Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
40oz

Map length measured in time

Recommended Posts

I was curious if anyone would like to assist me in constructing a formula to calculate the expected completion time of a map given the size, complexity, and combat involved with the map in minutes and seconds.

Obviously, having a player volunteer their honors system FDA demos as the standard to represent a mediocre skill bracket is not very reliable. However I suspect the future of printz' AutoDoom may occupy that role effectively in the future.

The data for all the weapons firing rates, the damage they deal, and the range of amount of shots required to kill each monster is in the wiki already, which I would assume is one of the major steps in conjuring up an equation.

However, the biggest hurtle I believe is coming to an agreement on what formulas best represent the balance of all the different possibilities that come to a number that best resembles the playing time of your average doom player. (it would involve a lot of averages, so the number would have to be blurry unfortunately) I have a variety of suggestions to brainstorm some ideas.

I feel this type of formula might help as a standard rule of thumb for a project designed around 2:00 minute maps, or for Soundblocks idea for a megawad designed to be representative of the default par times that he presented last year that didn't take off

Share this post


Link to post

Now that's something you'll never get anyone to agree on a precise figure, because there are many ways to define "completion time". Could be...

  • ...shortest time to run to exit, as if monsters didn't exist (Pacifist Speed definition).
  • ...shortest time to exit, by killing only as much as necessary standing in your way (UV-fast)
  • ...shortest time to exit after killing every single monster (UV-MAX)
  • ...the above, but at different difficulty levels.
The only aspect which could be somewhat automatized would be calculating the shortest path to exit based on pure pathfinding, including necessary switch and key hunts (BTW, if you can pull _that_ out, you'd revolutionize Doom self-playing bots by quite a bit), which would give you a good approximation of the fastest "good citizen" pacifist run (without employing other tricks or shortcuts).

Considering that most players go for an UV-MAX type of completion for their maps, I'd say that the "average" time will be even slower than the UV-MAX records, maybe by a factor of 2 to 3.

The only way you could make such an analysis is to write some software that tries to find relationships between map size, shape, # of monsters and the known COMPET-N records for that map. That's PhD-level data-mining and pattern-discovery work.

Share this post


Link to post

Doomwiki notes that Romero determined the original par times by speedruns rounded down (to nearest 5s?) and then added 30s.

It's more a process than a formula but the above seems reasonable - although maybe add 10s more for every minute above a 5min speedrun time?

Share this post


Link to post

If I was really hard-pushed to find a simplified formula, I'd define a "resistance factor" as the ratio between the uv-max record time for a map vs the uv-speed one (not necessarily pacifist).

E.g. for E1M1 it would be 0:29/0:09 ~= 3
For E1M2 it would be 01:38/0:22 ~= 4.5
For E1M3 it would be 02:24/00:40 ~= 3.6
etc.

There probably is a correlation between level size (Area? Perimeter?) and times, and a correlation between number of monsters per area unit and resistance factor, which could allow estimating the expected finishing times. For UV runs of E1 it seems that meeting resistance from monsters slows a player 3-5 times, if that's any indication.

Ofc there are pathological cases such as levels with small area but convoluted paths (e.g. a labyrinth) so rather than area, the smallest possible path between start/finish should be taken into consideration.

In any case, lots of statistical research is involved. Make sure you have plenty of free time on your hands.

Share this post


Link to post
traversd said:

Doomwiki notes that Romero determined the original par times by speedruns rounded down (to nearest 5s?) and then added 30s.

It's more a process than a formula but the above seems reasonable - although maybe add 10s more for every minute above a 5min speedrun time?

I think this is probably the smartest way to go about it. Like 40oz said, getting a random play to represent the average run bracket seems unreliable, but the truth is anything beyond that takes effort that frankly isn't worth it.

Share this post


Link to post

Ofc there are pathological cases such as levels with small area but convoluted paths (e.g. a labyrinth) so rather than area, the smallest possible path between start/finish should be taken into consideration.


Indeed I get the feeling navigation plays a bigger part in this than combat. You could have a 1024 map and a Deus Vult sized level sharing an identical monster composition. You could have two 1024 maps with similar fights but vastly different layouts.

To add insult to injury, with layout complexity comes the efficiency of visual cues, both on the mapper side (conveying information) and the player side (interpreting said information). Arguably this is only relevant in first attempts, but most players don't replay maps, so an average time probably should account for that variation.

Perhaps comparing FDA times to UV Max records could offer some insight in that regard. While select FDAs here and there may not be reliable, if you take a whole bunch of them outliers would have less significance.

Share this post


Link to post

This is something that needs empirical data, and either that has to be crowdsourced somehow, or generated by player-emulating bots averaged on thousands of attempts -but none has ever produced a Doom bot which is capable of completing arbitrary levels.

Share this post


Link to post

It'd probably be safer to disregard any speedruns. UV speeds won't work, because you'll have trouble collecting runs that don't use exploits/speedtricks. They "ruin" length calculations and Romero's method didn't even foresee them. Even UV maxes just go too fast if done by skilled enough players. They also completely remove any of that puzzle/maze/non-linearity bonus the map can have, a maxer will just blaze through it following a strictly predefined path.

It's nice to analyze the problem, Maes and phml brought up some good points, but it's nigh impossible to solve, imo. At least without a huge repository of demos that represent all sorts of players playing the same maps, especially average and casual players.

Share this post


Link to post

Probably the only realistic way is for the author and a couple of his buddies to playtest a map and write down the average time. This should work well, UNLESS the mapper is TimeOfDeath or anyone in his crazy rocket-jumping league :-p

Share this post


Link to post

Thanks for the input, guys. I too, was concerned about the quantity of variables, and the complexity of some maps, particularly for puzzles that aren't easily solvable, or doors that don't use door textures, etc. Theres no easy way to account for that.

I like the approach of getting a balance between UV Max and FDA, but I was hoping there might be a way to gauge the time using only the data of the level, without actually having to play the map, and using the average of FDA and UV Max to test it. If the numbers are at least within a minute of each other id consider it a success.

Share this post


Link to post

I'd go by uvmax time myself. There's too many variables that you can't take into account.

Share this post


Link to post

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×