Thursday, November 15, 2012

Design Portfolio - Robocap (Dynamic Table Limits)

Robocap (Dynamic Table Limits)
Q1 2010

In a live poker game, you can usually only play at one table at any given time for reasons that should be evident. However, online there is no such restriction. You can play two, four, eight or even twenty-four tables at the same time if you wish. In fact it is fairly common for players on sites like PokerStars to 'multi-table' - on average, 10% of active unique players are playing 2 or more tables at any given time. This has obvious benefits for both the operator and the player (as long as the player is a winner).

However, early on in my time at PokerStars it became obvious that multi-tabling would be a driving force behind many of the usability improvements and new features that I would design, because of one unavoidable fact - multi-tabling is hard! The 10% of players who are multi-tablers slow down the games for the 90% of players who play only one table, making the game less fun. This is not to say that every person who multi-tables significantly slows down the game

The severity of the problem was highlighted by a comparison that I performed between PokerStars (which limited players to a maximum of 24 tables by default) and it's nearest rival Full Tilt Poker (which limited players to a maximum of 8 tables by default, with this increased to 16 on request). For standard No Limit Hold'em full-ring games, Full Tilt Poker dealt 33% more hands per hour, despite the fact that more players saw the flop on average (meaning that more events occurred per hand). In fact, a 'normal' speed table on Full Tilt Poker dealt slightly more hands per hour than a 'fast' speed table on PokerStars. This was a startling statistic which needed to be explained.

Some of the discrepancy could be explained by differences in software. For example, Full Tilt Poker allowed players less time to act, and had faster card-dealing animation. Accordingly, during my time at PokerStars I designed a number of changes to the software to speed up the games (and these were continued by my successors after I left). But clearly differences in software alone could not account for such a major variation in game speed - nor could they explain the many complaints that we would receive from recreational players who wanted to introduce games where multi-tablers weren't allowed, because of slow play.

The answer was rooted in the Peter Principle. Multi-tablers typically push themselves too far, increasing the number of tables they are playing until they are playing more than they could really handle. Once somebody feels comfortable at, say 8 tables, it is very difficult to convince them that they should play fewer. Often they won't realise that they are slowing the game down, and even if they do, human selfishness means that they are unlikely to give up the possibility of extra profit or rewards and cut back on their tables to improve the experience of others. On the other hand, there were a few multi-tablers who were able to play 24 tables without causing delays, and were capable of playing even more, but were limited to a maximum of 24.

The problem was that the existing 24-table limit was fixed in stone, and bore no relation whatsoever to how capable a multi-tabler a person actually was. People are different - one player might struggle to play 4 tables without slowing the game, while another can play 24 tables easily without causing a significant delay. The only real solution, therefore, is to have a table cap that is different for each player, and which takes into account how capable the player is of multi-tabling without causing delays.

It turns out that the average time it takes for a player to respond (to take an action when it is their turn) is about 4.5 seconds. Clearly, if somebody can maintain that speed or better while playing at 24 tables, they should be allowed to play more. However, if somebody is exceeding that time by a significant margin on a regular basis, chances are that they are playing more tables than they can really handle and their table cap should be reduced so that they don't make the game an unpleasant experience for their opponents.

Because nothing like this had been done before in online poker, I designed a configurable rules-based system. Administrators could define a rule like 'If a player's action time is greater than 11000 milliseconds, and they are playing at 10 or more tables, and their existing cap is 10 or higher, then decrease their table cap by 2'. This allowed the company to roll out the system gradually and silently so that they could see the effect that the changes would have in advance of them actually being implemented.

Dynamic Table Limits were rolled out at PokerStars after I left the company in July 2010, and at the time of writing it appears that they are being used quite conservatively (understandable, since some players are not going to like having their table limits reduced). However I expect the system to have a major positive impact on the quality and speed of ring games online in the coming years.

More Information:

Back to Index 

No comments: