4

I'm currently developing a poker bot that plays online by itself in zoom poker and I'm looking at optimising the strategy.

So far I've noticed that setting an investment limit based on equity is more successful than EV. For example I set the call or bet limit to equity^4 +0.04. If the big blind is $0.02 and initial pot size of $2.00. So to invest $1 I would need 100% probability to win. For turn flop and river I then power the equity by a different number, which makes the criteria more or less stringent.

However it seems this strategy is still too simple. In your view, what are the key parts missing? General, is it better to play more or less aggressive pre or post flop?

  • Just set the bot to go all in everyhand. Easy. – user3492 Aug 14 '15 at 6:07
  • 1
    Sure, this is a glib response but I can imagine many zoom games where this strategy wouldn't be the worst. Especially on a site like bovada where identities are anonymized. You'd probably have to rathole a fair bit to have any prayer of making this profitable though. – Chris Farmer Aug 14 '15 at 17:55
  • Going all-in with everything in a game where everyone enters with monsters doesn't sound so profitable 8). You're guaranteed to win 5 pots in a row though, but not the 6th. – user1165 Aug 14 '15 at 18:55
  • I agree. That's why I said you'd need to rathole frequently. But the all-in-all-the-time strategy would play better in an anonymous zoom game than in any other currently offered online game, I think. Still probably not +EV, I'll concede. I'm not about to go try this out myself. Maybe at microstakes after a few beers when there's nothing good on TV.... – Chris Farmer Aug 14 '15 at 19:05
  • @ChrisFarmer Fortunately there's always something good on TV^H^HNetflix. – user1934 Mar 24 '17 at 5:27
2

In my opinion EV should not be used in all cases, that's the missing part.

An AKx hand on A-x-x will give you a monstrous EV heads-up or vs few but the combination of the specific gametype (Zoom) and the non-monster hand you have should ignore the monstrous EV and play like a TP or less.

Zoom is a game where all kind of monsters hits the flop and a TPTK hand, regardless its strength, is a hand that needs to be played like Bet -> Cbet -> Check/Call (pot control) and not by bloating the pot as dictated by EV.

Indeed the strategy is simple, as EV is a guessing business anyway; The winning odds you feed your EV formula are less accurate than in cash game for example, where the play is way more normal due to normal card distribution.

  • The winning odds are gained by a Monte Carlo simulation of around 5000 hands. This should give a relatively accurate odds of winning, wouldn't you think so? – Nickpick Aug 12 '15 at 10:36
  • @Nicolas, i'm not arguing about the winning odds, rather the educated guess of hand ranges between cash and zoom hand ranges. On cash games you can give a 99+,AJ+ hand range on UTG, on zoom games this range may be 22+,AK,A2s+. I don't know what hand ranges you give in your villains, but you'll find than most of the times you're against real monsters and your AK hand may have a huge EV to play. This doesn't happens as frequently as in cashes, that's my point. EV in cashes is good, but rather overrated on Zoom, my opinion. – user1165 Aug 12 '15 at 11:24
  • I see, basically there is an adverse selection of opponents as all the bad ones will fold early. So if somebody stays in the game they should be handled very cautiously. – Nickpick Aug 12 '15 at 11:37
  • @Nicolas, exactly this. I don't say you should not play AK or AQ, just don't rely too much on their monstrous EV if you are in a A-x-x heads-up. Zoom is a bad place for strong aces, in microstakes this is double-true, although that's my humble opinion. I would rely on EV on cash games because there the hand distribution is normal; everyone gets its junk share of cards, therefore EV is normal. – user1165 Aug 12 '15 at 12:46
1

The best way to define a good strategy is to get tons of hands and than use machine learning algorithms(Predictive Modeling, Data Analytics, and Artificial Intelligence) to analyze lots of situations and define a strategy as a result. With this you will be able to make a really good strategy for that kind of stakes.

The way it would works is you get a big sample of hands, and than you ask yourself a few questions.

Example:

How many people do a bet out of position after they raise preflop and check fold the turn? You precompile the result of this question from the millions of hands you have and use this as a betting pattern if it give you a good ROI.

The bot would than just query the database asking for instructions all the time.

It could also be done in realtime but adds in complexity.

The more hands you'd have the more effective the strategies would be.

(You made me curious i might start a bot also now lol)

  • You are welcome to join the project. It's open source at www.github.com/dickreuter/poker – Nickpick Nov 14 '16 at 17:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.