Quote:
Originally Posted by swoni
don't believe for a second you didn't mine hands, you said you coded to incorporate hud stat adjustments.
Also if you know any other bots, can you out them.
Also post graph.
WPN hasn't allowed mining for some time. You can't just open a table and start saving hand histories without sitting. I suppose I could have configured the bot to just screen scrape for data or something, but when I started, my initial priority was to see if I can get the bot to work well and play well. By the time the bot was sophisticated enough, I already had a good pool of hand histories and didn't really want to waste time designing a system solely for mining.
I'm not posting any personally identifiable info. My graph has some notable, and I'm sure unique, downswings and upswings. You have to realize that multiple players probably have 6-figure hand histories on me and it wouldn't be all too difficult to piece together an identity through some mutual cooperation. Everything that I've been willing post in terms of details and screenshots, I have already posted. Most of it has been deleted or edited out by the mods; I am not risking any chance of outing myself, sorry.
Quote:
Originally Posted by DaFooFighter
OP, could you tell us more about bot collusion. I do think that is the bigger evil by far. How much more work is needed? What is the process of card sharing in real time between the bots? Realistically at the moment what is the biggest stake a bot could be profitable at for both PLO and NL 6max? In your opinion, how long would it take for bots/ai to be able to beat the highest stakes?
I am actually not very familiar with bot collusion networks. From time to time, I would see posts or irc messages concerning some foreign network or another, but most of the time those messages were about troubleshooting and technical help, not recruitment. My understanding is that such networks generally have their own private, and heavily guarded channels. Bots will create logs in realtime: for example if the bot has AA preflop, there would be log output line indicating that. The next step would be sharing that info via a private network (maybe irc bots? maybe through a server? maybe a more direct approach?) and having an additional mechanism for each bot to be able to read the info as well as send the info on a hand by hand basis. There wouldn't be any need to share post-flop information since the bot could just read it off the table itself and presumably some shared code would tip it off on what the other bots in the game will do based on their hole cards, board, and current pot/player state. The far greater challenge would be masking the collusion so as to not make it obvious, and that would take a lot of time. The barrier to creating a colluding network is trust. You would have to find people who are extremely trustworthy, but also similarly competent and careful. That isn't easy. A workaround might to be just multiaccount across different skins of the same network or even the same site using personal info of people you know. You would have to configure some VMs and I guess set up a bunch of VPNs on each of those VMs, or have an entirely dedicated hardware setup. Again, doable but a significant financial and time commitment.
I would think that if a good poker player and coder spent a year or so, he could probably create a bot to beat the highest stakes of NL, but probably not as successful shorthanded or HU. Predictability and consistency are what a bot needs and in shorthanded games at the highest stakes where players are making rapid adjustments on a orbit by orbit basis isn't the most hospitable environment for a bot. Most good botters,like me, prefer to stick around limits where there are hundreds of players and dozens of tables to chose from. This allows a bit of anonymity and also, a lot of opponents are going to be multitabling and seldom deviating from long-term trends. It's much easier to grind out steady profit here than in a high-takes game where there are maybe 2 running tables, 10 regs, and insane competition. Grinding 100-50nl requires only 1500-2k of "working capital" in an account at a time, reloading/withdrawing as necessary. Playing highers stakes would probably require 5-10k in the account each session and it would suck to lose all of that at once due to confiscation or something.
Quote:
Originally Posted by miliboo
OP, you seem intelligent, well written etc.
How many other botters do you know/have contact details/screenames etc?
I have contact with a few botters. I was more active in different communities when I first started out since I had a lot more questions and often needed troubleshooting help. Now I mostly offer advice or help to noobs and don't really chat with/e-mail established botters unless something major happened on one of the sites. Good, careful botters will not out their screen names or playing profile so I can't really link a lot of my contacts to specific identifies across different poker sites. From time to time, I will see a couple of players' stats on PT that I suspect are running some default profiles or ****ty purchased profiles, but they don't usually last for more than 10-20k hands.
Quote:
Originally Posted by Dick Tracy
OP, don't you empathize with those your bots won money from?
Not wagging my finger at you, I'm just asking from a journalistic approach. I'd really like to know.
Not particularly, I spent ton of time and effort--probably more than most of those players--on analyzing gameplay, hand histories, etc. Nothing I made was "free money". I risked the same amount per hand as everyone else and I put in a lot of hours to develop and optimize the bot and its playing capacity. Yes, my efforts were directed differently than a lot of other poker players', but nothing about creating a *good* bot is easier than working to become a good poker player.