Statistics
Author
Feedback is essential for making good complete polished games. Unfortunately, good feedback is hard to get. The best beta-testers are generally veteran gamers, so their point of view is somewhat flawed. Casual gamers may not give very good reports, but are generally the target market for a game, and their first contact with a game is decisive (at least for shareware games).
Making a game that can record all the player''s actions could help. It shouldn''t be very hard to code either. Just record all the input events in the main game loop (keyboard, mouse, network messages, etc.), then run the game in a different game loop that reads the events from the log file. This makes things a lot easier for the beta-testers: all they need to do is play the game and send the log file.
Another useful tool could be automatic reports generated by the game, resuming game actions in a number of easy to read statistics. So, what would make good interesting statistics?
- average number of mouse clicks per action
- frequency of different actions
- usage ratio of different game items (objects bought / units build / time spent in different game areas / quests accepted / ...)
- player power growth rate
- learning curve - after how many games the "player power growth rate" goes to a maximum
How can these statistics be put to good use?
- putting most used UI elements in more accessible places
- balancing game items
- real-time finding of game cheats / exploits in MMO games: if a player''s power grows too fast, raise an alarm.
In what form can game statistics be best displayed?
"The best beta-testers are veteran gamers...their views are flawed"? What do u mean exactly? If they''re the best beta-testers then how do u mean their views are flawed?
Author
They''re the best because they know what is worth mentioning in a report. Their views are somewhat flawed because they are probably good gamers, and their point of view differs from the casual gamers point of view. And since most independent games must aim at the casual gamer...
quote:Original post by Diodor
...The best beta-testers are generally veteran gamers, so their point of view is somewhat flawed. ...
I agree, it''s the same as some game reviewers, they''ve just played way too many games sometimes and their view gets a little skewed.
Back to the post, I''m not sure, but I''d try to put in everything you can think of, and just hope that with all the cs degrees in the team, one of them took some statistics courses when fulfilling the math requirements.
(I''m in one now and I enjoy it.)
I agree with you now that i know what u mean 
. Why would you need to know the average mouse clicks? The others i can see, especially the learning curve. I like games that have somewhat of a learning curve but some are just too extensive.
Those aren''t bugs...they''re added features :-P
. Why would you need to know the average mouse clicks? The others i can see, especially the learning curve. I like games that have somewhat of a learning curve but some are just too extensive. Those aren''t bugs...they''re added features :-P
The problem with just gathering stats when you are testing your game out is that this is telling you only a very minor point about the game. It is telling you what people are doing but not the why. And if you knew why someone was clicking in the bottom corner 50 times even though there was no response you probably wouldn''t have put the big red X there in the first place.
The best place to find out what casual players are going to do when they play your game is to get all the casual gamers you know to play your game while you watch so you can ask questions either during play or after. Trust me we all know people who are casual gamers or not gamers at all. Your parents are a good start, people you work with, other people in school. If you approach these people and say "Hey I''m working on this game and wanted to know what you think about it." Most people would be more then happy to spend 30 minutes playing it for you in a non threatening environment.
The best place to find out what casual players are going to do when they play your game is to get all the casual gamers you know to play your game while you watch so you can ask questions either during play or after. Trust me we all know people who are casual gamers or not gamers at all. Your parents are a good start, people you work with, other people in school. If you approach these people and say "Hey I''m working on this game and wanted to know what you think about it." Most people would be more then happy to spend 30 minutes playing it for you in a non threatening environment.
I don´t see how this is new or groundbreaking. Detailled monitoring is part of almost every games test phase. Beta testers are usually employed to find bugs, playtesting is a different thing entirely.
What you monitor and how you display it will rely heavily on the game, what you should focus on are stats which represent player behavior, such as (assuming the game you had in mind when writing your post was a RPG) the players position/location, frequency of visits, duration of visits, idle times, unsuccessful actions (including clicks, which will telly you a lot about the initial learning curve).
The output of the data will vary depending on the type of data. Simple tables will to for most, graphical representations for others (if you have designed a labyrinth for instance you would want to know where, when and to what density dead ends are encountered and which parts are most frequented).
Another thing which is more difficult to monitor: frustration points (either through direct player feedback or indirect), which tell you when and where the players get fed up with the game (log the exit or load points for example).
there´s much more, but since it varies from game to game it wouldn´t make much sense going into detail here.
What you monitor and how you display it will rely heavily on the game, what you should focus on are stats which represent player behavior, such as (assuming the game you had in mind when writing your post was a RPG) the players position/location, frequency of visits, duration of visits, idle times, unsuccessful actions (including clicks, which will telly you a lot about the initial learning curve).
The output of the data will vary depending on the type of data. Simple tables will to for most, graphical representations for others (if you have designed a labyrinth for instance you would want to know where, when and to what density dead ends are encountered and which parts are most frequented).
Another thing which is more difficult to monitor: frustration points (either through direct player feedback or indirect), which tell you when and where the players get fed up with the game (log the exit or load points for example).
there´s much more, but since it varies from game to game it wouldn´t make much sense going into detail here.
quote:Original post by NicolasX
Why would you need to know the average mouse clicks?
Too many mouse clicks may indicate that the game is too complex. Although to be honest, I can''t see why you''d measure ''clicks per action'' as the number of clicks to accomplish a certain task is likely to be almost identical each time, isn''t it? Maybe clicks per minute or per second is more useful. Or the number of clicks in a row, indicating ''periods of activity''.
quote:Original post by MorganE
The problem with just gathering stats when you are testing your game out is that this is telling you only a very minor point about the game. It is telling you what people are doing but not the why. And if you knew why someone was clicking in the bottom corner 50 times even though there was no response you probably wouldn''t have put the big red X there in the first place.
Statistics are cheap and easy to compile and measure. You could get data from hundreds or even thousands of people using some sort of automatic stat-collection method. On the other hand, you only have time to interview maybe 5 to 20 people personally and ask them about the game. Not to mention the fact that you get a subjective view when you talk to them face to face, which is not always accurate. Both methods are important and should be used where possible. The statistical data can be useful for narrowing down potential areas to ask individual players about.
quote:Original post by Diodor
So, what would make good interesting statistics?
Obviously it will vary from game to game. You might want to measure the time spent in certain interface modes. For example, how long does the player spend in the inventory compared to time spent walking around? What proportion of the playing time is spent placing buildings rather than ordering attacks? There are some simpler stats too: win/lose percentages come to mind. Game session duration. Option choice (eg. game speed). etc.
[ MSVC Fixes | STL | SDL | Game AI | Sockets | C++ Faq Lite | Boost | Asking Questions | Organising code files | My stuff ]
Author
quote:
Original post by Hase
I don´t see how this is new or groundbreaking.
As opposed to every other thread
Not groundbreaking, but worth talking about. quote:
Another thing which is more difficult to monitor: frustration points (either through direct player feedback or indirect), which tell you when and where the players get fed up with the game (log the exit or load points for example).
How about putting a designer's picture in the corner of the game screen? Clicking on that image would hit the designer with a stick, making it very easy to express frustration
quote:
Original post by Kylotan
Although to be honest, I can't see why you'd measure 'clicks per action' as the number of clicks to accomplish a certain task is likely to be almost identical each time, isn't it?
My sister was playing Sims and had a very hard time locating certain house items. Knowing how many clicks it takes to buy each kind of item could reveal problems with the user interface.
[edited by - Diodor on September 4, 2002 4:32:32 PM]
Remember that when gathering sample data for statistical analysis if you sample is bad then the conclusions that you make from that sample data are bad. Basically it does no good to collect tons of information if it isn''t the right type of information. Maybe if you increase the information collected for each sample you can be better assured that you sample is good. For instance, don''t just collect the number of mouse clicks. Collect the number, frequency between clicks, position of clicks. Collect as much data about the clicks as possible. When you go to use the sample data you can determine which of the variables are relevant. Of course if you collect to few variables your conclusions may be misleading.
Cheers,
RandomTask
Cheers,
RandomTask
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
