| Fish Tank Clan :: Forums :: General Forums :: Server: Fish Tank Classic |
|
« Previous topic | Next topic » |
| Bring Back yTB | ||
| This topic is now closed |
| Author | Post | ||
| Saucey |
|
||
Registered Member #830
Joined: Wed Aug 08 2007, 10:30PMPosts: 109 |
I kinda liked YTB just cause it switched things up. like sometimes i wish when its a full game it would switch everyone around after awhile(all T's to CT and CT to T) or some funky monkey junk like that lol (certain people only play T unless forced to play CT) Edited Tue Oct 09 2007, 05:21AM |
||
| Back to top |
|
||
| alcosatz |
|
||
FT Classic Op
Registered Member #39
Joined: Fri Dec 02 2005, 10:30PMPosts: 1545 |
Thanks for the on-topic comments.  Work is progressing well on the team balancer and me being totally effing anal about some details is paying dividends!I know I said I would drop more details several days ago but work plus the occasional bender on the weekend has limited the time I spend on forums. I've still been grinding out code for the plugin, though, and I am by no means giving up on it. As a side note: I'm thinking about reallocating my old server for a public testing place for plugins since FT|GOD doesn't appear to need it anymore for DM. If you guys want to continue discussing some of my hints on page three about what I'm working on I would appreciate it. Here is a little bit more description of the process and factors used by the plugin to evaluate team balance: Player Acquisition/Analysis: - Track individual player KPD performance for the past month (entirely tweakable if a month seems to short or long) - Track individual player KPD in the current session - Determine an "effective" KPD by averaging (actually weighting the average with more importance on session KPD) - Track individual player win/loss count per session - Track individual player autoswaps and manual admin swaps (used for determining who is a good choice for swapping--no one likes being swapped too often or when you're having a crappy night [no wins!]) Team Data Acquisition/Analysis: - Track team wins in the 20 rounds - Track current winstreak - Determine team "strength" based on the above (W20, streak), average player EKPD ('E' is the weighted average of session to historical KPD) - Maintain a balance meter/threshold of team strengths that is decremented over time, which prevents instant team swaps (meaning: one bad round of a few people shuffling around is not going to necessarily cause the plugin to jumble teams) When conditions are right (balance meter drops to zero), then... Auto Team Balance Action: - Determine individual "candidacy" for a swap based on above KPD factors, as well as number of swaps that have occurred for each player (this session), and individual win/loss count - Do some sanity testing on the potential results (will the swap action create a larger imbalance for the benefactor of the swap? if so, try a less aggressive swap) So yeah, all the acquisition stuff is done and very solid at this point. Figuring out reasonable swap "candidacy" logic has been sort of a stumbling block, though. The good news is that I think I'm on the brink of handling that portion and, hopefully, it's smooth sailing after that. Obviously there will be refactoring etc. but once I'm over the most difficult part of finding swap candidates I will hurry up to get a public testing version of this thing out the door. I'm gonna keep working on it--hopefully I'll have something up for evaluation in the next few weeks! 
|
||
| Back to top |
|
||
| BuBBLe GooSe |
|
||
The Original MilfHunter
Registered Member #163
Joined: Mon Mar 06 2006, 02:24PMPosts: 1592 |
That sounds like a considerable amount of tweaking and testing to ensure each function works and works well in cohesion with other functions. It looks like apTB (Alcohol Powered's Team Balancer) should knock the ball out of the park! But will there be any added load on the server from all of this? I mean tracking 30 days of history for each player sounds pretty massive. I suppose SM is more stable and better equipped than ES and Mattie for this stuff, right? Alcohol-Powered wrote ... Team Data Acquisition/Analysis: - Track team wins in the 20 rounds - Track current winstreak - Determine team "strength" based on the above (W20, streak), average player EKPD ('E' is the weighted average of session to historical KPD) Quick question, would tracking 30 rounds instead of 20 make a significant difference? 30 rounds would be closer to an hour's worth of real time and would be a larger sample to draw from. Just a thought.
|
||
| Back to top |
|
||
| alcosatz |
|
||
|
FT Classic Op
Registered Member #39
Joined: Fri Dec 02 2005, 10:30PM Posts: 1545 |
BuBBLe GooSe wrote ... That sounds like a considerable amount of tweaking and testing to ensure each function works and works well in cohesion with other functions. It looks like apTB (Alcohol Powered's Team Balancer) should knock the ball out of the park! But will there be any added load on the server from all of this? I mean tracking 30 days of history for each player sounds pretty massive. I suppose SM is more stable and better equipped than ES and Mattie for this stuff, right? Bots have made testing much easier. It's a lot of information to go through but there are copious amounts of logs to audit and make a corollary of whether the numbers are being crunched correctly. The good news is that all of the "heavy lifting" of calculations happen at round end with a little at round start. Very limited data is gathered on kill events but the impact should be minimal (nearly all stats mods/plugins capture these). The historical data shouldn't be a problem since it is stored in an external database, and the functions to store/retrieve this data are asynchronous (and ONLY happen on player connect/disconnect). This means the server does not "wait," and in turn cause players to wait, for these operations to complete. BuBBLe GooSe wrote ... Quick question, would tracking 30 rounds instead of 20 make a significant difference? 30 rounds would be closer to an hour's worth of real time and would be a larger sample to draw from. Just a thought. I suppose it could make a difference. Players have pretty short memories, though, and the I doubt that many of the same people are playing one hour later. The importance of the past X rounds diminishes greatly if only two or three people are still present. A perfect example of this is the total team win counts displayed on the scoreboard. People look at them and assume because one team has 50 more wins FOR THE ENTIRE DAY that those must have happened in the past 50 rounds that were just played... very frustrating when your team drops 10 rounds straight and people say teams are fair because your team won a crapload of rounds hours before you started playing. :/ Edited Thu Oct 11 2007, 08:48PM |
||
| Back to top |
|
||
Powered by e107 Forum System
|
|