. Knowing you a playing a bot, playing a system you should be able to play optimally against it.
|
|
This is not my signature. I just write this at the bottom of every post.
|
English
Deutsch
Español
Français
Italiano
Nederlands
Pусский
Svenska
06-02-2012, 08:10 PM Merge Network 6.0 Looks to Retain Current Player Base
Linear Mode
