gamelab history

2018-10-24: v1.1.10

Fix e-mail notifications for administrative purposes. These would internally error out due to changing function call return values in kcgi from many versions ago.
2018-03-17: v1.1.9

Fix playing games on Android's browser, which had issues with the Notification API. Fix canonical URL of project. Fix buttons on Chrome to look better.
2018-01-25: v1.1.8

Maintenance release: bring up to date with newest dependencies and simplify Makefile a bit. Fix two bugs: a double-free when re-sending errored e-mails to players and a superfluous free when given bad payoff values during game configuration.
2016-07-24: v1.1.7

Allow for negatively-valued lottery values as now documented in the manual. Link to a detailed map of the database schema (again, courtesy of sqliteconvert). Also allow experimenters to specify a set of free-form question and answer pairs in lieu of the default pre-play questionnaire.
2016-07-01: v1.1.6

Since Mechanical Turk players and regular players are not compatible regarding the lottery, do not include them when computing the lottery. This allows Mechanical Turk players and lottery players to mix. While there, update the database schema documentation.
2016-06-10: v1.1.5

Have the assignment time of Mechanical Turk players be set by the player's number of rounds, not the total. Considerably clean up the experimenter interface internals, though there should be no (or very little, beyond polish) visible difference to experimenters. Also focus on using experimenter instead of administrator. This is a stable release of the system: it has been used for several production games.
2016-03-29: v1.1.4

Fix showing the pre-history on the play panel (not just the history panel) on the first round of play.
2016-03-01: v1.1.3

Fix the QualificationRequirement portion of the Mechanical Turk request and make qualifications be part of the configuration process. Add the facility to automatically submit bonuses to Mechanical Turk. This button is revealed after completion of the experiment if and only if the experiment is associated with AWS credentials. (Only Mechanical Turk players are notified of their bonus, of course.) Fix the Finish notification to Mechanical Turk to be through the gamelab system, not the client's browser: this behaviour is usually disallowed as a browser security option.

Also fix showing the pre-history on the initial round of play.
2016-02-23: v1.1.2

Integrate with a GitHub account as a read-only repository. Also run a Coverity scan against the software. This found no meaningful bugs, but is a useful addition to regression analysis.
2015-11-20: v1.1.1

Mechanical Turk submission has been fully integrated into the system, warranting a minor-number bump. To use the Mechanical Turk facility, you'll need only your AWS credentials. The rest happens from the experimenter console when creating the experiment. The manual and quickstart guide have also been looked over and checked that they match with the current version of operation. While doing this, un-break the distributed archive by including some forgotten but necessary files.
2015-11-16: v1.0.24

Now that the session identifiers are properly embedded, remove support for the Mechanical Turk survey link functionality, which was just a crutch. Moreover, move the Mechanical Turk configuration into the experiment parameters and accept the requested HIT identifier as well as whether the request is to be sandboxed. Also make sure that all Mechanical Turk variables (worker, assignment, and HIT identifiers) are proper uppercase alphanumeric strings.
2015-11-16: v1.0.25

Finally have HIT be automatically created given its name, description, your AWS access and secret key credentials, and the number of workers. Deprecate currency (AWS uses only USD). Allow for both sandbox and production mode. Add a quickstart guide and the version history. Finally, make the lottery amount be a dynamic field when starting the experiment. If unspecified, a lottery is not computed—much simpler than needing to specify the lottery status and manually editing the instructions file!
2015-11-09: v1.0.23

Have the entire system (except the administrative console) automatically handle embedded session identification in the query string. This allows external program Mechanical Turk workers to participate without needing cookies.

Clean up the login pages for smartphone and tablet-sized devices.
2015-10-13: v1.0.22

Allow the administrator to control whether games and actions (rows) should be randomised prior to being shown to participants. Fix where Safari would allow the enter key to submit row weights, where on all other browsers it was specifically disabled. Use the Notifications API to note changes to the current round. Switch to using long-lasting cookies instead of session cookies (fewer superfluous logouts). Add administrator flag to show relative rounds (instead of absolute rounds) for players. Cleaned up the captive portal to make it easier for participants to join that way.
2015-10-13: v1.0.21

Address a number of small issues encountered in live trials: join players after answering the last question instead of when they actually try to play the first round, add administrator notice of running the mailer process, disallow negative accumulated tickets (in the event of negative payoffs) from crashing the system. (This last part will be later fixed even more.) Make the administrative console even clearer. Remove the end variable from the experiment table (unused) and properly reset (upon wiping) the total variable as 0, not 0/1. Fix a presentation bug in showing hypothetical payoffs where the payoffs were fractional and the payoff would show zero instead of the fractional result. Bring the manual up to date with the current system, partially documenting the JSON API in the process.

Add the ability for the administrator to disable history altogether. This prevents the history of other players' plays from being transmitted, effectively making games history-free. This is useful for testing the difference in play between information sets.

Also hard-code the Font Awesome at compilation time instead of within source files, allowing administrators to deploy their own icon server instead for LAN environments.
2015-09-28: v1.0.20

Add ability for e-mailing players when the round rolls, or when the first round begins. This involves a long-running child process. (Note: Apache2 has a feature where it waits for all descendents of its CGI process, so if you use this web server, you might see a zombie'd process if you use this feature.) Also add the ability for the experimenter to prematurely end experiments by advancing to the last round.
2015-09-15: v1.0.19

Allow for Mechanical Turk players to log in using a survey link instead of the traditional external question. This is because gamelab requires cookies, and when logged in through Mechanical Turk, these cookies are third-party since they are fetched secondary to the Mechanical Turk interface. Some browsers will ignore these by default, such as Safari, putting an extra burden on players to allow cookies. In the future, we forsee allowing cookie values to be specified in the query string for Mechanical Turk players.

Clean up some typos in the questions.

Significantly clean up the JavaScript front-end bits after running over them with jshint. Also fix some timer issues (timeouts not firing) raised during Mechanical Turk tests. Lastly, allow malformed decimal values like .5 instead of 0.5. After some user feedback, add a loud notification that the round has rolled.
2015-09-10: v1.0.18

Accomodate for older browsers by appropriate setting element event handlers (e.g., onclick) directly instead of as attributes. While there, add better reload-timeouts for players when still in the pre-play phase.

Complete Mechanical Turk integration by recording the assignment identifier and stipulating that Mechanical Turk players report back to the Amazon site conformant with their rules. Hard-code whether we're in the sandbox or not. Eventually this will be relegated to a button when Mechanical Turk is baked more into the system.

Fix a bug wherein players who would have played through the end of the experiment (e.g., added in round 1 with a maximum of 10 rounds, but 10 rounds in total) would accidentally be shown a final round.
2015-09-08: v1.0.17

Allow Mechanical Turk players to log back in with the same HTI identifier. Moreover, don't unset mturk status (on the experimenter side) when experiments begin: mturk players can only log in after the experiment commences anyway, so this is useless. Show experimenters a message when mturk or captive players are enabled consisting of the necessary links.

Put graphs of current game into Play tab for easier reference. These are formatted to display all at once, but be smaller and stacked.

If an experiment has a grace time, show that grace time instead of the total round time for players. When it expires, use the usual Next round text. This involved actually enabling the minimum-round value, which was being ignored by the server! Next, have the browser check the server, once all games have been played, for round rolling.
2015-08-25: v1.0.16

Squash a bug that caused games to abort after 64 rounds. This was due to some legacy weighted-average calculations that have long since been deprecated, but are still computed. These facilities have now been removed. This was quite an invasive change, so pay careful attention to any payoff irregularities.

Have the administrator page check if it should re-load when viewing round-percent experiments. This keeps the administrator page a bit more up-to-date with running games.

Finalise the Mechanical Turk instructions, which required adding a conversion rate for points to currency. Finalise the Mechanical Turk preview page as well.
2015-08-25: v1.0.15

Continue polishing questions so that they have a success post-question phase, the game-choice boxes are properly checked, and some minor formatting issues managed. Remove the history matrix, as this information is already captured in the graphs. A significant bug has been ferreted out and fixed related to counting players for round advancement. This was found by the bots, which now can selectively enable compression to reduce load.

Add a new view for administrators: the projected number of participants over the rounds. Also lift the restriction that administrator e-mail information must be set: if the e-mail is not set, emails are not sent. This is reflected in the administrator instructions.
2015-08-24: v1.0.14

Flesh out the questionnaire, which is an optional sequence wherein players must correctly answer a series of questions before playing. Also record the number of attempts (and starting time) of each questionnaire instance.
2015-08-18: v1.0.13

Split out the option to have Mechanical Turk players from that of ordinary captive players. While working on administration, finally clean up instructions. As it is now, experimenters can select from several defaults (lottery, no lottery, and Mechanical Turk) or upload custom instructions. These can still be changed in the usual way.

This also marks the first no-wait database implementation: connections will now wait perpetually in order to be serviced instead of timing out. This is part of a broader set of actions to reduce database contention for large (multi-hundred) player games.
2015-08-13: v1.0.12

Begin adding support for Mechanical Turk. Experiments employing mturk use the same process as before to start a captive experiment. Unlike in normal captive experiments, however, the application makes available a special URL for submitting as an mturk HIT. This URL acts like the captive portal except both registers and logs in participants, immediately pushing them through to the lobby (or directly into the game, if so configured). The fact that players are mturk-registered is logged and displayed when reviewing player status.

While here, clean up the lottery-winning (and non-lottery-winning) interface for players. Just polish: nothing exceptional.
2015-08-11: v1.0.11

When logging in (either via JSON, such as with captive players, or via HTML as when using the form page or initial e-mail), have the system detect whether the lobby should be used or the home page. This prevents a needless and at times confusing redirect after hitting the homepage.

Lots of aesthetic clean-up for the experimenter: more straight-forward status, cleaner buttons to push, a status graph to see how many participants have played, an a printout of the current top earners (by tickets, not payoff). Also clean up the presentation of earnings for players: one style for lottery-less play, another for playing with the lottery. This involved some internal simplification of how lotteries are computed.

Stipulate a valid user e-mail address only for participants added over the experimenter console. Otherwise, identifiers can be free-form. This allows captive players to use non-email-based identifiers.

Lastly, significantly improve the bot playing utility to handle multiple games and randomised (instead of consistent) strategies.
2015-08-08: v1.0.10

Add the ability for players to have a brief questionnaire when they begin. With this feature, players are put into the lobby before joining the game. They are then required to answer some pre-made questions. When the answer with success, they'll join the game itself. This feature is still experimental: the questions have not been written.
2015-07-31: v1.0.9

Continue to add features to the bot player to discover bottlenecks in gamelab. Document the JSON API in the manual. Simplify the experimenter interface. Fix round advancement in the event of zero players (a hole where players joind in a burst, then finished playing while rounds remain, then new players join). Allow the experimenter to upload a prior game-play history to supplement the current experiment's displayed history. In using this feature, players are shown the prior experiment's game-play as if it were to have come before the current round. Allow the lottery drawing to be suppressed—this is useful when playing games where payouts depend upon the number of tickets, not a lottery drawn from them.
2015-07-25: v1.0.8

Add the ability to test out game deployments with an automated bot player utility. This utility simulates participants who connect, register, and play through a simulation. The play is maximally disruptive (extremely high-frequency attempts) to establish limits of gamelab installations. The bot player will be used as a proof-of-concept for fully programmatic play, wherein participants can program a bot to play on their behalf. This feature will be documented and described in later releases of gamelab.

In doing so, fix a very subtle bug in how round advancement occurs with percent-based play. Also, change the database to use write-ahead logging, which reduces the amount of lock contention in heavily-loaded experiments.

Smaller changes: fixed up the manual with more specific language; simplified the experimenter's Experiment Parameters panel; and fixed the experimenter instructions (their template values we're being filled in). The Save Experiment feature has been temporarily disabled.
2015-07-08: v1.0.7

Introduce ability for players joining after the experiment has started. To do so, experimenters must specify a per-player maximum number of rounds as well as a per-experiment maximum. Players are added by the experimenter in the usual way. When the new players join after the experiment has begun, they wait in a lobby until a slot is available. Once a slot is available (the participant's new role is dynamically assigned as the role containing fewer players), the participant is automatically assigned to the next round.

Remove the unused rank field from the player table.

Perform an audit of using sqlite3_column_int instead of sqlite3_column_int64, and make sure that all return values of int64_t are properly checked and cast.
2015-06-13: v1.0.6

Add database field for checking that the experimenter password and e-mail have been re-set. Change icons for the new experiment page to make it easier to understand what has been done and not done. Add more documentation regarding the use of Font Awesome icons. Add (conditionally-shown) link to captive player homepage on administrator player planel. Add a minimum amount of time for percent-based play: a minimum of time must pass before the round percentage is checked, giving slow players a chance to get their plays in before auto-progression.
2015-05-14: v1.0.5

Allow rounds to advance when a given percentage of players per role have played, defaulting to the maximum time of not enough players play in that time. This allows rounds to advance much faster if players play in bursts. Add ability for administrator to manually advance rounds as well.
2015-05-01: v1.0.4

Renamed auto-added players as captive. Remove logout button for captive players. Remove remember whether to see instructions for captive players. Add cache-control mechanism for player JSON that depends on the current round, relieving the gamelab server of producing the full experiment JSON request with each reload. Added several useful bits of information to the database: client user-agent to each session (and sessions are now retained instead of deleted), creation time for each game-play, and link from game-play to active session (for source correlation). Merged winning instruction text [back] into template itself.
2015-04-30: v1.0.3

Consolidate round-advancement into a single location, opening the door for more advanced round progression. Restructure the player page with additional graphs, replacing all tabular information.
2015-04-22: v1.0.2

Added an auto-add mode where, upon being set by the administrator, participants can register themselves for games. This makes it easier to deploy gamelab in a setting (e.g., laboratory, classroom) where participants' e-mail addresses are not known or where the e-mail verification phase would take too much time. The administrative console has also been overhauled and now allows for saving experiments before deployment. HTTPS access is fully enabled for all components.
2015-03-25: v1.0.1

Second [non-public] release of the system. The push behind this release is to reduce the load on the Gamelab server: a lot of HTML and JavaScript which previously had been dynamically served is now statically served. The paths in these files (which were previously filled-in by the server) are now hard-coded during compilation. Also added some informative bits to the login pages regarding browser usage and the login URI to the player e-mails.