PA Mentor shows how your in-game stats compare to other players, based on the data collected by PA Stats. The percentages mean that you stat is currently as good or better than that many of all players at this current point in time with same team size. Detecting the team size should work automatically if you have PA Stats installed. Otherwise, click those +/- to show the stats for teams of that size. You can install this mod from PAMM. Version history, source code etc. are at https://github.com/orfjackal/pa-mentor
I'm trying to offer constructive criticism to help improve the mod. I'm not quite sure how this helps players. Games are so completely unique and matches that last 14 minutes are very different than matches then last 2 hours. So the total unit count varies at different times. And you're right, the number of players is also important. So this would need to track the stats based on the time and the players, as well as the game type. A 4 way FFA is very different than a 2v2. Neat idea though. Just a little too early developed to be overly useful.
I've now fixed some bugs in the percentile calculations. It used to show too low percentages, even 0%. Only the server side has been changed, so you don't need to update the mod. The stats will be fixed in about 2 hours, after the server finishes recalculating the percentiles. (Amazon EC2 Micro instances are not quite meant for CPU intensive operations. My desktop calculates them from the same data in 5 minutes.)
I've now implemented calculating the percentiles based on the number of players in the team. After about 12 hours (EDIT: i.e. now), after the PA Mentor server does its periodic recalculations, you should see the percentiles be much greener. With 0.1.0 version of this mod the percentiles are based on single player teams. Tomorrow I'll release the next version which will let just adjust the team size. I haven't yet figured out a way to find out in-game that how many players the current player's team has, so there are +/- buttons in the UI for adjusting that manually.
Last time I checked it was horrible, PA Stats grabs that data in the lobby before the game and stores it in the local storage. You might want to try that instead.
Yeah, so just require PA Stats and read that same value out of local storage. No need to do extra work!
PA Stats isn't supposed to be a framework and I won't give any guarantees on how stable or unstable is what happens internally. It's just not reasonable to do, since it's version management is not controlled by pamm.
Thanks for the tips. I'll try using the value that PA Stats stores. I can always make it default to 1, if PA Stats is not installed or there is a breaking change.
Oh also there is a bug where sometimes in rare cases it fails and stores bogus data. Results in reporting to PA Stats to fail completely. I have not been able to reproduce it though :/
Version 0.3.1 released. It makes this mod compatible with build 61250. Note that it still shows the stats from PA Stats that are based on the previous build, so the balance changes are not reflected immediately. This situation will gradually resolve itself over the next 2 weeks when PA Mentor forgets all the old stats, but I might improve it to update the analytics sooner.
The OP doesn't explain what the percentages mean. Your stat is currently better than X% of games of the same size at this time stamp?
I'm loving the idea of this mod! but why is "currently better than x% of games at this time stamp" rather than my data / averaged data ? the latter is what I initially expected it to be. I.e. at this time stamp, I have 100 units, whereas the average number of units is 150 units = 66%.
Average is a quite useless number without knowing the distribution. I have plans to draw a graph of the distribution as a function of time, similar to this graph. Just haven't yet had time to do it. Which would you like to have higher priority - such graphs or stats about the unit composition (e.g. whether others have T2 or nukes)?
respectfully, I disagree! I don't mean to come off as complaining or bitching - I am in total awe and appreciation for this - and all! - mods. I just (personally) feel like average would be better. average, by definition, accounts for distribution. currently, I only know that I'm doing better or worse than other games, but I don't know by how much. showing me where I'm at relative to the average of the other games would tell me. and I'd rather having info about unit composition, rather than distribution graphs... but that's just me! again, I want to buffer my opinionated comments with gratitude for your efforts on the mods!!
Average (mean) is useful only with normal distribution and a couple others. But even then the amount of variance has a big influence on how good "50% above average" is. With log-normal distribution and others the difference between mode, median and mean gets skewed, with mean being the most skewed: (When PA Mentor reports your stat as 50%, it means that you are at the median.) That's why I'm planning on making those graphs.
aww man, you're tossing around terms I haven't used in over 10 years! hah. instead of going and refreshing my memory (and giving myself horrible flashbacks of undergrad) i'll just leave it in your evidently capable hands! out of curiosity, is this graph going to be part of the mod's in-game UI?