Thursday, June 12, 2014

Lee Jones and the Switch of Doom

During my tenure at PokerStars, the four questions I was asked most often were:
  1. Is there a cashout curse?
  2. Is the software random?
  3. Can you see my cards?
  4. Is there a doom switch?
Of course, these questions generally didn't take exactly these forms - they were generally expressed as facts ("Yeah, I'm running really bad on RiverStars right now because I just cashed out, but that will get better in a week or so.").

If you've been reading this blog for a while, you know that I have a great deal of respect and affection for PokerStars, even though I left the company more than seven years ago. But I suspect that you also realize that I shoot pretty straight. And since, as they say in Texas, I don't have a dog in this hunt, you should assume that the following is a mostly unbiased assessment.

Is there a cashout curse?
I definitely understand why players have come to believe this. The assumption behind the notion of a cashout curse is that online sites want to discourage players from cashing out. Using this type of indirect negative feedback could have nothing but disastrous results for us.

The two possible outcomes if we did rig the software in this way are (1) players wouldn't realize that they were being discouraged, in which case they would just decide that they were running bad, the software was rigged or (rarely) that they couldn't beat the game, or (2) players would realize that there was a 'curse,' in which case the more diligent players would start tracking it seriously, giving them evidence they could provide to the community at large.

My experience is that recreational players tend to play considerably looser once they have money in their accounts, particularly if it got there suddenly (like a big tournament win). As an example, a guy who plays small stakes cash and tournaments, who is used to having $100-300 in his account, plays the $10 rebuy tournament and wins $5,000. Most don't cash out right away; they may play bigger, or play more games, or play looser, or a combination of all of those things. Then he decides to cash out $4,000, but he doesn't revert to his prior style of play immediately. If he's a typical recreational player, he has a negative expectation, and that -EV is amplified by the fact that he's effectively playing bigger.

Verdict: No cashout curse.

Is the software random?
We struggled with this one from day one, as all online poker sites have. I've heard hundreds of stories purporting to be definitive proof that [fill in site name here]'s random number generator (RNG) isn't random, or that boards are somehow being rigged.

The usual reason I hear for the alleged lack of randomness is rake churn. It's better for [site] to have all players win roughly the same percentage of the time. The result is that all of the money circulates, but it generates more rake as it's circulating. So it's in [site]'s best interest to make sure that all players win and lose about the same amount, so the only negative liquidity is rake. The reasoning is wrong but sound, which makes the argument considerably harder to refute.

Here's the big problem with this (also with most of the other alleged means of rigging games): either (1) the software would all have to be written by Isai Scheinberg himself, (2) PokerStars would have to pay its programmers $10 million/year to keep the secret or (3) they'd have to have them all killed immediately after the code was finished. This is the problem with any conspiracy theory: it's hard to keep a secret. I can't say it any better than this:
"Three can keep a secret, if two of them are dead."
                                                              -Benjamin Franklin


One of my favorite stories related to this comes from one of my shifts in support. All PokerStars employees were required to take at least two hours of support shifts per week, a brilliant idea that really opened my eyes. I cherry-picked the questions I answered, and always looked for the ones with topics like "Got my AA cracked for the 20th time." We heard things like this so often that the programmers added a tool to the back-end system, allowing a support person to enter any hand and see the player's last n results with that hand. For example, I could specify that I wanted to see a player's last 20 hands where he started with AA.

The email, which I saved, said:
Dear Fuckstars, 
I just lost my 23rd straight hand with AA and you can stick your fucking rigged piece of shit site up your cheating fucking assholes. I also lost 12 straight flips and 6 hands in a row with AK vs A and some turd. I don't know how the fuck you guys stay in business, but you can suck my dick.
While I wasn't going to act on his suggestion, I did decide to take this particular support request. Here are his results on the preceding 23 hands in which he had AA:

Won: 18
Lost: 4
Tied: 1

That put his win rate with AA at 78%, which is pretty damn good. Oddly, his 18 wins were consecutive - meaning that he had just lost four times in a row (the tie was the last hand). That explained the tone - those four losses became amplified, and in fairness to him, the last one was excruciatingly ugly (he flopped a set vs. TT, but the board ran out four flush cards). The tie was similar - AA vs. JJ, the board ran out five cards of a suit neither had.

I examined each hand to see if there was anything specific I should mention, and I did find one fun hand in a $1-2 NLHE cash game: he had three-bet preflop and got called with two black aces. His opponent check-raised him on a flop of 678 of hearts, he reraised and the guy jammed almost $300 into a pot of less than $100. Our AA player called and found he was looking at Kh Qh, making him about a 32:1 dog. But the board ran out 8s 8c and he won; of course, he didn't mention that hand in his email. I never checked, but I suspect that the holder of the Kh Qh probably sent us a different excoriating email.

I sent him a very detailed analysis of his last 23 AA hands. Just for kicks, I also ran his last 23 KK hands. Not a whole lot there, except that he was up against AA in two of them, and managed to win them both, once flopping quads and once flopping a flush draw that came in. I mentioned this to one of our support supervisors, who said, "You'll never hear from him again, and he'll keep playing." She was right on both accounts.

On a related matter, on more than one occasion, we submitted large numbers of hands to a third party for analysis as to randomness. We're talking tens of millions of hands here, and in each case, the result was the same: there was no evidence of the hands being anything other than entirely random. And contrary to Sheldon Adelson's contention to the contrary, random number generators are able to randomize a deck of cards far better than a live dealer.

Verdict: Yes, the software is random.

Can you see my cards?
This is a tough one, and I was tempted not to even approach it. I can only answer this for PokerStars, and the answer is no. At Isai's explicit instruction, the programmers made certain that no one could see cards while hands were in progress. The tools the support team used relied on what was committed to the database, and hands weren't committed to the database until the hand was completed. To be completely honest, though, I've seen programmers do some stunning stuff over the years, so it would be disingenuous for me to say it's impossible.

The counterexample to this is UltimateBet, which was involved in the most well-publicized cheating scandal in the short history of online poker. As the story goes, their programmers created some 'superuser' accounts in the development stage so they could observe hands playing out in real time. These accounts were disabled but never fully removed from the system, and at least one of them was used by various nefarious types in the clumsiest cheating effort imaginable. So while I can speak to how PokerStars dealt with this, there's no question that there is at least some risk, particularly at sites with less to lose than PokerStars.

Verdict: On PokerStars, no one can see your cards.

Is there a doom switch?
Every time I heard this, I imagined Lee Jones in a giant, Willy Wonka-type construct, with a huge green switch, wearing a headset and awaiting instructions about who to hose. This is, in reality, the same question as "Is there a cashout curse," although it's a little more personal. Poker players need to find a reason why they're losing. The players who blame the dealer when they play live are the players who blame the software when they play online. No, Victoria, there's no doom switch, although I think Lee in a velvet coat with a giant lever in his hand is an awe-worthy image.

                                 That's Lee with the hat.

Verdict: No, there's no doom switch.

1 comment:

  1. The no doubt excellent picture of Lee with his doomhat is broken.

    ReplyDelete