Client

Tournament Reporter (TR) is a start-up company providing web-service to poker players based on their proprietary Poker Tournament Analysis Model. TR's tournament data analysis helps each player find out his/her own skill score and ranking among all the players, to help the player decide his/her tournament tactics and develop skills. To build up their image as the leader in the business, TR must launch their novel service website as soon as possible to attract customers, and demonstrate the vitality of their business model to investors.

Challenge

To complete the Poker Tournament Analysis Model, software developers are faced with the following challenges:

1. The skill score of a player is given on the basis of his/her tournament record, including maximum gains and losses, dates of each gain and loss, total gains and losses in recent tournaments, etc.
2. The importance of a tournament record in skill score evaluation decreases with time in a non-linear fashion, i.e. the earlier the gain/loss happened, the less important the record is.
3. The result of a new tournament will not only affects the skill score of all the game participants, but may also affect the ranking of other players.

In order to provide service to poker players using this analytical model, TR's E-business website must have the following functionalities:

• Mass number of daily updates: There are about 2,000 online poker tournaments every day, each with 50-3000 players participating, which makes the total number of poker players as high as 100,000. A huge amount of tournament data must be downloaded from the 10 biggest poker sites. The amount of records to update is up to 6,000,000, the number of calculated attributes for each participating player is about 500, which makes up to 350 million attributes to be updated daily.
• Problem identification: Sometimes, the data may be duplicates and should be eliminated automatically by the system. The system should be able to identify other problems with raw data and inform a human operator to fix each of them.
• Querying service: A registered player is allowed to query his/her own skill score and skill rank at any time. A query could be like "how do I compare with the Top 5% ranking players", or "what is the average skill score of the upper 50% players".

Core Algorithm:

1. Large number of daily calculations: Although only a few dozens of thousands of players participate in poker tournaments in a single day, all the rankings will be affected. Everyday, the system has to process tournament results and update scores for every player on file, before generating new rankings for each player, and all the necessary calculation must be completed before next day morning. Large volume of entries, coupled with the non-linear model, posts great challenge to the algorithm design.
2. Peak time congestions: During peak time, the system may receive thousands of inquires, each requesting a response with more than 2,000 attributes. Timely response to these inquiries is critical to the system's performance.
3. Maintenance issues: In addition to keeping the records of all the poker players, the system must maintain a log file to record daily calculations, system resource usage, computing time, performance benchmark, etc. for future references.

Online services: TR hopes to introduce the skill appraisal and ranking service on its own website to attract poker players and generate enough web-traffic. Therefore, the application has to appeal to the target online audience.

1. Value Proposition: TR needs to demonstrate the value of the ranking service to poker players. For instance, a player might be able to improve his/her games by knowing the comparative ranking and performance evaluations.
2. User-friendliness: The web-based application must be easy to use with a steep learning curve. In addition to a simple interface, the website also needs step-to-step instructions to guide users through the process.
3. 24x7 Availability: The website must be updated every day, and provide 24 x 7 services. This requires automatic backup of the database and quick recovery of common malfunctions. Quick and effective technical support is also essential.

The Solution

Among the outsourcing vendors, ExperExchange proposed the best solution and won the project to develop the system. The proposal includes the Active Indexing & Group Indexing Algorithm to handle the large volume of calculations.

1. Active Indexing & Group Indexing:

   Based on the following two important observations, ExperExchange proposed the Active Index & Group Indexing Scheme for the system to complete core calculation:

   1. Active Index: At any time, only a small fraction of the players would participate in tournaments and/or make online inquiries, compared to the 1.7 million player population. We assign an Active Index figure to each player based on his/her tournament and inquiry pattern, and assign calculation priority based on the figure. The higher the Active Index for a player, the more often he/she plays therefore requiring more frequent recalculation and update. Queries from players with a low Active Index can be calculated as-needed without degrading the system response. By prioritizing calculation resources and minimizing unnecessary updates, we achieved fast response without committing too much computing resources.
   2. Skill Score Group Index: Any single tournament will not affect most players' score or ranking, and the attributes will change over time in a non-liner fashion. We group the players based on their current rankings, with the same Group Index figure assigned to all the players in the same group. Only an extremely small portion of the players at the extremes of a group would need to be moved to the neighboring group as a result of new tournaments in a single day. Hence, the amount of as-needed calculations and updates is further reduced.

   Our simulation showed that, our algorithm reduced calculations to 1-2‰, and only 0.1-0.2% of the players would move to a different group in any single day. It suffices to re-group only once a week to guarantee ranking accuracy. By reducing the number of calculations, the Active Index & Group Index Algorithm also successfully removed system bottleneck at peak time.

2. Server Cluster:

   Although the Active Indexing & Group Indexing Algorithm helped alleviate system workload, the daily calculation was still too heavy for an ordinary server to handle. We used a server cluster with three servers, with a program dispatching tasks among these servers. The program also monitors the computing process and re-dispatches tasks if one server fails. The system structure also allows more servers to be added when necessary.
   
3. Website Usability:

   The following principles were observed in the web design (see Fig. 1 & Fig. 2, www.tournamentreporter.com):

   1. Content of the site must be concise and to the point, and include all key information to address most customer requests within one mouse click.
   2. Service items are shown in flash files, to demonstrate the service and guide users through the system.
   3. Display skill scores and rankings in histograms, curves and intuitive graphs to minimize the amount of text.
   4. Any tracking, report request or player switches can be completed within 3 mouse clicks.
   
   

   

   Figure 1
   
   

   

   Figure 2
4. Daily maintenance of the website:

   To help Tournament Reporter operate and maintain the E-business website, ExperExchange designed an automated business flow to reduce human hours. In addition, we set up an Offshore Operation Center in China for Tournament Reporter, with operators standing by to support the system. They would monitor daily data collections, inspect the automatic calculation of skill scores and rankings, manage data updates, check the reports on system status, handle unexpected system breakdowns and malfunctions. By using an offshore center, we further reduced TR's maintenance cost by 50%.

   
   

   

   Figure 3. Tournament Reporter Business Flow Chart

Major Benefits

We design a great site, using all the tools available. Photoshop, tablets, HTML, Illustrator, caffeine, dreamwaver, AjAx, - we use all the latest and most established technology to create a web site that is visually appealing and yet fully functional.

Technology