Twenty-Four Tables

When you're playing twenty-four tables at a time, you really need to be in the zone. It's not even action; it's reaction. You've got a HUD — a heads-up display — that shows you what's happened in the hand and what the general tendency is of the other players at the table, colour-coded. You're almost feeling the contours of the decision rather than explicitly deciding anything.

For a while I was starting to get carpal tunnel from the mouse. So I built a little tool that wired up all the actions I needed to take onto an Xbox controller — press one button to raise, press another to fold, the bet sizes precalculated. If I was in a hand I didn't know how to deal with, I'd push the big green Xbox button in the centre of the controller and the hand would get flagged for analysis afterwards. That way I could teach myself what was the right decision, what was the wrong decision, and hopefully know what to do next time.

For the after-game analysis I used tools — some commercial, some I'd built myself — that tried to calculate as accurately as possible the cost of any decision. If you raise here, you'll make twenty dollars on average. If you call, you'll make fifty dollars on average. That sort of thing. The tools would work out the decision trees both for where you were in the hand right now and for what was likely to happen against the playing styles and likely holdings of the other players still involved.

Tiny little rock in the middle of the Irish Sea

At some point, playing poker online in the US was starting to get more difficult, and I was also just getting very bored of it. I was originally going to go back and finish my last semester at uni in the fall of 2011, but I ended up getting a job interview at PokerStars to run their ring-games product line. When they offered me the job, I decided that was going to accelerate me more in my career than any sort of entry-level job I'd get after graduating. So I moved over to a tiny little rock in the middle of the Irish Sea and spent four years running that product line.

The leeches

The thing most frustrating about the table caps was that if you were really good at the muscle memory and at knowing what to do, twenty-four tables just started becoming a comfortable rhythm you could breathe within. You still had space to expand from there. Most of the people sat at twenty-four, though, were painfully slow. There were leeches on the system. They didn't make the game more interesting; it's not like they played particularly well. They just happened to know enough to leech money out of the system, and in the meantime, they slowed down everybody else's games by a substantial amount. It was a real epidemic at the time.

The signal

It was a well-recognised problem inside the company, and fairly easy to get it approved as something to try out. One of the things I liked about working at PokerStars then was that if something didn't require a lot of development work, or that could be done in configuration, senior management was happy to experiment with it, provided the upfront cost was low and the change could be reversed if it didn't work out. We just had to be honest with ourselves about whether or not, once it was implemented, it was the right move — not getting sucked into the sunk-cost fallacy.

Honestly, figuring out the signal was fairly obvious on its face. How long it took you to respond to a decision when it was your turn to act — because that's ultimately the thing that slows down the game and that matters to the player experience. After that, it was a matter of how to get the information. The feature was popular enough internally that it kept getting discussed across different groups on a semi-frequent basis, and at some point we realised the data already existed. It was just in a subsystem that wasn't usually used for player features; it was usually used to detect and prevent cheaters. So we had a fairly novel problem: using data that was meant for one system to affect a completely different one. In the end it was the best way forward, and meant we could ship in weeks instead of months or quarters.

Rollout

Most of the work on the rollout was in the messaging and the internal support documentation. The feature was going to improve quality of life for most of the players in the system, but anyone who had their cap lowered who had previously been playing the maximum was going to be sore. There were complaints, people said it shouldn't apply to them. Most of those escalations came to me as the person in management who owned the product change. People got a bit pissy. But we saw a material change in how quickly they were responding at the tables and how quickly tables moved in general, once it really got underway. Approximately five percent lift in hands played and revenue across the multi-billion-pound product line over the following six months.

The Guinness record

One of the fun things about this is that we ended up setting a Guinness Book of World Records entry for it. At the PokerStars Caribbean Adventure in January 2012, Randy "nanonoko" Lew played 23,493 hands of no-limit hold'em across twenty-five to forty tables in eight hours, ending the session up $7.65 — without eating or leaving the table. He was down $1,200 at one point and locked it in with aces on a final three-way pot. The record was for most hands played in eight hours while maintaining a profit.