03-29-06 [poker] - 3

03-29-06 [poker]

I think it's correct to always make solid bets, not too small not too big. Some players make the mistake of betting big with bluffs (to make people fold) and small with value hands (to make people call). Others bet big with their good hands (because they want to win a lot of chips) and small with their bluffs (because they don't want to risk many chips). Both ways are bad because they're very easy to read, however, the first way is MUCH worse. See why?

03-29-06 [poker] - 2

03-29-06 [poker]

Watching "Live at the Bike" tonight. It's the 25/50 NL game, which is the biggest game at the Bicycle Casino. There are a few regular "pros" who play, but they're really awful for the most part. The highlight tonight is this guy Frank Mariani, who apparently is one of the owners of the Lakers. As expected, he's just unbelievably awful, playing every hand and donating like crazy. He's dropped about $25K so far, which is 500 BB's, five buy-ins, quite a lot. This is what I've heard, that sometimes the highest games have the biggest fish (obviously the super low games have huge fish too, but the middle ground is a bit dry).

03-29-06 [poker] - 1

03-29-06 [poker]

In Limit poker, how should your bluffing relate to pot size? On the one hand, when the pot is big, you get a much bigger bonus when you do win the pot. On the other hand, it's far less likely your opponent will fold. One thing's for sure - when calling, you must call much more when the pot is big, even if it seems obvious your opponent must have your beat and you only have something like ace high, when the pot is like 40 bets it's not the time to save your 1 bet, you must call, and lose 90% of the time. It seems to me that perhaps bluffing frequency shouldn't change much with pot size. If you bluff 25% of the time on the river, that should be roughly unrelated to pot size. Of course you should change that if your opponent plays badly. For example, if your opponent just calls based on hand value and not pot size (eg. they just look at their cards and don't account for pot odds), then you should bluff much more in big pots, perhaps 100% of the time, since they will fold too often.

03-29-06 - 2


I don't go to Best Buy for the music, the movies, the games, or the electronics. I go for the "stuff". I can't get me enough stuff.

03-29-06 - 1


Are Chang, Zhang, Xiang, and Giang all different names? Or just different English spellings of the same name? (it seems Jiang and Giang are just different spellings of the same character, but the others are all separate).


03-27-06 [poker] - 1

03-27-06 [poker]

Small interesting hand with Gus in Superstars 3. Johny Chan open raises with 77. Freddy Deeb pushes all in with AKo. Gus is in the BB with JTs. Gus goes into the tank. Why is he thinking so hard? Someone just raised and pushed, you fold, right? Well, not so obvious. First, Johny had about half his stack in, so he's surely calling. Johny probably has something like Ax or a pair. Freddy probably has a pair or AQ or AK. Now, if either of them has a big pair, like 99+, that's very bad, but if they have hands like they did, a low pair and two high cards, then JTs actually wins more than 1/3 of the pots ! Furthermore, Gus is in the BB and the blinds were big, giving him a nice overlay. In the end Gus winds up folding, but I think it's very close, it's not a trivial fold at all. I think there is one big factor making it a fold, which is that if you fold you still have a pretty big stack and can push people around, so you don't need to run a race like that at basically even money since you'll have more +EV opportunities to steal.

03-27-06 - 4


Braise! Poach! Sautee!! I think Americans overuse baking and broiling. Salmon should not be grilled or broiled, it should be poached. It gives it a buttery texture, a tender flake that's truer to the fish. Asparagus should be sauteed (and the ends should not be broken off like the fools on Food Network advise - the spears should be peeled. All kinds of meats are succulent braised (as are vegetables like broccoli or brussel sprouts or carrots).

03-27-06 - 3


Consumer followup report : a while ago I left Wells Fargo for the American Express internet bank. I also dropped my Cingular cell phone and got a T-Mobile "To Go" pay as you go phone. In both cases it was because I was furious about the constant ridiculous fees they were dropping on me. The result of both changes is a huge success. I've been incredibly happy by the lack of charges and low fees at both services, and I have never wanted any of the services that the more standard providers offer (eg. I've never needed to actually go into a bank for anything). I highly recommend both means. If I could get more of my basic bills in a pre-paid or no-fee format I would. Now if I could just reduce my $300/month natural gas habit...

03-27-06 - 2


Why do humans have ass hair? Even monkeys don't have ass hair, it's one of the most hairless parts of their body! I suppose ass hair hides your pink eye from embarassing view, but it inconveniently holds your refuse close to your body and creates dingleberries! Seems like some sort of evolutionary mistake.

03-27-06 - 1


The way Giada de Laurentiis smiles is really freaky. She's actually the grand daughter of the great Dino de Laurentiis (producer of Dune, woot!). I just imagine her scary passive-aggressive-guilting Italian mother always saying "Smile! bigger. BIIGGER! what, you don't have a smile for your mother? Show your teeth, you got sucha pretty teeth, your papa spent so much at that orthodonist, give us a big smile, bambina".


03-26-06 - 3


G. Love was coming to Pozo, but it was cancelled >:( that woulda been cool. Chris Isaak is playing the Chumash Casino. Poor Chris Isaak, he was kind of alright, now he's playing after Blood, Sweat & Tears, Earth Wind & Fire, Emerson Lake & Palmer, and other three-name washed up bands from the 70s who are doing casino tours to pay alimony.

03-26-06 - 2


It seems to me paying taxes on my poker winnings is -EV. Say my taxes are X. If I don't pay and get caught, I'll suffer penalties and interest that might double it, to 2X. Paying the taxes is only +EV if the chance of getting caught is > 50%. I don't think it's even close to that, maybe more like 10%. This is a gamble just like everything, but the +EV seems to be very clearly not paying. It's sort of odd to me to make good EV decisions and make poker winnings and then make a bad EV decision and pay tax on it.

On a related note - I think car insurance is pretty clearly -EV. The car insurance companies just skim a huge profit, and furthermore they don't really provide a service the way catastrophic insurance does, since if the "worst" happens and you car is wrecked, that's not so horrific. They're not really amortizing anything for you.

That is, say the bad thing happens, it costs you -C. The chance of it happening is P, so the expected cost is P*C. A zero-profit insurer would charge you about P*C. Now, if C is very large and P is small, you should be willing to pay a premium, because if C happens you're screwed and have no recourse. This happens for like farmers and other disaster coverage, where if the bad P event happens their business is destroyed, they can't pay C, so it's worth it to pay a premium. On the other hand, if you can easily cover the cost C, then paying any premium for the insurance is ridiculous, the insurer is basically providing you no service.

Of course in the real world unfortunately car insurance is needed because of punative damages and all that nonsense, people suing you if you happen to injure them.

03-26-06 - 1


There's a movie coming out called Old Joy , starring Will Oldham (Bonnie "Prince" Billy), perhaps my favorite musician these days, an avant-folk superstar. I can relate to how the J-Lo fans must have felt back in '97 when she broke out in "Anaconda". How exciting!


03-25-06 [poker] - 1

03-25-06 [poker]

I think I'm going to become a Party Poker affiliate. There don't seem to be any disadvantages to doing so, and I can provide you all with nice bonuses if you sign up through me. I'll put adds for Party on my web site and be able to sign up people through my links. If you're thinking of getting a Party Account - wait and sign up through me !! It will be much better for you...

It seems like the only requirement to be an affiliate is that you have to sign someone new up once every 90 days. So, that's 4 people a year if you space them out. I think I can probably handle that, and the benefits are grande.

Well, that took about two seconds, so here's your link :

Use sign up bonus code "CBPARTY" to make sure you're tracked and get your 20% to $100 bonus.

If you're a serious poker playing friend of mine, email me for details.


03-24-06 [poker] - 2

03-24-06 [poker]

Well, I'm having fun at 200 NL. It's gotten me inspired again. I was getting bored and burned out grinding 100 NL. I was still making a lot of mistakes and didn't feel like I was making progress. Now I'm sharper on my game, and back on the program of trying to move up the levels. My original goal was to make it to 1000 NL before the World Series. It seems very unlikely I'll make it, but I still could if I move up once a month. I need to stay focused and try to improve and move up to 400NL. It's no fun unless you're making progress, like most things.

There are some mild differences in play between 100 and 200. There are more good players are 200, sometimes a whole table is full of semi-pros and I just leave that table. It's a bit harder to find tables full of donks, but they still exist. Overall, everyone is much more aggressive, even the donks. Actually the strategy of waiting for big hands and pushing them still works because people are aggro and will make big calls. If you were weak/tight you would get destroyed, because you'll be pushed off hands too often. There aren't many weak/tights at 200 and they mostly just lose money quickly.

The most annoying thing at 200 is there are a lot more shorties, and most of the really bad players tend to be shorties. It's such a dumb ridiculous move, these guys who take their $100 and buy in at a $200 table instead of a $100 table. It just means the blinds are bigger for them and their stack will get eating up faster. Often when I hunt down a mega-fish at the 200 level, he's on a tiny stack ($50 or so), which is annoying because it makes it hardly worth chasing.

One thing I really miss about 100 NL is it's the only level where $1 = 1 Big Blind. (god damn I hate that I can't write BB since BB often means "Big Bet"). When you have to compute odds or the 5/10 rule or something you can just look at the bet amount and that's BB's. At 200 I have to go, ok, he raised me $22, so that's 11 blinds, okay, I fold. It's another mental step which is bad. A major part of poker for me is cutting out mental operations to make it more mechanical.

03-24-06 [poker] - 1

03-24-06 [poker]

If you have a 3:1 chip lead in heads up, should you race with the worst of it? Assume the ICM model, so before the hand you are 75% to win. (the ICM model is the assumption that your chance of winning is proportional to your chip count). If you lose an all in, it will be 50/50, an even match. What odds do you need to get all in ?

0.75 = P * 1 + (1-P) * 0.5
0.75 = 0.5 * P + 0.5
0.25/0.5 = P
P = 0.5

So, you should take any race where you're >= 50% to win. If you take any worse race, you're giving up EV. Actually, this is not really interesting, it's built into the ICM model. The ICM model assumes that you're running your chips against each other at a 50% win rate. It's an open question how accurate the ICM model is; there are some good data mining projects which suggest ICM is not quite right.

Now, that only applies to calling an all in. If you're putting him all in you have the bonus of fold equity, and if you have to take a race as a dog it's balanced by all the times he folds. For example, say you push something like Q9 or K2. If called, it's probably a 60/40 race with him in the lead. If the blinds are tiny, pushing is a bad idea. What size do the blinds need to be for pushing to be right?

His stack is 1, yours is 3, and he's in the big blind of size B, you're in the small blind. You push. He calls with the top 20% of hands (roughly any ace, any pair, better kings and queens). 80% of the time you win +B and the stacks become {3+B,1-B}, so your chance of winning becomes (3+B)/4. 20% of the time he calls and you race as a 60/40 dog, in which case your chance of winning is 0.4 * 1 + 0.6 * 0.5 = 0.7

0.75 = 0.8 * (3+B)/4 + 0.2 * 0.7
0.75 = 0.6 + 0.2 * B + 0.14
0.01 = 0.2 * B
B = 0.01/0.2 = 0.05

So any time the big blind is >= 5% of his stack, this push is goot. (that's M = 13.33 for him). This is actually sooner than I thought from intiution. I don't usually start making this push until his stack is more like 8 big blinds or less, which means I'm actually waiting too long. Good Sit-N-Go pros know all this and are super-aggressive pushbots once they have a chip lead.

03-24-06 - 3


OMFG, the Andy Milonakis bit called Spoons is like the funniest thing since that Simpsons episode...

Homer : Marge, where's that ... metal deely ... you use to ... dig ... food...
Marge : You mean, a spoon?
Homer : Yeah, yeah!

03-24-06 - 2


Wow, this is hillarious. There's a group here in San Luis which is protesting the new Urban Outfitters store that just opened here. (and they have a web site which is pretty swank). Urban Outfitters they're protesting, because of the novelty books and toys they carry. These are probably the same mothers who get their teenage daughters boob jobs, buy them thongs and short-shorts with paws on the butt, etc. We have an Abercrombie in town, which they didn't seem to mind, because that's good old fashioned American values - they support traditions like the whole football team gang-raping cheerleaders. Okay, I'm going a little off track here. Anyhoo, it's ridiculous. There's nothing that hinders juvenile indiscretions like a "Mustache Rides, 25 cents" T-shirt.

03-24-06 - 1


A thread should not use IsWindow for a window that it did not create because the window could be destroyed after this function was called. Further, because window handles are recycled the handle could even point to a different window.

So, basically this function is just pure unsafe unreliable garbage and you never should have even provided it in the SDK, right? And why the hell don't you provide me with a safe GUID for windows? All you have to do is give me a handle (which is recycled) plus a recycler refcount, very simple & makes this not a P.O.S.


03-23-06 [poker] - 1

03-23-06 [poker]

Poki-type AI's have the very bad property of exaggerating their mistakes. You have some AI function A() which returns {fold,call,raise} probabilities for a given hole and state. You use the same A() in your opponent model - that is, you assume your opponents think the same way you do, and this A() is used in the Bayesian way to guess the chance they have each hole, which you then use to measure how often you're ahead or behind in the hand.

If A() is too tight/weak, it will only bet with very good hands, and check or fold a lot. If you then use the same A() for your opponents, then you assume they are only betting with very good hands, which makes you even more scared!!

Similarly, if A() is too loose/aggressive, it bluffs way too much and plays a lot of junk cards, then you will assume that your opponents also are betting a lot of junk. This will make you want to bluff them even more, because you give them no respect for having hands, you'll try to bluff their bets & call their raises with really marginal stuff.

If you had an accurate opponent model, these flaws in A() would still be flaws, but by using the same A() for the opponent model it amplifies the problem, which is a horrible feedback loop.

Of course bad human players make the same mistake - they assume that the opponent plays the same way they do.

03-23-06 - 2


Whoah, Sharon Stone in Basic Instinct 2 just looks scary. I've never been a Sharon Stone fan; when she was young she sort of reminded me a typical Florida beach girl, probably roller blading and going to Nascar races with boyfriend Bobby the high school quarterback. Now she's the epitome of an Urban Cougar, scary old lady who thinks she's still hot, on the prowl. I definitely don't need a money shot of wrinkled old poon-tang, or implants in old skin (they look sort of like a canteloupe inside a plastic grocery bag).

03-23-06 - 1


You know sometimes you get into an argument with someone, and they're just like completely ridiculous and unreasonable, and you just can't possibly progress in the argument. Like they keep bringing up unrelated things, and using false logic, and accusing you of saying things you never said, etc. etc. You just have to give up and move on. Unfortunately, this jerk seems to be just about every Republican in power these days, and giving up means just letting them do whatever they want with the country.


03-22-06 [poker] - 1

03-22-06 [poker]

I think a lot of people in the 2+2 hand analysis community don't understand a key idea - the "correct" plays that we espouse are only correct because of the way average opponents play these days. If they played differently, the correct plays would be different. Because of this, the correct plays in a tournament are different than in a cash game, even at the exact same blinds, not because of any stupid "survival" issue (which is bogus, BTW), but because average villains behave differently. Similarly, the correct plays at the $1 blind NL and the $10 blind NL are different plays even with the exact same hands and boards. (for example, check-raising on the turn is much more often correct at the higher levels because of the higher aggression level and willingness to gamble). Furthermore, the correct way to play today was not correct 10 years ago, and probably will not be correct 10 years from now.

In fact, online, I feel like I can see play trends evolving all the time. Certain moves come into vogue, everyone starts doing them, and then people start countering them, expecting them all the time, then other moves come into favor. One move I've seen recently pop up a lot is people leading into the preflop raiser as bluffs on the flop. That is, someone raises in late position (usually me), a blind calls. The blind misses his hand, but he leads out pot size. You rarely saw this move a few months ago, now it's quite common, because a lot of late position stealers would fold the flop to this lead.


03-21-06 [poker] - 1

03-21-06 [poker]

Another take on the difference between Limit and No Limit : in a typical low-stakes Limit hand, making a good decision requires you make a good estimate of the probability of him having one of a very wide range of hands - he could have top pair with various kickers, middle pair, overpair, sets, lots of draws, etc.. you need to assign some chance to each and try to maximize against that range. In No Limit, profitability depends more on being able to just the chance that he has exactly *one* hand, or maybe a few. Like when you have KT and the board is KT9, you need to estimate the chance he has QJ and how he would play it.

03-21-06 - 1


Poker AI is a really fun problem. The thing I really hate about it is the community of poker AI developers. There's zero sharing of ideas and work. Everyone thinks their stupid little AI is the holy grail, or is worth money, and they don't want to share anything about it. Even the academic guys don't give out details (the Poki group doesn't actually describe a functioning AI in their papers), perhaps because they don't want people implementing clones which could be used illegally or sold. I miss being able to talk over problems with other smart people, it's just massively helpful.

When you know what you need to do, two full workers are perhaps only 150% as efficient as one. On the other hand, there's this magic network thing that happens. If you have 10 people each working on their own thing, they're 100% efficient each on their own. Now, if you just let them talk to each other a few minutes each day, their productivity goes up massively! A few words from someone else can help you avoid going down long dead ends or send you on paths you never thought of


03-19-06 [poker] - 1

03-19-06 [poker]

I've gone back to a bad habit recently in my poker play, which is giving too much credit to my opponents. The opposite is just playing by pot control, vanilla basic play, like - he bet big, I won't call without a big hand, I don't care that it doesn't make sense, or - I have a big hand, I'll bet big, I have junk, I'll fold. Playing by pot control is a winning strategy against crazy random nonsensical players. In the past few sessions I've been doing things like thinking "he just pushed all-in in a tiny pot, he wouldn't do that with the nuts, he'd try to milk it more and not scare me, okay, I call with the 4th nuts", or "I know he has a weak hand, and I've shown tons of strength, I'll push all in and there's no way he can call". Well, guess what, he did just push all in with the nuts, and he did call my all in with a weak hand. Back to basics.

I think it's since I moved up levels I've been feeling like I need to step my game up to the next level, and I'm trying too hard to make good plays. I need to go back to just playing the way I was, basic solid poker, and then move on from there.


03-18-06 - 5


The PS3 is starting to look like a big disaster. Actually, as soon as I saw the processor specs I should've known it was a disaster. The only question I have is how much that will affect Sony's bottom line. I think a short position on Sony could be a good way to go.

In other investment news, I'm starting to feel more and more like the U.S. is headed for the crapper in the next 10 years, so I'm thinking of getting my money out of dollars. I like to do that by buying foreign stock. The obvious places to invest are India and China, perhaps there are some other good options.

BTW, why do I think the US is headed for disaster? 1) massive foreign accounts defecit, somewhat propped up by just printing more dollars, 2) massive government budget deficit, 3) continued loss of real jobs & real earning rate in productivity sectors (eg. if you take finance and real estate and natural resources out of the equation and just look at the "skill" sector of our economy, it looks really bad), partially due to the growth of foreign countries' skill base and outsourcing, 4) large amount of personal debt and leveraging, which may destroy consumer spending power if - 5) the housing bubble may pop or even just slow, which will crush consumer leveraging, which is a major force in our economy.

Basically the US economy is currently propped up by - massive oil profits (and other natural resource wealth), which is not only temporary but also destabilizing, massive consumer leveraging in the form of credit cards & home loans, and massive government deficits.

03-18-06 - 4


In honor of Cuba's great performance at the World Baseball Classic, you all should watch the movie "Soy Cuba", a beautiful black and white pre-communist semi-documentary about the island. Die, capitalist pigs!

03-18-06 - 3


I started playing with Poker AI again. It started when I was contacted by the author of GNU Holdem who's integrating some of my work. Then I found "mabots", which is a free implementation of the TTH (Turbo Texas Hold'em) AI. That suddenly opened up an exciting possibility - I could run a brute force simulation of poker! The key is that the TTH bots are so simple and fast that you can actually run them about a million times a second. That allows you to explore the full branching tree of possibilities in the hand, and simulate what your opponent would do in each case. I've written about this type of poker play before, but the idea of running a full-tree brute force simulation just seemed ridiculous. Even with how fast the TTH bots are, you can only run the full tree heads up, in a multiway pot you either have to do a monte-carlo sampling of the simulation, or prune the tree to condense branches with similar outcomes.

So, I got this TTH simulator working, and the good news is it completely destroys the TTH bots. At a table full of random TTH opponents, the simulator beats them for 12 PTBB/100. That's just an insane ridiculous win rate at limit hold'em, the best players in the world win at 3 PTBB/100. The very best TTH bots at the same table win at around 5 PTBB/100. Even that is high and it's because many of the TTH bots are so weak. Heads up against the very best TTH bots, my simulator bot wins at a rate of 30 PTBB/100.

The TTH bots are pretty bad, and my simulator quickly finds their flaws. For example, if the hand checks to the river, the TTH bots will often stab at the river with no hand. The simulator will then raise them with any two cards, because the TTH bots will fold to the raise. The TTH bots generally make very predictable stabs at the pot and also fold easily to aggression, so you can trap them and bluff them. I saw another cute example :

Board is [9 T Q] and my simulator has Q7 and TTH has A4. TTH checks to me and I check behind (!!). Yes, it looks drawy, but I get more value because : Turn is a [3] and TTH bets out with air. The simulator raises (!!) and TTH calls. The river blanks and it goes check-check. The simulator estimates that it gains +1.5 big bets from this line rather than just betting the flop, because TTH will just fold the flop to most bets, but here we trapped in 2 more big bets. The rare cases where we get beat are outweighed by all the times he has junk and we win those extra bets.

Basically the simulator is able to play 100% perfectly against the TTH bots. Of course it can't see their cards, and the TTH bots do have a tiny bit of randomization, so my simulator can still make "mistakes" in the silly Sklansky Theory of Poker sense of a "mistake" (eg. if I could actually see its cards, I might do something else). However, the simulator plays 100% perfect poker in the sense that it's the best you can possibly play against them - I put them on the best possible hand range, and in each case I can predict what they would do in response to my actions, since I just run their brain to find out.

Unfortunately, this Simulator is not very useful as a general purpose AI. The weaknesses of TTH are so predictable and unrealistic, that it makes the simulator play strangely. That is, the Simulator is perfect against TTH but it's very very far off a game theory optimal AI, it's sort of at the opposite end of the spectrum, it's tuned itself to be perfect against TTH which then makes it highly exploitable to other play styles. For example, the Simulator against a Poki AI is a losing player.

So, this turned out to be a purely theoretical excercise. Perhaps I need to read some more game theory papers. I suspect there may be some way to prevent the Simulator from becoming so exploitable to other strategies. For example, maybe if I just added some chance that instead of doing the TTH action, the opponent does something completely random at each decision point? That would certainly keep me from zeroing in too much on the TTH opponent, but not sure if it's good or not.

In other words, if you have an AI A1, and you construct the optimal strategy that beats it, A2 = Sim(A1), then you can construct the optimal strategy that beats it, A3 = Sim(A2), etc.. in theory this is a sequence that converges to a fixed strategy AN which is the game theoretic optimal strategy, that is AN+1 = AN (as N -> infinity), the best strategy against AN is just AN. I'm not even sure if this simulator approach will ever converge, or if it will just oscillate wildly, but certainly I've seen that for the first few steps in the sequence, it can swing wildly. Presumably if I made a simulator than ran my simulator, it too would have very strange exploitable flaws.

03-18-06 - 2


Okay, I'm officially up to 200NL. I had a few sessions where I took some beats and roughly broke even, but today I had my first big winning session. It feels a lot bigger than 100NL because everything is doubled, it's actually getting to a level where it's significant money, not play coin. It's fun having a new challenge, but I do feel a bit burned out. I can't seem to play long sessions recently, I only play about an hour and then I'm done for the day, which is nowhere near my goal of around 1000 hands per day.

03-18-06 - 1


Seniority-based pay is bad for capitalism. Some small amount of seniority-based pay increase is good because it encourages employees to stick around and ensures a valuable knowledge base built from experience is in the company to transfer to new hires. However, most traditional old companies and pay structures (especially in unionized and public sector jobs) have severe seniority based pay scales. These strongly encourage people to lock into one job and stay there for many years without changing companies. This is bad for employee liquidity, which is terrible for capitalism (see my previous rant on why employer-based health care is so bad).


03-16-06 - 1


Okay, I'm not normally one to defend wife beating, but I'm watching these old Mike Tyson fights on ESPN Classic and it got me thinking about Robin Givens. Tyson, if you don't know, has the brain of a child, and was just a fearsome physical specimen back then, a destroyer of men and muscled like a horse. Robin was a very smart woman, she can't possibly have been in love with Mike, she must've wanted him for his money, his fame, she must have been using him just like Don King and everyone else in his life used him. If she thought a second she would've known he was dangerous too, a child brain with scary explosive boxing power. I can't really feel sorry for someone who makes their bed with lions.


03-15-06 - 2


I bet you could make a variant of Hold'em that made people have hands a lot more often. Call it "Action Hold'em" or something. The fishies would love it of course because they're action fiends, and the pros should like it too because it means you get to play a lot more big pots with the fishies, you get a lot more real hands per hour so your profit per hour is way higher. Obviously there are already games like Omaha that have a lot of big hands, but they're too complicated or technical. We want the simplicity of Hold'em preserved. There are a few easy ways to do it -

1. Add a joker. I don't really like this option because it adds a big technical aspect involving theory of the joker, and it makes the game feel cheap, like some wacky home game. Also, I'm not sure this stimulates action, since you have to sit around and wait to get a joker in the hole a lot.

2. Get rid of low cards. If you reduce the deck to only cards >= 7 or so, it makes it far more likely for people to connect with the board, etc.

3. Various dealing manipulation algorithms. This is very easy and effective online, but can't be done with real cards. That's a negative. Also, it would make all the moron fishies think the game was fixed in some way. There are lots of possibilities here, like only dealing flops that hit people holes, or probably the best is simply never dealing anyone junk in the hole. Simply do a random deal, and if someone gets pure junk in the hole, throw it out and try again, repeat until everyone has something.

4. Add another street, the "Ocean", after the river, so there are 8 cards to make your hand from instead of 7. This greatly increases the chance of straights, flushes, and other big hands.

03-15-06 - 1


I figured out TiVo's latest fuck-up. GSN (Game Show Network) had all the program schedules off here, it was showing the programs eastern time schedule, but airing them pacific time, so it was all off by 3 hours. TiVo decided to fix that apparently, by dropping GSN from my lineup and adding "GSNP". Unfortunately, the stupid fuck-wads didn't update any of my scheduled recordings or season passes to move them from GSN to GSNP, so all of my recordings on GSN just silently failed to happen, worst of all "High Stakes Poker" didn't record (the best poker show on TV, by far). This is such shoddy programming, I spit on them. Pfftt!


03-14-06 [poker] - 1

03-14-06 [poker]

Why is pot control so good? On the one hand, it's an obvious idea that you will win in the long run if you can play big pots when you have big hands, and small pots when you have weak hands, eg. pot size is proportional to hand strength. On the other hand, isn't it correct to just get the most money in as possible when you have the best hand? eg. if you have one pair (a pretty weak hand), but think he has a worse hand and will call, shouldn't you go ahead and play a big pot to maximize profit? These are like macro and micro views of the same problem. The macro view (pot control) is an overall play strategy. The micro view just looks at each individual choice (fold, call, raise), and says for that choice you should take the option which maximizes EV, which could in theory conflict badly with pot control. How do we reconcile these two pictures ?

First of all, recall that you should be playing Bayesian Poker (see previous articles, etc.). Bayesian Poker at each decision point assigns your opponent a probability for every possible hole, P[H], and also an opponent model of what he'll do if you do certain actions, P(A|H,S). For this to be complete, these probabilities must include the chance that he's playing really weird this hand, that he's being tricky, that he mis-read his hole cards or the board, that he's just not paying attention, that he mis-clicked a button online, etc. etc. In practice, we don't actually think of all that when we do "hand reading". What we do in practice is just to think of their primary play style and figure out what hand they would play that way. eg. based on the play so far and the board, I put him on hand X, or perhaps hand Y. I'm not actually thinking of a full range of probabilities and all these other possibilities, because it's just too much to go through.

Now, this type of "deterministic" opponent model is pretty good, it's right 90% of the time, and if you make decisions just based on "putting him on a hand", you will most of the time make the right decision. The problem is those few times he's deviated from it, you might make the wrong decision. Now, those events are rare, so they're not a big deal, right? Well, in Limit Hold'em, that's true and we can just stop the discussion. In No Limit, however, the pot size can vary, so that a rare event can be disastrous.

Just as a quick example, let's say you decide you're never going to fold the second nuts. Most of the time that's a fine strategy and you'll win many pots that way. Once in a rare while, though, he'll have the true nuts. If you lose a huge pot in that case, it may be disastrous to your EV even though it's incredibly rare.

Pot Control is a way to hedge against these rare disasters. It means you only play big pots with good hands, which makes it even less likely that your opponent can have a good hand when you play a big pot (because the chance of you both having very good hands is very rare). If you will play a big pot with one pair, there might be 300 holes (out of 1200) that make better hands, but if you only play a big pot with a set or better, there are usually 50 or fewer holes that make better hands.


03-13-06 - 1


Poker Superstars III has Howard the Great announcing, which is awesome, but still the structure is so fucking awful that I can't watch it. The blinds are so big, there's no play at all, no skill, no moves, it's just a push-bot fest. I think any of the online Sit-n-Go pros would be a favorite over these "big names" who are okay, but not experts at proper push-bot play (and there are a *lot* of online sit-n-go crushes who totally rule at this type of structure).

03-12-06 [poker] - 2

03-12-06 [poker]

You can open raise almost any two cards, continuation bet and take the pot down. You can make a big reraise preflop with almost any two cards and take the pot. When you do it with KK and everyone folds, you start thinking "why bother even doing this with good cards, I can do it with garbage". Of course that's not true. Yes, you can do it with any two cards and in some cases you should, but having cards greatly increases your equity for the cases where they don't just fold.

03-12-06 [poker] - 1

03-12-06 [poker]

(see 03-09-06). On trying to reduce the draws :

The board is B, you have a hand H1, your opponent has H2. You're trying to simulate all the draws from the deck {D} which has excluded the cards already on the board or in your hands. You can generate a hand rank which is R(B,H1). A draw card may or may not affect your hand ranks. The hand rank is a list of numbers, the first is the type of hand (no pair, one pair, etc.). The second is the rank of the card of the most important part of the hand (eg. the rank of the highest card in your flush, the rank of the paired card in your one pair, etc.), the next value is the second most important card.

For example, a house 88844 would have a rank {House,8,4}. Two pair JJ44A would have a rank {TwoPair,J,4,A}. One pair 44AQ2 has a rand {OnePair,4,A,Q,2}. Note that only "No Pair" actually has 5 significant ranks.

In some cases a draw card doesn't affect your ranks at all. If you draw card C, and R(B,H1) == R(B+C,H1) and R(B,H2) == R(B+C,H2) , then C was a total blank in terms of affecting your hands. However, you may still need to consider C because it may have been a scare card. We can approximate and ignore any changes to the hand rank beyond the first two. That is, for one pairs we consider cards that change the kicker, but not cards that change the second kicker.

We can define a "scare card" as a card which changes the hand rank of a likely holding, even though neither one of you may actually have that holding.

03-12-06 - 3


Downtown San Luis has pretty major (in a relative sense) streets named "Toro" , "Morro" and "Chorro", which I find very confusing. Of course Santa Barbara is much worse with "Cabrillo" , "Carillo" , and "Castillo". You hear lots of amusing cell phone conversations with people giving/getting directions that are like "Ok, I turn on Carillo, wait, Castillo? okay, okay, Carillo, oh, you said Cabrillo?". Perhaps it was a joke played on us by some whimsical city planners of olde.

03-12-06 - 2


So, I installed "Windows Defender" hoping it would do something about Party Poker's snooping, but it doesn't seem to address that domain at all. On the plus side, it is pretty F'ing cool. It provides a service that should've been in Windows all along - whenever some app tries to change your system settings, Defender blocks it and puts it in a list for you to approve. Then you can either apply the changes or refuse them. It prevents apps from changing your file extension associations, your startup apps, etc. It's reasonably unobtrusive too, at the moment I recommend it.

In the mean time, I'm still searching for a decent anti-snoop (anti-hook) app that doesn't gobble CPU or cost $50. I might have to just figure out how to write my own. Unfortunately there doesn't seem to be a Hook for the settings hooks ;(

03-12-06 - 1


A Google search for "cpx win32 profiler" points to my own rant asking for said thing. Curse you Google, that's no help!


03-11-06 - 2


Old cooking tomes will tell you bread should be stored in a bread box, or failing that, in a paper bag. Not refrigerated or kept in plastic. The goal is to not ruin the crunch and flakiness of the crust. That's all well and good, but bread left out like that goes stale in about 24 hours. In real life it's best to store bread in an air-tight plastic bag and refrigrate it, if you're going to toast it when you eat it, since the toasting will restore the crunch, and it will keep much fresher for a few days.

03-11-06 - 1


Fucking Excel holds an exclusive lock on any file it has open. WTF is wrong with you MS? How can you not follow basic standard practices of software and OS behavior !?!?!


03-10-06 - 2


OMG, someone linked to Prahlad's SoundClick tracks - 18 tracks by my favorite freestyle rapper. I think "Green Streams remix" is actually half decent, but listen to some of the others, some of them are HIGH-LARIOUS. "Spirit Rocks" is awesome.

03-10-06 - 1


"Microsoft Malicious Software Removal Tool has changed your homepage". LOL, mother fucker. LOL.

(BTW, the whole idea of Microsoft making an anti-malicious-software tool is about as funny as the LAPD giving anti-brutality conferences, or the U.S. making "humanitarian intervention".)


03-09-06 [poker] - 3

03-09-06 [poker]

Did I mention playing poker with a gamepad is awesome? So much easier to play lots of tables. (and everything with a gamepad is rad)

I haven't played much in the last few days; I finally played some today and I could feel the rust. Human poker playing is about rhythm and feel, and you really need practice and need to feel in the groove. I usually have to blow some chips off before I get back into it. Today was pretty slow, and then a beautiful thing happened. A table busted up, people left, and it was just me and one guy left. I love playing heads up, I can push my edges much better, and wonder of wonders - this guy stayed at the table, and no one else came to join as often happens online. We played about 20 hands and I had his whole stack. He rebought and we played like 10 more hands and I had that stack. He rebought one more time for a half stack and I took that too. By that time some more people joined the table to spoil the fun, and he ran off.

My "roll" is now only a few dollars away from the threshold I set for moving up to the $200NL game, so the next time I play it'll be at the next level !!

03-09-06 [poker] - 2

03-09-06 [poker]

I mentioned before about the importance of different streets WRST stack sizes. One thing that's not obvious is that the deeper the stacks, the less preflop matters. Think about this - say you have AK and your opponent has 47. He has the advantage of knowing you play good cards, you have the advantage of having better cards. How do your better cards help you? Well, the only real advantage of AKo is that you have a better chance of making a better pair. But he's not going to pay you off with one pair anyway. Both of you can make two pairs, straights, trips, etc. Now, you have the advantage that if you BOTH make two pair, yours is better, so you will get a small edge from that over time in the long run, but that's very small.

03-09-06 [poker] - 1

03-09-06 [poker]

In my poker simulator, I have to consider drawing the next card. I currently just draw all possible next cards, which is 45 cards when drawing the turn heads up. That doesn't really seem necessary, since a lot of cards don't change our relative hand values. I think it's mostly easy to tell when a card matters, though there are funny cases, like when a high board card comes out that counterfeits my kicker. For example say I have A8 and he has A3 and the board is 55K2 , any draw of a 9 or higher changes the hand but isn't obvious since neither of our hand ranks change. Ideally what I'd like is to merge all draws cards that have the same affect on the hand, but it seems non-trivial to tell what draw cards really are the same.

The first obvious thing is that if there are no flush draws, then the suits are irrelevant, so drawing a 3c or a 3s is the same. If there's one flush draw then you only have to consider if the flush suit comes or doesn't, so it's like there are 2 suits instead of 4. Then the other big group are all the blanks. There are a lot of blanks though that can change the "situation" even though it didn't affect the actual hands. For example an ace that comes out looks scary and will change behavior even if it didn't affect the hand values. Some blanks may make straights possible even if they didn't complete any actual straights.

It seems possible to me to cut it from like 45 draws to maybe 10 on average, which is a pretty nice speed up, but it's not at all trivial.

03-09-06 - 1


I've just discovered that the clib srand() is a P.O.S. ; if you srand() with low values then the next few calls to rand() give you very predictable results. So, I did some hacky thing which seems to work ok -
void mysrand(int value)

	value ++;
	srand(value + value * 67 + value * 1031);


03-08-06 [poker] - 2

03-08-06 [poker]

Poker is not some mysterious problem that's hard to solve. The perfect way to play poker can be formulated exactly : the ideal way to play hold'em . That, of course, is assuming you have a perfect model for how your opponent will act with given cards in a given situation. That is, if I have OpponentModel(board,hole,history), then I can run IdealPlayer[OpponentModel] and it will play as well as is possible by any means given those inputs. No human could ever possibly beat this IdealPlayer, assuming that I had an OpponentModel() which simulated exactly how they played. Note that IdealPlayer[] doesn't know their hole cards and doesn't know what cards will come in the future, it just knows what they would do in concrete situations.

As an aside, if you wanted the "game theoretical" optimal solution, you simply assume that the opponent plays the same style as you, and set OpponentModel = IdealPlayer[OpponentModel], and you have a recursive equation which you can solve. Of course you can't actually solve it, or maybe you can but it's a hell of a bitch and no one has done it yet. (this could also have multiple solutions, some of which are false solutions)

In practice, you actually can run this IdealPlayer[] on modern machines with a simulator sort of like IBM's Deep Blue. The problem is finding a good OpponentModel which is accurate and fast. OpponentModel gets called around 1 million times per decision, and ideally you could call it even more. That's not bad, but if you try to use another IdealPlayer for OpponentModel, then you get 1M*1M and it's out of control. So, you have to use some sort of cheaper heuristic CrappyOpponentModel, and that leads to problems. It means that IdealPlayer plays perfectly against CrappyOpponentModel, but that may not actually be very good play.

For example, CrappyOpponentModel might fold too much if you just bet like crazy. That will give you an IdealPlayer that's just a maniac, betting all the time. In order to train an IdealPlayer to really play great poker, you need an OpponentModel that's on a pretty high level. In terms of the standard levels of thought thing, if OpponentModel is on level N thinking, then IdealPlayer is on level N+1. The problem is that to take each step of levels you need to run the previous level about 1M times. So to do level N thinking you're O( (1million)^N ).

03-08-06 [poker] - 1

03-08-06 [poker]

It's a huge mistake to overthink in poker. You need to figure out what level your opponent is on, and then just think one step past him. eg. if he's on level N you need to be on level N+1. If you try to act on level N+2 or N+3, you can do very badly. It's because of this that mediocre players can actually do really well against donks, while very good players who are stuck on higher level thinking will do very badly.

One example that I run into a lot is thinking that someone is "setting me up". Here's an example from a few days ago :

I open raise in the CO with 87o. I have a very tight image, so I like to put in some raises to take down some pots. The button calls. The flop came something like QT3, a total whiff for me. I continuation bet like normal, and the button raised to 3X my bet. I folded, and the button showed K8o - pure garbage. Ok, so the button raised my cbet and showed it.

A few hands later I get 99. I open raise in early position, and the same player who was button in the previous hand calls. Flop comes all rags, like 367. I continuation bet again, about pot size. Aggressive Player raises to 3X my bet again. Now what? My hand is actually really weak, he can have a set, a straight, two pair, even against two overs I'm not that far ahead. Then I start thinking, he knows that he just showed me a bluff and I must think he's a bluffer - he wouldn't do this again on a pure bluff, he must've set me up and now he actually has the goods !? Then I realized I was being too tricky, I could tell from this guy's play that he was just really bad and I didn't think he was on that level of thinking. His level was more like "this guys is a tight/weaky, he folded before, I'll make him fold again!". So, on that level my hand is good. I pushed all in, and he called with just two overcards, and my hand held up.

03-08-06 - 1


It seems like there's no profiler built in to VC any more? I need to do some profiling; I guess I have old copies of VTune from old dev work but not even sure if they work with VS.NET (they were VC6 I think). I really wish there was something like "CPX" for the XBox, that thing was so F'ing sweet.


03-07-06 - 1


I need something like "SnoopFree" to block spyware hooks, that's not going to eat 50% of my CPU. Anyone have any ideas? StopZilla maybe. Adaware Plus seems to do this, their free version is pretty cool, I use it all the time for spyware scans, but I'm not sure if Adaware Plus is the goods or what.


03-05-06 [poker] - 2

03-05-06 [poker]

When you're drawing, you really want to be up against good hands. If your opponent has a good hand and you miss your draw, okay, that's just chance. If your opponent has rags and either bets you off or is able to check it down and win, that's horrible for you, because it means that you weren't really drawing at any profit in the first place!

Say you have A2s in the SB. CO limps, you complete the small blind, button checks. Pot is now 3 BB. You flop a flush draw. Now, you can either lead out for $2 , or you can check and you're sure someone will bet. So either way you are paying $2 to improve and see the turn.

The great thing about leading out here is that it reduces your opponent's hand range to hands that will pay you off more if you hit. If you check/call he might have complete rags, which don't pay you off if you hit. Any time you're on a draw, you really want to be up against good hands, because they give you implied odds. If you might be up against rags, it makes your draw much worse!!

03-05-06 [poker] - 1

03-05-06 [poker]

Let's consider limping vs. raising small pockets, playing for sets. You flop a set about 1/7 times. Assume you always have the best hand if you hit your set, and don't if you don't. Half of the time you hit your set you get action, and in that case you win 10X the pot size on the flop.

If you limp for 1 BB , say you see a flop and the pot has 5 BB in it. No set, you lose -1BB. If you hit a set, half the time you bet and just take down the pot = +4BB. When you get paid, you make +49 BB. Net EV =

-1 + (1/7) * ( 0.5 * 5 + 0.5 * 50 ) = + 2.93

If you come in for a raise for 4BB, assume just 1 caller and the pot has about 13 BB in it. No set, you will continuation bet and take down the pot about 50% of the time, 50% you also lose your cbet. When you hit your set, 50% of the time you just get the 13 BB pot, and the other half you get 10*13 BB, but we'll cap that at 100 BB because that's the usual full stack (we use 104 below because you also get your 4 BB back). Assume your cbet is 9 BB, then Net EV =

-4 + (6/7)*(-9 + 0.5*(13+9)) + (1/7) * ( 0.5 * 13 + 0.5 * 104 ) = + 6.07

So, definitely better, but not 4X better. We can see the key benefit is that by building the pot early, it scales up the sizes of everything which allows us to win a much bigger pot when we hit. The cbet factor is actually not necessary for this to be profitable, but it does give us a nice extra bit of value.

03-05-06 - 4


Grid Wars is half way decent if you don't want to buy Mutant Storm (you should though).

I've been trying to find a 2d vertical scroller shootem-up for the PC that uses a game pad, and I can't find one (!?). You know, something like those sweet old arcade games, you power up with multi-shot, etc. I'd love to make one of these games, but making games for the PC that require a gamepad is not +EV.

03-05-06 - 3


Read a tiny crappy thing about Copernicus in the NYT and it made me think along this line : when I read about old scientists like Copernicus, Darwin, Newton, etc. one of the things that always strikes me as being so different is the concept of working alone over such long periods of time, and coming out at the end with this work that's ground-breaking. I know this is somewhat of a fantasy, but to some extent it did happen back then. The loner scientist could work on an idea that they'd hardly talk to anyone about, sometimes over many decades, and over all that time no one else had come to the same idea and beat them to the punch. These days it's hard to imagine doing anything like that, you have to work from a base which is already highly developed, you have to work with a team, and you have to work very fast, and even then your idea is probably a small step, and many other people are surely working on the same thing. The last thing that I can think of that's similar was Einstein's General Relativity, where he really did disappear into his office in his home in Princeton from 1911-1915, and when he emerged there was this new idea which no one else was working on. (which of course is not really true, since Einstein got a lot of help from others and was frequently publishing his progress and giving talks).

03-05-06 - 2


Well, I've given up on "ProcessGuard" and "SnoopFree". ProcessGuard is just horrible for a developer because every time you rebuild your app, it thinks it's a new app and requires you to validate it again. I guess if I wasn't a dev PG would probably be okay. SnoopFree on the other hand actually blocks some useful things, but it seems to just chomp on the CPU. I'm guess it's hooked itself in all kinds of kernel calls to monitor what's going on, but it's just ridiculou, it takes 50% of my CPU under pretty normal use. So, I'm back to Party snooping my system.

03-05-06 - 1


Hernandez wrote me about problems with rdtsc on hyper-threaded machines, which I believe, but I actually forgot until just now that I had written some paranoid safe and accurate Timer code . I mainly wrote it because of the freaking SpeedStep on my laptop, but it should be a safe timer that handles the QPC jump-back bug, and TSC oddities, etc. Basically it uses several timers, tries to use the finer grained one, and if it detects the finer one is wrong it steps down to a coarser one.


03-04-06 - 1


The whole RIM/NTP thing seems like a ridiculous travesty. NTP is a patent holding company, something which is unfortunately pretty common these days. They bought up some ridiculously general patents to an idea that was completely trivial and obvious, and their sole business is charging people who infringed on them. Ironically, RIM successfully got all of the patents invalidated as bogus, but that didn't prevent the courts from imposing an injunction to shut down the Blackberry service, so RIM settled and NTP made a $600 million dollar payday for being crooked bastards. This is sort of like the old Stac case, except that RIM's lawyers are apparently not as good as Microsoft's.

The most absurd thing about it all is that many business analysts have said that this is a "great day for capitalism" because an inventor's intellectual property rights were preserved, and that if anything this shows that our patent laws need to be strengthened to defend innovators and research. Holy shit, you guys have no clue and are so bad for the universe.


03-03-06 [poker] - 3

03-03-06 [poker]

So, I found out PartyPoker installs a bunch of system hooks to spy on you to checks for bots and things like that. Among other things they install a key logger, and they take screenshots of your whole display every so often, and send all that info back to headquarters. On recommendations I've now installed "ProcessGuard" and "SnoopFree" which hopefully will block some of that. Pretty evil shite, those bass-turds. Unfortunately both PG and SF are pretty user unfriendly, and I have to pretty frequently click popups that say "yes, allow that". It's sort of like trying to use Internet Exploder in high security mode - every thing you do you have to click a thousand boxes saying "okay, allow that", blah blah blah.

I just added code to GoldBullion so you can play poker with a gamepad. It's pretty sweet, much more comfortable than a keyboard or mouse. Mutant Storm is still the best way to test a gamepad.

03-03-06 [poker] - 2

03-03-06 [poker]

The later you get in the hand and the more you know about your opponent's hand, the more value you get on your bets. On the river, for example, if you know your opponent's hole, you get 100% value since you can bet when you're ahead and if he calls your EV of betting is +bet. The earlier you are in the hand, the less ROI you get on your bets, because 1) a lot of cards can come that beat you, and 2) his hand range is wider, including hands that beat you. What this means is, you want to try to put in more money later when his hand is better defined and you know that the board hasn't come to beat you.

03-03-06 [poker] - 1

03-03-06 [poker]

Raising suited connectors is great, but not for the reasons people think. Yes, disguise is nice, yes, being able to win without hitting your hand is nice. The real reason why raising suited connectors is great is because it reduces your opponent's range to better hands. Rather than playing a lot of small limped pots with them, you play fewer raised pots. When you have an SC you really want your opponent to have something good enough to get stacked with if you hit your draw. What you really don't want is for them to have something like K5o, which is a favorite against your 67s, but will not pay you off when you hit. You want to force them to have a big hand to play against you, which means you'll pick up more pots when they have nothing, and then when you do hit you can win a big pot.

The sum of EV components with an SC goes something like :

- cost of opening preflop
- cost of drawing postflop
+ winning small pots when they have nothing and fold to a cbet
- winning medium pots when they check-raise or trap your cbet
+ winning huge pots when you hit your draw
If they're on junky hands, all those components are there, but you won't win that huge pot. Another factor for raising suited connectors is that you drive out higher suited hands. 67s really doesn't want Q3s of the same suit in the hand because it can break you if you both hit your flush.

03-03-06 - 1


I thought Cach� was pretty great. It was kind of funny seeing it in the theater and hearing half the audience going "oh no, what the fuck, this is bullshit". The acting is pretty fantastic, all the interpersonal relationships are very realistic and interesting.

This Video of Spore is pretty awesome. WTF is up with the foliage, though? There are like 10 blades of grass in the world guys, get a proper decorator renderer. The procedural creature stuff is awesome. The world looks pretty boring, actually it reminds me of this ancient Amiga game "Drakken" (unrelated to the later Surreal game "Drakan"), which was this RPG with these giant, empty worlds scattered with random crap. I find these games cool in a purism sense, but they're missing a huge aspect of why I play games (when I do) - which is to see the cool art that someone has very specifically made to be rocking cool. Spore seems like a bunch of different games tacked together, which may or may not be cool. The question is whether tacking them together makes the whole better than the sum of the parts. In the worst case games like this often become as bad as the worst of its parts.

I read that the PS3 might cost $900 per unit to produce, and presumably they'll sell it for around $300, which is a ridiculous loss of $600 per unit !! I'm sure they justify this internally by saying it gives them the gateway to the home media system, which they can leverage to sales of lots of high margin products, like games and Blu-Ray DVD's. In practice, many people have tried this idea in the past (sell a big loss leader to get the user attached), and it's never ever worked. It was MS'es big mistake with the Xbox, which they seem to have corrected for the Xbox360. Sony got it right with the PS2 (making it reasonably cheap to make), and seem to have totally fucked up here. Supposedly the expensive piece is the blu-ray DVD reader, which is stupid.


03-02-06 [poker] - 3

03-02-06 [poker]

When someone makes a really rotten play against you and beats you in a pot, you've basically "paid" for the information that they will make plays like that, and you need to keep playing against them to make it back. It's true that poker is "one long game", but there's also a cost of acquisition of information. If you leave that table and go sit with some more unknowns, you'll have to pay to learn about them. Once you're sitting with someone who has large exploitable flaws which you've identified, that's a very valuable state, and leaving that situation is -EV, even if you can find more bad players in the future.

03-02-06 [poker] - 2

03-02-06 [poker]

I added a new feature to GoldBullion to show my stats by card group. These are my win rates with various hole card types, in terms of PTBB/100 (1 PTBB = 2 big blinds) :

groupstat BigPair		474.37
groupstat Pair 		137.79
groupstat SuitedBW		52.8
groupstat OffsuitBW	68.26
groupstat SuitedC		-14.89
groupstat OffSuitC		-12.48
groupstat SuitedA		27.13
groupstat SuitedK		-2.76
groupstat SuitedTrash	-2.39
groupstat Trash		-9.63

This is all at 6-max. First of all, paying the blinds at 6-max costs -12.5 PTBB/100 , so if I just folded every hand my win rate for all of these would be -12.5 So, with "Trash" , I'm basically folding all of them and occasionally playing them and winning a tiny bit.

There are two big surprises here, though. One is that I'm doing better with Offsuit Broadway (things like KJo) than I am with Suited Broadway (same hands, but suited). That could just be an anomaly from variance, or it could mean I'm getting too stuck on suited broadway, or that I play them to a raise too often, something like that.

The big problem is the connectors. Suited Connectors and Offsuit Connectors are big losers for me, even worse than if I just folded them every time I got them!!! That's clearly wrong, I must be playing them wrong, so I need to look into that.

I also added a new feature to show in real time the list of holes that are currently beating you and the chance you have of improving to beat them. If there are lots of holes that beat you it show a representative from each group. For example if your hole is
[ K T ]
and the board is
[ T 9 7 ]
the display shows something like :
[ T T ] Set 3.5%
[ T 9 ] 2Pair 17.6%
[ Q Q ] Pair 20.6%

While I was testing it I was playing on the play money tables and I had [ Q 6 ] and the flop was [ Q 6 6 ] and I found myself thinking woot, I flopped the nuts, and then I noticed it there in the display :

[ Q Q ] House 4.4%

In play news, I'm back on a brutal brutal dry spell. I'm finding the coding much more satisfying, because Visual Studio only gives me rotten bad beats once in a while. As a nice example of how bad I'm running, I got all-in with [ Kd Qd ] on a [ Jd Th 2d ] board against just top pair. I have 2 overs + a str8 draw + a flush draw = booya!! And ... I missed. When you're running really bad it becomes almost funny, every hand I'm just waiting for god to find a way to fuck me. One of god's favorite cruel jokes is for the board to pair when I have two pair, counterfeiting me against the donk I was going to bust who had just one pair.

It's not surprising that two pair being cracked seems to happen so often - it does happen a lot !! With two pair on the flop vs. an overpair, the overpair wins about 30% of the time, it's slightly better than a str8draw and not quite as good as a flush draw.

03-02-06 [poker] - 1

03-02-06 [poker]

One of the cool things about reraising AK preflop in a 100 BB game is that it effectively makes the stacks smaller. AK is a great hand for making one pair hands, and a good one pair is great in a short-stack game, but it sucks in a deep stack game. If someone opens for 4 BB and you reraise to 12 BB, then the pot will have about 25 BB if they call. Now you each only have 88 BB left, which is just over 3 pot size bets, and it's prefectly reasonable to play one pair for your stack in that case.


03-01-06 [poker] - 3

03-01-06 [poker]

Wow, the live game tonight was brutal. I felt like I was just so pitch-perfectly on my game, it was disappointing not to win. On the plus side, I did get to make a big bluff with THE HAMMER (72o) and showed it, that was nice. I play enough serious poker all the time, I like to just mess around a bit in the live game, it's so low stakes I don't really care too much about winning. I do want to win just so I can beat Dustin and hopefully bet player of the year, but more important is pulling some fun moves. You can read Dustin's version of some of the hands at The Chapel Perilous .

03-01-06 [poker] - 2

03-01-06 [poker]

If your stack is >= 100 BB, you can c-bet the flop almost 100% of the time. If your stack is <= pot on the flop you can cbet the flop almost 100% of the time. In between, c-betting so predictably is a mistake. C-betting predictably is worst when your stack is around 10-15 BB preflop. See why?

You can call with a flush draw on the flop when calling puts you all in. eg. if he bet pot, if your stack is <= pot, you call. Similarly you can call pot with a flush draw on the flop if your stack is >= 10X pot size on the flop. In the region between you must fold. See why?

03-01-06 [poker] - 1

03-01-06 [poker]

On the river the pot has $60 in it, he bets $40 into you. You have $60 left. Should you fold, call or push? Well, any of them is reasonable. Many people make the mistake of thinking that calling & pushing are the same decision here, they are not. In order to call, your hand must be good >= 29% of the time. You won't have much left, but there's no need to put in any more chips unless you're good >= 50% of the time.

Another common case is when you have something like two pair on the river, but the flush card has hit. He checks. You think he either has one pair or a flush. Should you try to get in a value bet here? The crucial factor here is actually whether he will raise with anything but the flush. If he will only raise with the flush, you can fold to a raise and you should value bet. Let's look at that. Let's say 75% of the time he has a pair and 25% of the time he has a flush. The pot is currently $100. You bet $20, which is the most he'll call with a pair. With a pair he'll call, with a flush he'll raise and you fold.

EV(check) = 0.75 * 100 = $75

EV(bet) = -20 + 0.75 * 140 = $85

Obviously. But now, what if he'll raise without the flush? What if P of the time he has just a pair he pushes for $100 more and he also does that with the flush? First of all, should you call that? The pot is $140 and you have to call $100 to win $340. You win (0.75*P)/(0.25 + 0.75*P), so

EV = -100 + (0.75*P)/(0.25 + 0.75*P) * 340

Let's say he chooses P so that it's neutral whether you call or fold (which is the correct game theory thing to do).

100/340 * (0.25 + 0.75*P) = (0.75*P)
100/340 * (1/3 + P) = P
100/3 = P * 240
10/(3*24) = P
5/36 = P = 13.889%

So, in the case that he pushes you just fold and it's zero ev. Assuming he still calls in the other cases you get :

EV(bet) = -20 + 0.75*(1-P)*140 = $70.4

That's worse than the EV of just checking, and that's still assuming that he calls your bet. If he just folded his one pairs it would be disastrous. Say for example he folds his one pair 50% of the time, then your EV is :

EV(bet) = -20 + 0.75*(1-P)*(0.5*140 + 0.5*120) = $64

Note that this is even worse than if they just always folded their one pair and always pushed their flush. Of course we could get some value back by betting more bluffs in addition to two pairs which would make it neutral for them to fold their one pairs, etc. etc.

03-01-06 - 2


K.T. Tunstall is like the flavor of the month, but she did this one man band thing on Conan that was actually pretty rocking.

03-01-06 - 1


OMG , scroll down to the bottom of Rakebreak and listen to Prahlad's rap ! (some of you may have seen another sample of his freestyling on a WSOP show, which was equally awesome).

old rants