Hacker News Reader

Comment by nostromo | Hacker News Reader

The short version:

Because of the recession casinos are desperate for high-rollers. To lure them in they offer perks. Some of those perks include slight modifications to the rules and a "loss discount" -- for example: losing $500k only costs you $400k.

This guy negotiated rules that changed his odds to 50/50. Then he negotiated an 80% loss discount, which meant for each $1 of upside, his downside was $0.80.

So to pull this off, you need $1m to put down, great negotiation skills, even better blackjack skills, good luck, and a very sophisticated understanding of stats.

View on HN

No, that still doesn't make sense. The loss discount was on a per-trip basis, not per hand, so basically if at the end of the week you were down $100k, you only owed $80k.

The rules were generous but NOT totally even. The house still had around 0.25%, which is low, but not unheard of. Even on the Vegas strip you can find Blackjack games open to the general public under 0.50% house edge. Of course, that's with optimum general strategy, not "playing your hunches", which could easily open the house edge to 10% or more depending on how dumb you're playing.

My guess is that he was card counting, but not with a traditional system, and he wasn't using it to size his bets, just to make strategy decisions, which would be harder to detect.

The odds of beating 3 separate casinos, over a reasonable number of hands, even against a very low house edge, are quite small.

I agree, but it appears that while he doesn't "count", He plays a more general "averages" game. Per the article: "More useful, for his purposes, is running a smaller number of hands and paying attention to variation. The way averages work, the larger the sample, the narrower the range of variation. "

Edit: I would guess he's developed some algorithm for keeping track of swings and general variation, so he plays not hand by hand the way counting goes, but rather "batch" or "streak" While will also have wild swings of him making and losing lots of money, but if he keeps tabs on these variations, he can swing them in his favor over time.

Probably the best evidence for your argument is the fact that he apparently negotiated a hand-shuffled shoe. I'm amazed anyone would agree to it, since it's pretty much a giant warning sign that the guy is planning to exploit poor shuffles.

The shuffles needn't be poor. The existence of the hand-shuffled shoe just means he has the opportunity to count or shuffle track. The former is impossible if they use a continuous shuffle machine and I imagine the latter is either impossible or too difficult to use effectively. I agree it makes no sense for the casino to agree to that if that's outside their normal rules. I guess they want high rollers to think they've got an edge, but the casinos realize that most high rollers don't actually know how to count fast enough to realize that edge. And until this guy, they've always been able to identify the successful advantage players (not counting group players a la the MIT blackjack team) before losing millions of dollars...

another factor: as an inside guy, he probably knew their precise algorithm for catching counters. I suspect he does count, but knows exactly how to vary his play to keep just south of the model targeting him as a counter.

What are you talking about? A CEO got fired because of this. Anyone and everyone was watching this guy.

What I'm saying is, they have mathematical models that they follow to decide whether someone is counting. I suspect this guy was in fact using information about cards played to vary his future bets -- in effect counting, but he was doing it in a way that did not fit the house's models for what card counters normally do.

He clearly knew more than the people who were watching him, so the fact that all the 'experts' were paying attention doesn't mean he wasn't following a winning strategy.

This is correct. If the variation is high enough, the per-trip discount starts to approach the value of a per-hand discount, which of course is tremendously favorable. If he's a real whiz with the numbers, he could even have altered his play to slightly less absolute advantage but greater variation, and therefore greater advantage in the model given per-trip discount vs probable # of hands played per trip.

What struck me as very odd was the description of the 800k play: He had two tens and two elevens, and decided to stand. That makes no sense to me, as he would have lost nothing by getting four more cards.

Also, the article made some reference to variation, without explaining the edge. I'm guessing there was a real point lost there.

When you double down, you have to stand, and you can only double down on your first hit. The article makes it seem that he doubled after getting the card, when reality he made that decision before he got the 2's and 3's.

He split his 8's to 4 $100k hands, and then doubled down on each, for 4 $200k hands. He just got really unlucky on each hand, that the point value was so low.

The odds of beating 3 separate casinos, over a reasonable number of hands, even against a very low house edge, are quite small.

No one is writing stories about the millions of other high rollers who lost. It's entirely possible his outcome was driven entirely by chance alone. No card counting, no special expertise other than knowing the odds of the game. Just sheer luck. Unfortunately, that doesn't make for a very interesting read.

It depends on the number of hands he played. As it mentions in the article, the house edge is based on billions of simulated hands. If he played nonstop for 1 week, that is at most 14,000 hands (100 hands/hr x 20 hrs/day x 7 days). Card counting or not, this is a statistically insignificant number of hands.

This article describes someone with a lot of money playing blackjack with bad bankroll management who quit while he was ahead.

The std. deviation of blackjack is quite low. 14,000 hands would be a fairly large sample for blackjack, for live hands. Now if we were talking about something like video poker, then yea, 14k hands would be nothing.

No, you are hugely underestimating variance in blackjack and the amount of hands needed to approach statistical significance.

If you do the math, you should see that after 14,000 hands with 0.25% house edge, the chance of a gain is around 40%.

Given a chance of being up of 40%, the chance of being up (and this is ignoring the magnitude of the win) over 3 separate trials is 6.4%.

"... The odds of beating 3 separate casinos, over a reasonable number of hands, even against a very low house edge, are quite small. ..."

what about the MIT Blackjack team? (I assume you're refering to Blackjack) ~ https://www.youtube.com/watch?v=gXR5sKR3f-E&feature=rela...

They were counters. I meant the odds of beating 3 casinos without counting, which is that the article was claiming.

"... They were counters. I meant the odds of beating 3 casinos without counting ..."

from the article it looks like pre-counting (odds) using software isn't factored or detected.

even better poker skills

He plays blackjack.

Thanks, updated

Thank you for the summary. I hate when articles open with a teaser then bury the explanation.

Then you probably shouldn't click through to anything published by The Atlantic. (Or for that matter, the New Yorker, the New York Times, etc).

This isn't a blog post or a new article. This is a short story. I thought it was interesting and I thought it was well written.

HN has a strong bias against anything artfully written (read: more than the bare minimum of words). I guess I shouldn't be surprised considering the crowd (mostly engineers, myself included) but it is sad to me when people do the 'TL;DR' thing.

> "it is sad to me when people do the 'TL;DR' thing"

HN has a strong bias toward allowing people to allocate their time based on their preferences. The tl;dr convention is tremendously useful because it allows people to make an informed decision as to whether or not to read the article, based on a summary which is (typically) more accurate than the headline.

In particular, if someone thought the article might have something new to say about the math of blackjack, the original comment would have dispelled them of that notion and therefore saved them some aggravation.

On the other hand, if someone was looking for a short story about the personality behind a famous gambler and didn't want to get bogged down in math, your comment may have enticed them to read. Thus, both comments are of value.

The premise of "tl" - "too long" - is a negative. While a convention, it's also dismissive, and that's what makes me sad. It reminds me of Emperor Joseph II telling Mozart there are "too many notes" in The Marriage of Figaro.

I would prefer people said "summary" or "precis" instead of "tl;dr".

In any case, it's only one of many conventions. Another convention is that "The Atlantic", "New Yorker", etc. write high-quality short stories and don't dive into technical details. It's sad that people feel that other readers here are ignorant enough of history to need additional clues about a 150+ year old magazine.

Sadly, I think this is an increasingly recent phenomenon (get off my lawn). It's kind of funny given the style and length of HN favorites like pg and Steve Yegge.

Steve Yegge and pg are more on-topic for HN.

I love the TL;DRs just as much as I love the artfully-written short stories. I find they both serve a useful purpose for me.

Why is it sad? I too am curious about what happened but I'm not interested in reading a long, drawn-out abridged biography of the characters right now. A TL;DR saves me from that pain, and everybody who WANTS to read the full glorious article is more than welcome to. I don't usually find this style of writing to be artful (instead I frequently find it tedious and boring), but neither raldi nor I are asking for the authors to change their style. We both simply appreciate that there's a way to skip the drama and get to the crux of the matter. Thanks OP. (For what it's worth, I really enjoyed "The Eudaemonic Pie", dealing with the same topic of beating the house. That was well written.)

The problem is these articles are low signal to noise.

You get the info on the subjects childhood, house, car, wife/husband, children, upbringing, college, style of dress, appearance, office, eccentricities, pets, etc, etc.

There's a lot of human interest element in these stories.

And of course the are paginated to increase page views for advertising.

Not that there aren't good articles occasionally in the publications you cited, but there is way too much "Sunday paper" in it.

Reply on news.ycombinator.com