I've recently discovered that I've been folding way to much in the BB vs an SB steal. I think its probably safe to assume I'm folding too much in the BB vs other position opens as well.

I'm curious what others poster believe to be a reasonable defense range vs Btn, CO, HJ & UTG. Assume a reasonably ok/solid regular in a 1-2 structure. I think it would be helpful to list ranges rather than Top X% as I believe they can often differ.

I started this thread in the Stat thread, but I think it would be more useful as its own thread. I'm going to re-post the initial discussion with Unguarded here.

Unguarded:

Myself:

I didn't mean to imply there is one standard range of hands that everyone should be using. I realize defense ranges are dynamic and change based on a number of factors. However, I think with the parameter of a reasonably good reg opponent who doesn't have any major flaws post, its reasonable to list basic defense range. I posed this question more to a get base line to see if a certain range(s) differs drastically from my general approach. If they do, I likely have a fundamental flaw in my BB defense.

Unguarded:

I'm generally folding K7o, Q8o vs an Reasonable UTG opener. I don't have any real quantifiable data to support this, but I just feel like playing these big-little dominated hands is going to get us into trouble postflop.

Same thing with the Big-Little suited hands. I know the suited-ness adds value, but it really enough value to make playing K2s OOP vs and UTG open profitable?

Always err on the side of playing too loose in the BB. I think the ranges I gave are more likely too tight than too loose. Folding too much in the BB is a massive leak. Calling too much is barely going to be noticeable to your winrate.

I use a pattern: if they open A4s, K7s, Q9s, J8s, T8s, 98s, 87s, A7o, KJo, QJ, then I'll play hands with a kicker two steps below: A2s, K5s, Q7s, J6s, T6s, 96s, 85s, A5o, K9o, Q9o, J8o, T8o, maybe 97o.

I'll add suited hands if he's a frequent turn checker.

Against the button lagtag, I'm playing almost anything suited(I'll fold 94s-) and any 2 cards 6+. I'll fold the weakest connectors like 43o-.

Hard to say vs. the earlier positions but in a recent 2+2 article we found that the BB should never fold to a SB open.

It would be a nice read !

It was a good series. We have a thread in microforum discussing it.

Here you go:

http://www.twoplustwo.com/magazine/i...d-vs-blind.php

Phil, one thing I don't understand about bots pre-flop:

I always thought that the programmer is the one who constructs the bot's pre-flop play in most cases. For example, the hands that are placed into the 10 buckets for the bot in the article... the programmer obviously did that, right? For the bots in TIP, the pre-flop play was all over the place. And Sonia's was obviously way different from the GTO bots.

If most of the bot's pre-flop play is designed by a programmer who likely isn't very good at LHE, how can we learn much from the bot's pre-flop play?

On BVB play... I am wondering how much of a difference it makes when there are 4 folds pre-flop vs. when it is just a straight HUHU BVB match such as in your article? In a 6-max game, 4 folds means that 4 people likely had 1 or 2 low cards in their hands. How much does this affect the likeliness of BB having a good hand when we open from the SB?

Also, must give Oink some credit... he was 100% convinced that we should never fold BVB against normal opens over 5 years ago and was pretty loud about it. I have been calling 100% for years and telling everyone else to. Quite a few of us have been sure this was right. To me, it was obviously correct simply because we know you should only be folding 10% or so HUHU... surely position means we can just call everything. It honestly continues to surprise me that almost no 6-max regs defend 100% BVB nowadays.

I have been playing a lot of 3-6 handed against neobot. I have about 7k hands against it. For some reason, it still folds sometimes BVB... I would estimate about 5%. It also opens really wide from the SB... nearly as wide as most people do HUHU... hands like 32s, J4o, Q2o, 54o that most of us just muck. Early on, it was sometimes open limping BVB but it seems to have stopped. Its 3-betting range also seems more mellow and balanced than most humans'. I've emailed the people who run the neobot website asking some questions about how the bot works, but haven't gotten a response for a few days now.

Cool article... just wondering how much we should really trust that bots' pre-flop strategy and why. And thought I would share my experience with neobot so far. I plan to keep sparring with neo, so let me know if you have any questions. Prolly gonna spar with this bot from the article too now, lol.

Oh, and is the bot from the article available to the public? Can't find it if it is.

thx !

I think the BB defense range depends on the opponents. If your opponent opens really wide you should defend more and if your opponent opens tight you should defend less hands. But I think the stakes you play is important as well, because

a, the good regulars at 1/2 are not as good as good regulars at 10/20. The good regulars at 10/20+ are the best FL players and they usually play very loose.
b, the rake at 1/2 is significantly higher then at 10/20. As I know the rake is around 3BB/100 at 1/2 and around 1,6BB/100 at 10/20. It means that there are hands that you can defend profitably at 10/20 but not at 1/2. Well, the diference is probably not too big, only some hands that you can defend at 10/20 but you should fold at 1/2.

Unguarded's range against a 50% Button opener seems okay for me, probably I would defend a little bit less but maybe it is my mistake. On the other hand I dont really like his range against an UTG open raiser. Maybe it is okay at 10/20+, but I think at 1/2 it is too wide. Players at 1/2 open really tight from UTG and HJ, so IMO defend with Q8o style hands is a little bit too much.

And I think it is the same in the BB vs SB situation. Yeah, at SB you should open a tons of hands, at least 60% and against that range you can defend almost every hand. But usually 1/2 regulars are not as good as Tpiranha or Unguarded They do not open enough hands from SB, so I think hands like 84o are not good enough because on one hand their SB range is stronger than it should be and on the other hand the rake at 1/2 is higher than it should be

Nowadays most bots are powerful enough not to bucket preflop, so with those bots it isn't really an issue. In general there are usually only small qualitative shifts between bots that do and don't bucket preflop so I feel confident that the strategies in those articles can serve well as a general guide (and since many preflop strategies can be very complex I prefer to simplify, regardless). This is something Dan also found with the bots from the articles.

I haven't played Neobot but those ranges def sound exploitable to me (although maybe it is exploiting you somehow :P ). The biggest surprise for me was just how low the 3-betting range seems to be in BVB compared to HUHU.

I think Dan has some plans for that in the future. His site is here: http://propokertools.com/

You can also contact him on 2+2 his alias is bachfan.

Haha, neo isn't doing it to exploit me... I only go by the neo vs. neo hands to make sure that isn't the case. But I think it's likely that neo still needs more time to learn and will eventually stop folding BVB. I played 25k hands against Sonia and she was definitely growing in strength from 10k-25k hands before they decided to break my heart by shutting down the site. She was 1000x more fun to play against than GTO bots.

Ya, the 3-betting range is really interesting for Dan's bot. I agree that it is due to having position... no worries about flops checking through. I also like that it uses a mixed strategy. 2008 Polaris's 3-betting strategy always seemed weird to me.

I am going to try to talk Leader into running some nerdy calculations on the difference between BVB 6-max and BVB HUHU, lol.

Thanks man, didn't realize bachfan was Dan. Also didn't even know about your articles or the new book until this week, lol... nice work and and thanks!

Sure thing. The new book is pretty nerdy so you might like it. Watch out for the article next month too!

I have been working on improving my AI abstractions quite a bit in the last few weeks; I plan to re-run the blind vs. blind AI using no buckets preflop and my (much) better abstraction, but my machines are currently pegged computing other things. That said, it seems to be a pattern that AI strategies for pre-flop play don't seem to vary all that much (at least for the GTO-approximators like my AI).

Personally, I'm most curious about confirming the 3-bet percentage and the sb open percentage; I'd be surprised if going to a bucketless preflop abstraction resulted in a BB fold percentage of more than a percentage point or two for reasons already posted; zero still seems the most likely.

I'll post again when I have something.

Cheers,
Dan aka bachfan

And yes, I do plan to release something this year that people can play with. More on that later.

This is an excellent question. My personal philosophy/strategy in building AIs has been to use as little human judgement as possible. I like to give the AI build process as much information as I can and let it figure out what to do with it.

For example, in the AI agents mentioned in the articles, you might ask "well, who decided on the buckets?" I did, in a sense, but the buckets were generated using a strict numeric criterion; I didn't sit down and divvy up the buckets by hand or anything like that.

Cool, thanks bachfan! That is good to hear. I got the impression that 2008 Polaris's pre-flop play was too heavily influenced by the programmers. Your bot almost always uses a mixed strategy with every hand, which is how I would think a GTO bot should be playing. I also like it because it frees me up to have a lot more fun pre-flop in wide range situations since mixed strategies seem to not only be ok, but best.

I am very curious about how you think pre-flop folds affect a BVB situation. In your finished product, I think it would be really cool if you give us an option to simulate 1-8 folds before the bvb battle starts. I have always wondered in general how different it is to open from the button 10-handed vs 3-handed. In the 10-handed case, there should obv be a lot more aces and face cards left in the deck when it gets to us since the previous 7 players didn't open.

We have to be careful, though. Some of the mixing may be because it is "best", but some mixing is surely an artifact of how I constructed these AIs. For instance, when you get rid of the buckets preflop, there are plenty of hands that are played one way and one way only that would have been mixed using buckets; a good example would be 98s, which the BB always re-raises with my latest AI, but uses a mixed strategy with a bucketed AI. (talking about huhu bot here)

I've had similar thoughts. The problem is, we need to define a range that everyone folds, and that needs to be folded in (sorry!) to the AI construction itself.

You could always jut let the user customize the ranges, lol... then you don't have to deal with ppl saying "But in live poker, at my stakes, at 3:31 AM, during a full moon, there's no way the HJ opens A7o ever!!!! These ranges suck!!!!!"

I personally always love being able to customize things when possible.