User Pass
Home Sign Up Contact Log In
Forum > Test Server Discussion > Where are the tests for the new SAs?
Page:
 
wormser1971
no title
offline
Link
 
Originally posted by Catch22
All my builds will be open as soon as I have time to set them.


This guy drops the ball !!! http://goallineblitz.com/game/player.pl?player_id=1939012 (24 last season)
And his name is Digital Bort.... that's LOL funny!!!! Good thing you guys have a good sense of humor!!!
 
Catch22
offline
Link
 
Originally posted by wormser1971
This guy drops the ball !!! http://goallineblitz.com/game/player.pl?player_id=1939012 (24 last season)
And his name is Digital Bort.... that's LOL funny!!!! Good thing you guys have a good sense of humor!!!


lol he has 12.8 in catching, I would hope he would drop it.
 
Staz
offline
Link
 
Originally posted by Tony Dungy

Thank you for *finally* admitting on behalf of Bort........what we have all known since the beginning of this game - that testers have had an unfair advantage.



Originally posted by r8
1) He said that's why we don't do those tests.

2) Testers haven't been around "since the beginning of this game."

http://koti.mbnet.fi/tibe9mm/Troll_fail.jpg


 
boomer82
offline
Link
 
Originally posted by Catch22
I'd guess season 16 tbh. Still have a lot more to test and then have to make adjustments and re-test.


So, as far as the old SAs go, none of them will be tweaked until all of them will be?
 
Staz
offline
Link
 
Originally posted by boomer82
So, as far as the old SAs go, none of them will be tweaked until all of them will be?


I'm hoping that's the case. Would skew a lot of things if something like Monster Hit was given a buff while Cover Up remained underpowered (for example purposes only. I don't know if either is flawed).
 
britdevine316
offline
Link
 
Originally posted by Staz
I'm hoping that's the case. Would skew a lot of things if something like Monster Hit was given a buff while Cover Up remained underpowered (for example purposes only. I don't know if either is flawed).


LOL, like anyone cares what cover up does...its just a stupid stepping stone....even if it never made me fumble by having 10 in it, no one would pump it up
 
Staz
offline
Link
 
Originally posted by britdevine316
LOL, like anyone cares what cover up does...its just a stupid stepping stone....even if it never made me fumble by having 10 in it, no one would pump it up


But, if Monster Hit became powerful enough to force more fumbles, and you were able to stop your HB from fumbling twice a game by pushing it to 10, I think you'd give it more consideration.
 
Octowned
offline
Link
 
Originally posted by Catch22
I ran regression testing on the SA's - 20 game base set with no SA's and then would run 5 game sets with varying levels of the SA's. I made changes/recommendations to the SA's to Bort based on these test results.

FWIW, I'm doing the same thing right now with all the old SA's - it takes a while, especially since I have a lot of other things I'm doing. I'm about 3/4 of the way through the offense SA's.

Test results will not be made public though, games are hidden from viewing, not even testers can see them.


20 game test with a limited set of builds. The Statistics Masters in me says

Granted, I'm assuming the test builds are high levels, so I'll assume a few will work for my team
 
tragula
title
offline
Link
 
Originally posted by Octowned
20 game test with a limited set of builds. The Statistics Masters in me says


You need to multiply by the numbers of plays per game. I am not sure how the test is set and what type of probabilities they try to measure. I expect that the average user should be able to "feel" the difference of a SA going from 0 to 10 by comparing two games. If this is so, I would say that it is not so .
 
Catch22
offline
Link
 
Originally posted by Octowned
20 game test with a limited set of builds. The Statistics Masters in me says

Granted, I'm assuming the test builds are high levels, so I'll assume a few will work for my team


I'd run more if the process was simpler but its not. Basically I have to enter two team ID's for a test game and then wait for the test game to complete (usually 15-20 minutes) before I can start another one. So a 20 game test run takes about 8 hours and that's if I am constantly starting new ones which isn't always the case since I am often doing other things at the same time.
 
Octowned
offline
Link
 
trag - for sure, it's not a size 20 sample size. But in 20 games, how many types of builds could you really be testing? This game is way too complex for 20 games.

catch - totally understand, just trying to support my idea of an automated testing process at least a tester understands what regression is, though I am curious what you are using as your data points!
 
slashxtreme
Lead Mod
offline
Link
 
Originally posted by Octowned
20 game test with a limited set of builds. The Statistics Masters in me says

Granted, I'm assuming the test builds are high levels, so I'll assume a few will work for my team


Is it possible to run a batch file? To where you could automate the process by having a macro (or a batch executeable) input the info for you?

20 games isn't a huge problem, it's basically the N of iterations. you need to just make sure you have the appropriate amount of cell values on levels of the SA to make sure each cell has 25+ examples.
Edited by slashxtreme on Mar 25, 2010 08:53:36
 
Hikariu
offline
Link
 
Originally posted by Catch22
I'd run more if the process was simpler but its not. Basically I have to enter two team ID's for a test game and then wait for the test game to complete (usually 15-20 minutes) before I can start another one. So a 20 game test run takes about 8 hours and that's if I am constantly starting new ones which isn't always the case since I am often doing other things at the same time.


Catch, you can schedule two at a time Just make the home id the away id in the second game and bort's process will Q up two separate sims.
 
rjssob
HOOD
offline
Link
 
Is the latest version of the test server fully upgraded to be a mirror image of what is in the production SIM today? I know this is pretty obvious, but if these tests are being run against old code, you really aren't accomplishing anything.
 
Staz
offline
Link
 
Originally posted by rjssob
Is the latest version of the test server fully upgraded to be a mirror image of what is in the production SIM today? I know this is pretty obvious, but if these tests are being run against old code, you really aren't accomplishing anything.


The way I'm seeing it is that the test server is exactly the same as the live server with the exception of proposed changes are obviously uploaded there, and tested, first. If it's loaded to the live server, I'm sure it's still on the TS.
 
Page:
 


You are not logged in. Please log in if you want to post a reply.