Page 16 of 26
Re: EvenMatchOmni
Posted: Wed Mar 30, 2022 8:04 am
by pooty
We decided and implemented the lower PPH balance. And joins ALWAYS goes by number of players first and foremost. So the team that's lower numerically is going to get the new player.
This accounts for a stronger team having slow start, and it also accounts for other joins... Say blue is losing, then they get a strong player taking it from 4 blue v 5 red to 5 v 5, at that point its PPH balance, otherwise, blue could get two players, potentially strong players and easily flip/ruin the balance.
I could see joining the losing team, if numbers were equal and the match is in the latter stages but that's likely not possible.
Re: EvenMatchOmni
Posted: Wed Mar 30, 2022 8:11 am
by captainsnarf
pooty wrote: ↑Wed Mar 30, 2022 8:04 am
I could see joining the losing team, if numbers were equal and the match is in the latter stages but that's likely not possible.
That shouldn't be too hard to do actually. The old code is still there commented out for determining who is losing based on game score/nodes.
Something like
Code: Select all
if(Level.TimeSeconds < 60*20) // 20 minutes
{
// use PPH based joining
}
else
{
// use Game score based joining
}
Re: EvenMatchOmni
Posted: Wed Mar 30, 2022 11:27 am
by pooty
Hmm. Cool.
So maybe there's an "in between" option.
Check PPH difference/threshold
Check time/game scoring (is this match scoring or is it like nodes controlled etc.)
Figure which has the bigger difference and use that to determine where they go...
Say Blue (2000), Red (1500) - 500 pph is big threshold I'd say throw to RED
But say Blue (2000), Red (1900) - mostly a tie, then put them on the "losing" team.
I think we already have a threshold value (450 is current setting IIRC) we could use. As a side I think that 450 might be too high still... most maps a good above average player I think is around 300, so 450 could be a "player and half". That value right now is only used for a shuffle check.
Re: EvenMatchOmni
Posted: Wed Mar 30, 2022 11:52 am
by Enyo
Well, we've had a lot of matches lately where the team with the lower PPH balance score wins. So, it wouldn't make sense in those cases to put a new join on the winning team just because it has a lower PPH. The new join should go to the losing team.
The logic should go something like this...
1. If a team is short a player, they get the new join
2. If player counts are even, the team with lower PPH gets the new join, UNLESS...
3. If player counts are even and one team is "losing" (whatever that threshold is), then the losing team gets the join.
Re: EvenMatchOmni
Posted: Wed Mar 30, 2022 1:18 pm
by captainsnarf
The reasoning behind the current way is to not punish people for winning and strictly balance teams by PPH only. Teams are balanced, just not playing up to potential.
We can change it and there will be more close games. Not sure what is fair though. If I scramble to get to the center node in MassD and the other team doesn't, I shouldn't then be punished by the other team getting the new join. I worked hard to get the center first.
Re: EvenMatchOmni
Posted: Thu Mar 31, 2022 3:05 pm
by Enyo
captainsnarf wrote: ↑Wed Mar 30, 2022 1:18 pm
The reasoning behind the current way is to not punish people for winning and strictly balance teams by PPH only. Teams are balanced, just not playing up to potential.
We can change it and there will be more close games. Not sure what is fair though. If I scramble to get to the center node in MassD and the other team doesn't, I shouldn't then be punished by the other team getting the new join. I worked hard to get the center first.
Good point... you can't balance behavior. Therefore, isn't balance in ONS impossible, or at the least misguided? How is balance defined anyway, wouldn't true balance ultimately lead to frequent stalemates (is that fun?), like Dr. Scrotey once pointed out? Maybe we should listen to a guy that has a PhD in physics with a particular focus on quantum computing.
Re: EvenMatchOmni
Posted: Thu Mar 31, 2022 6:09 pm
by pooty
Maybe we should listen to a guy that has a PhD in physics with a particular focus on quantum computing.
Yep. We don't need to over-engineer it. Between the considerable original work from Wormbo, and Snarf's improvements...I'd say its probably about a good as its going to be.... and sometimes its just fun to beat the crap out of the other team.
Re: EvenMatchOmni
Posted: Thu Mar 31, 2022 8:18 pm
by YEAAAHHHHHHHHHH
Yeah really what’s the goal here? Nail-biters every time? I think it’s good how it is with minor adjustments as found fitting. It was fun to suddenly wreck the other team on Red Planet the other night.
Re: EvenMatchOmni
Posted: Fri Apr 01, 2022 8:03 am
by pooty
We did have a case, I think it was Minus?, where SK10 joined and we believed he got put on the "stronger" team. Snarf, can we log the joins with pre and post balance numbers to make sure its working? I am not 100% he went to the stronger team since there were several matches last night where PPH wasn't the whole story, as the high PPH team got wrecked.
Re: EvenMatchOmni
Posted: Fri Apr 01, 2022 10:08 am
by captainsnarf
sure I can add some logging. Also we still have some non-mulligan mulligans. looking into that also