Dedimania forum

okilju_o · 2013-10-23 00:13:45

Hello everyone who is still browsing this forum !

I was wondering if there was somewhere (maybe recorded by the dedimania) an world average time of all players PBs on 1 map. I guess there isn't, but could it be technically possible ?

The world average time would be a very useful piece of information. For example after some statistical analysis, it would allow you:

- to determine how hard a map is
- to locate your pb in the mass
- to rate your pb objectively

without having to look at all the records. However, for that, you need the widest possible sample (if not the whole pack) and server stats are not visible, and often not representative enough.

*** beware, some maths follow ***

You may think calculating Dedi 1 - your PB is a good way to rate your PB, then of course it depends of the map length. But even (Dedi1-PB)/Dedi1 isn't satisfying. The difficulty can vary in huge proportions and if you add the map length and popularity as parameters, the result doesn't mean anything.

Example: you score a 10'20" on an easy map with dedi1 at 10'10". That gives you a 0,99% difference. Very close.
Now you score a 57'50" on a very well known hard map with dedi1 at 56'50". That gives you a 1,77% difference.

Even if the 57'50" was really hard to achieve, you would still need to make a 57'06" to have the 0,99% difference
Popularity (you can put that in numbers) and difficulty (you cannot put that in numbers) make it impossible to find an equation for how good your PB is.

With the average time + the number of PBs + the dedi 1-30, assuming the distribution of scores is a Gaussian, you can locate yourself like: "30% of all records are better than me, and 70% are not as good". Of course, it is an estimation and you would need to check if the distribution is really a Gaussian but that seems likely.

*** no more maths ****

If you think something/someone else should record that kind of info, please let me know Nadeo maybe, for example ?
Feel free to comment, ask if anything is unclear...

slig · 2014-03-24 00:44:28

The problem is that such thing is meaningful only if enough tops on a map are really competitive ones, ie if enough good players have drove a good time on the map. And because it's probably not the case for the big majority of maps, an average of the rankings on all maps a player have times on will be meaningless.

Perhaps that it would become a little meaningful if considered only on maps which have been played a lot, but again it would be biased because probably a lot of the most played maps will be short or very short maps, which are not the most representative of the level of players... But again, for each map that a player have played only few rounds his average ranking would be artificially wrong.

Globally, i don't see what solution would give meaningful results, so wasting resources on this seems useless.

okilju_o · 2014-06-02 02:15:38

Actually, a world average time is meaningful no matter how good the times are. If the map isn't popular, the fact that few good times have been made is balanced by the fact that few very bad times have been made also, because the number of times is low. If the map is popular, very good times have been made, but also plenty of bad times.

To understand better, you can picture yourself a two-dices test. You can calculate the average and the standard deviation with only a sample. And then if you take these 2 variables as fixed, you can predict the frequency of the numbers.

I need to document myself on how racing statistics work, the fact that you can improve your own time changes some things.
However, I think it would be really great to be able to guess, at any moment in the life of the map, how frequently bad/average/good/insane records will be made and their corresponding times. That, knowing only the first 30 and the average of all of them.

Lastly, I don't know if it is technically feasible to update that world average when a player beats his PB without actually recording all the PBs (which is precisely what I try to avoid), so... you tell me Mathematically speaking, you only need the number of PBs, the previous and the new PB during 1 operation and then you can erase them.

Dedimania forum

#1 2013-10-23 00:13:45

Average times

#2 2014-03-24 00:44:28

Re: Average times

#3 2014-06-02 02:15:38

Re: Average times

Board footer