logging doom stats (to file/real-time)

Hi, I am helping out with a sleep project that is studying the influence of dreams in skill acquisition.

The experiment goes like this: participant plays doom for 30mins, then sleeps, is awoken by experimenter after a REM episode and describes his/her dream. next day he/she plays doom for 30 minutes again.

The question is: is there a correlation between the inclusion of doom-related elements in dreams AND skill improvement?

Up until now, the experimenter has recorded the games as video clips and has gone through them to analyze the participant's performance, scoring them by hand, which is an extremely tedious process. What I suggested is to somehow gain access to a log created by Doom's engine so that we can automatically assess performance that way.

Is there such a log? If so, can one choose what to record on the log? If so, can one record, for example, trajectories? and times of kills?

Thanks

ernesto

Well, the game lets you record demos in "lump" format which the engine can then replay, like the ones you see when it starts up. These in turn can be analyzed down to the Player's exact actions by tools such as LMPC (or here is the author's website for the utility).

nice! and how do I record it in lump format?
thanks for the help!

OK, I had a look at the LMP utility and it looks super. I did not find, though, information about kills, either of monsters or of the player. Kills are essential because they're a good measure of the player's skill. Another question: how can I reconstruct the full trajectory of the player after reading the LMP file? I guess I'll have to have a look at the map and then, for each tick, determine whether there was a wall in front blocking each of the player's intended movements, no?

Yeah, the data only tells what the player is doing and is them placed in correlation with the level. PrBoom+ may be of help as well, as it has a "smart totals" feature which shows the number of kills as the demo is in progress, as well as a camera mode where you can step out of the player's view while the demo is playing, and other demo related features.

portugueezer said:
and how do I record it in lump format?

Assuming you use the DOS executables, in the command line type the following:

Doom2 -warp 02 -skill 4 -record mydemo

That's in DOOM II, in DOOM:

Doom -warp 1 2 -skill 4 -record mydemo

That will record a demo called mydemo.lmp taking place in level 2 (or episode 1, level 2) in Ultra-Violence. The DOOM FAQ lists all the normally available command line parameters. To use PrBoom+ for this, the commands are similar, though you'll need to tell it to behave like the corresponding normal engine (and game) with "complevel".

prboom-plus -warp 02 -skill 4 -record mydemo -complevel 2

prboom-plus -warp 1 2 -skill 4 -record mydemo -complevel 3

See also its usage.txt for some notes on additional features.

To get full information on a player's skill you'd really need to know such things as shots that miss vs shots that connect, total amount of damage taken, etc. The only way to get such stats would be to get a programmer to create a modified source port for you, as none of the existing ones actually keep track of that sort of stuff. This would allow you to save such statistics in whatever format you desired and analyze them automatically with tools.

Going by KILLS/ITEMS/SECRETS/TIME alone is a rather coarse picture of Doom playing skills.

Well suppose the participant hasn't played Doom before in his/her life? Actually getting to the exit could be quite a hurdle for some people depending on there eye-hand co-ordination, reflexes, how they remember things etc. etc.

I don't know what kind of controls, population sample, etc. they are using the experiment so there's not much use speculating about that. I just find it interesting.

This sleep project is in its early stages. The idea is to run comparisons between performance pre- and post-sleep and to correlate this probable improvement with the ammount of dream about the game. There are many ways you can characterize game performance, from detalied information regarding fine motor skills (could do this by analyzing trajectories and see how much they deviate from straight lines, reasoning that novice players would 'wobble' instead of using the shortest trajectory between two points, or count the number of times that the player bumped into a wall) to, for example, knowledge about the map (by determining which parts of the map were visited).

I like the idea of tayloring your levels to force players to use particular cognitive strategies. Up until now this hasn't been done, but it's a good idea.