Poker-AI.org

I am working on a C# bot and now need to setup a testing environment. I have searched the net but most threads I find are several years old and things seem to change fast in botting. I therefore suspect much of what I have found is totally outdated.

Currently, my planning looks like this:

a) A first step is to use PokerAcademy 2.5 for testing. (my thinking is that if I can't beat the bots in PokerAcademy, I should stay away from the poker sites)

b) Next step might be opentestbed: http://code.google.com/p/opentestbed

The problem is that i find very little technical information on either of these.

How can I get a bot written in C# to play in PA and in opentestbed?

Are there any strong bots to play against in opentestbed?

Are there any other alternatives for testing?

How do you test your bots?

I have always tested my bot in play money games. I have found its better to do that as you see a wide range of player types and actions. I let it play for hours and take notes on mistakes and situations that I had not programmed for.

micros...

Hi,
If someone is planning to write a poker server/client in C# as a testbed I'm willing to collaborate. I will need one in the future as my bot uses decision trees so I will need to test specific hand combinations, lines and boards over large numbers so "live" testing is not very useful in my case.

Hmm, I wonder how similar that is to playing with real money? Don't you find that people play much loser with play money than they do with real hard-earned money?



Yes winnie, I totally agree, I think! But what does it mean? micros... ????

1c/2c or 2c/4c tables at a room you want to bot

Yes it is very loose but that is good for bot programming. This is good as you will have all the bases covered even for unrealistic scenarios. I have found HU SNG and Cash to be fairly normal but 6max and 9max can get a bit crazy.

I have been doing this for ten years and I think it is the best way to do it although it is slow. Usually the problems show up quickly and it does not take as long as you would think however.

OK guys! Thanks for your input so far, and please keep it coming!

My latest, firmer and slightly revised, planning for testing now looks like this:

a) A first step for off-line testing will be to use PokerAcademy 2.5. I will use combot as a base for a C# to meerkat interface (see here: http://poker-ai.org/archive/www.pokerai.org/pf3/viewtopicfd37.html?f=3&t=3393)

b) Haven't decided yet if I will use opentestbed as a next step for off-line testing: http://code.google.com/p/opentestbed (is this project dead?)

c) First step for on-line testing will be play money on some site where stealth is not so important.

d) Then testing on same lo stealth site with real money micro stakes, and then progressively move to larger stakes when I am confident my bot is reliable.

e) When everything works and I have a hight stealth setup, I will start playing to win on some easy sites and then on difficult sites like PS.


Your comments please?

You can use Open Holdem on some sites I think

Not sure what you mean by that spears.

AFAIK, Openholdem is alive and kicking and can be use on most (or all?) sites. I was actually planning on using Openholdem as a screen scraper engine for my bot via a user DLL plugged in on OH. Unless someone has a better idea.

OH is ridiculously easy to detect, so it can be applied only on sites which don't do any bot checks.

Should be no problem for me, I hope. I plan on running the poker client in one VM, and OH and my bot in another.
That way, the only botting sw running in the same Vm as the poker client will be Bring, which has a howto on stealth.
Do you think this is stealthy enough for PS?

I haven't read that Bring howto, but, unless it involves installing some kind of rootkit, it should be deteactable. Also, there were reports of PS screwing with screen scrapers by randomly changing pixels colors/locations. Plus, you'd need to solve the PS's captchas.

You have some "high stakes" at playmoney tables. You'll find people playing more tight because you need to be a little bit serious to win enough chips to reach those limits.
There are also rooms where you can win real money by playing play money so the level might be better.

it is best of price/efforts ratio

2nd place — nodeposit bonuses
3rd place — instant bonuses

According to the howto, Bring should run under another user so the poker client (PC) user cannot access it's process information, also Bring is "hidden" in a folder owned by that other user and which cannot be accessed by the PC user.

Obviously, this is not perfect stealth. The PC has to run with admin privileges, so it can take command of the folder where Bring hides and have a look inside, thereby discovering Bring and raising the bot flag. One solution might be to hide the Bring folder with a rootkit, but I guess also rootkits can be detected. I know very little about rootkits, so please correct me if I am wrong.


Yes, that is a big problem. I haven't started studying this, so I don't know if it is possible to detect captchas with screen scraping. I think that if I hook on to the PC process I could detect a popup window opening. But that would require a process running in the same environment as the PC, which I think the PC could detect. How do I solve this?

I'm not a rootkit expert, but as far as I know, if it's done right the poker client has little chance to detect it. Having said that, the more popular route for screen scraping nowadays is just to have the client in VM and control the VM from the host (VM engines have VNC interfaces to do that). This way you're 100% legit, apart from the flag raised by running in the VM.



Detecting is easy, I'd be more concerned about solving them in the limited time available.

Which room are you talking about?

i'm not talking about specific room