Antithesisoftime
Joined: Aug 20, 2014
|
 Posted:
Jul 10, 2019 - 23:36 |
 
|
| neilwat wrote: | https://fumbbl.com/p/match?id=4075943
Is this supposed to be the result of a really broken system. It does not look that bad to me as a result. You didn't even take an additional apothecary. |
Wait, seriously. This is all over 3 MNGs? |
|
|
ramchop
Joined: Oct 12, 2013
|
No, all this over feeding a troll. When will we learn? |
|
|
awambawamb
Joined: Feb 17, 2008
|
| Posted:
Jul 11, 2019 - 00:52 |
|
| The_Murker wrote: | | (everything) |
this, totally. |
_________________
"la virtù sta nel cielo e nella terra, ma anche nelle nuvole e nelle stelle"
 |
|
Arktoris
Joined: Feb 16, 2004
|
| Posted:
Jul 11, 2019 - 00:58 |
|
how can anyone complain about sharing a bloodbowl match with mrt1212? |
_________________
Hail to Manowar! The latest charioteer to DIE for bloodbowl! - Slain, by Ghor Oggaz |
|
mrt1212
Joined: Feb 26, 2013
|
| Posted:
Jul 11, 2019 - 01:12 |
|
| Arktoris wrote: | | how can anyone complain about sharing a bloodbowl match with mrt1212? |
Weird, right?  |
|
|
MenonaLoco
Joined: Jan 05, 2016
|
| Posted:
Jul 11, 2019 - 03:36 |
|
|
ph0enyx13
Joined: Nov 14, 2015
|
| Posted:
Jul 11, 2019 - 04:14 |
|
I'm mostly a league player, but I have to say, I am trying to build a bigish orc team in box so that I can protect smaller teams from getting these 1000+ tv mismatches. (Granted, I am also trying to just play orcs in general as a way to get better at the game). I think the way box is set up does encourage certain teams over others. I've had humans death spiral and lose so much with them that I generally would rather play them in ranked. Granted I've death spiraled humans in league too and it makes me play teams with big banks like undead/orcs or teams that don't care about player deaths like Amazons. |
|
|
mrt1212
Joined: Feb 26, 2013
|
| Posted:
Jul 11, 2019 - 05:21 |
|
ph0enyx, you're on the right track - I in fact have done similar in the past in the opposite direction - I activate a team in the low 2000s to protect my even larger teams from a matchup that could spell peril and spiral for them. Orcs, dwarves, lizards, chaos. They're good body guards for high elves.
Also, part of the fun journey to be had, if you have any sort of imagination that can rationalize dedicating years of your life to playing this silly game...is the story of how your orcs became the body guards for the coach's stable. Just saying. |
|
|
ArrestedDevelopment
Joined: Sep 14, 2015
|
| Posted:
Jul 11, 2019 - 05:31 |
|
I really wouldn't worry about a one million TV inducements gap when you've never had a team last longer than fifteen games in box. You've got to be a pretty old team first, or 15+ and massive TV to even see 300+ Tv gaps. Not a dig at you Ph0eyx13, just trying to stop a thread like this scaremongering. |
_________________
 |
|
garyt1
Joined: Mar 12, 2011
|
| Posted:
Jul 11, 2019 - 23:30 |
|
I know Christer and admin like to be tolerant but it seems long past time to ban Smallman. He plays a lot of matches but in his absence total matches would probably go up rather than down.
..To the topic it would be interesting what Christer thinks of Nelphine's analysis. |
_________________
“A wise man can learn more from a foolish question than a fool can learn from a wise answer.” |
|
Nelphine
Joined: Apr 01, 2011
|
| Posted:
Jul 12, 2019 - 00:35 |
|
Actually, if Christer is interested in my analysis (I may have squee'ed at the shout out *shifty eyes* *dances*), I'm also interested in the racial suitability modifier. I expect there are at least 2 factors that could influence it:
One which is based on win rates of the given races at the given tvs, as calculated elsewhere on the site, with a suitability modifier that reflects those win rates in an attempt to give even matches. But does this factor account for coach bias, meaning, coaches of certain teams will inherently win more assuming equal skill than coaches of other teams (wood elf vs ogre for instance); and therefore the win rate of a given race at a given tv should be further modified by the win rate of the coaches involved.
(However, this may turn out to be so insignificant as to not matter, but I have no idea that's why I'm interested!!)
But the other major factor is what games will result in longevity for the box, and therefore account for 'fun' matchups - meaning bash, and in particular, clawpomb (I know it doesn't exist anymore, but it's the extreme example) are less 'fun' opponents. Another example is dwarf vs Amazon - it may not be likely to pitch clear, but a team practically designed to counter your team's advantages may be un'fun'.
My guess is the box doesn't use this second factor, but I could see using pick stats from ranked as a basis for it, and I can think of other influences that might be relevant.
Regardless of those factors, I do think that I might just have the wrong gut for racial modifiers, based on my own pixel hugging emotions. And that shows that there are both huge advantages, and disadvantages, to hiding the racial suitability. I'm inclined to agree overall, it shouldn't be public. |
|
|
harvestmouse
Joined: May 13, 2007
|
| Posted:
Jul 12, 2019 - 09:30 |
|
My thought has always been that box would have suited a divisional type division. Kind of like the old faction but more.
Coaches could get badges/out of game rewards for coaching different races to higher divisions. So you get a 'fill the grid' aspect too.
TV really isnt reliable for match making outside of new to newish teams. |
|
|
MenonaLoco
Joined: Jan 05, 2016
|
| Posted:
Jul 13, 2019 - 00:40 |
|
bah, why not make it simple?
Scheduler should Just have the <15 games protection on, and just pair per tv.
No matter what. Abuseable? not much, if admins can modify as they wish. |
|
|
JellyBelly
Joined: Jul 08, 2009
|
| Posted:
Jul 13, 2019 - 00:55 |
|
Just implement the same rule in Box as in Ranked, that you can only be drawn against the same coach 1 in 10 games. It would have practically no effect during peak times, but would boost the US Box hugely. |
_________________
"Opinions are like arseholes, everybody's got them and they all stink." - The protagonist, Fallout 2
"Go for the eyes, Boo! Go for the eyes!!"  |
|
Christer
Joined: Aug 02, 2003
|
Ok, so I never responded to this thread, so let me add my perspective.
First off, before I go off on a tangent, let me respond to the original post.
OP got scheduled with a TV 1260 team vs a TV 2310 opponent. This TV 1260 team was at 106 matches played at this point, meaning low-mid TV is a strategy.
Now, back before the 15 game protection was introduced, this was a viable meta strategy: Staying low to hunt new-ish teams who were yet to build strong players. Now, I believe this is a bit predatory and these games were outright not fun and in my opinion not in the spirit of the game.
Rather than going out and banning people for doing this, or retiring teams, the formula was changed to make these teams much less likely to be paired against new teams in preference of older teams who had had the same opportunity to build their teams.
However, the problem here is that OP (and another notable poster in this thread) thinks it's an inherent right for these teams to be able to do this without risk of getting paired against a higher team. This is clearly not the case, and this match perfectly shows the direct consequence of the 15 match protection zone.
The claim that (and I'm paraphrasing here) "this is clearly the best way to play" was absolutely true before the 15 game protection zone, but categorically changed with the change. You are still free to minmax at low TV, but you have to realize that this means that your team will strongly be preferred by the system to be matched against the older teams which means high TV in the current meta.
As a response to this, I'm certain that some people will claim "but my team was just bashed to oblivion by my opponent and it's unfair that my TV 1400 broken elves has to face a 2000+ killer team". I respectfully disagree with that sentiment in general. Teams do not have an inherent right to survive in the blackbox. Although the probability of getting your team actually killed off is minimal in the current ruleset, it can certainly be hard to rebuild after a streak of a couple of bloody games in a row. That's tough, but is sort of the whole point of the Blackbox division in my mind. Blackbox is about coming up with a team build that can be sustained over time, even in the rough environment that it is. This is to me the challenge of getting a high CR in B as opposed to the opponent selection component that R has. This distinction is important to me.
With that out of the way, let's go off on the tangent about the formula:
First off, responding to all questions about racial factors: They're not in effect anymore. It turns out the race effect was practically always pushing the win probability towards 50% because the races are practically even on a large scale. Thus they were disabled entirely in the last CR update, and while the B scheduler technically still calculates them, there is no effect as the underlying factor has been normalized { Mathematically: p ^ [ 5 ^ (1 - 2 * 0.5) ] = p ^ [ 5 ^ 0 ]= p ^ 1 = p }
Second, Nelphine was asking about a detail about how the scheduler works, or suggesting a change to "fix" the problem. This is unfortunately based on a misconception on how the system works.
The scheduler doesn't pick a specific team for a coach and tries to pair it against everyone else. The scheduler takes every pair of activated coaches and compares *all* of their activated teams to find the best match between these two coaches (in terms of highest suitability). It is these potential matches that are placed in a pool and pairings are made from these.
In the draw in question, you'll see that the chosen pairing shows the 1260 underworld team in the chosen pairing, and the 1410 chaos renegades vs another coach (1510 high elves) in another match that was considered. The system simply picked the match that was scheduled as the best match between these two coaches, which is why it at no point considered the 1410 renegades vs the dwarves.
So the question is, why was this the case?
If you look at the description of the formula (linked from the box page), the most obvious thing you see is a graph for base suitability. You'll notice that the base suitability plummets really quick once you go beyond 50k (normalized) difference. Without actually double-checking this, he game in question is something like 2310 / 1260 * 1000 - 1000, or about 833k normalized difference, which puts it base 500 suitability. The 1410 team would equivalently be 2310 / 1410 * 1000 - 1000 or 638k difference. Again, this is very very close to 500 base suitability.
Now, after this TV difference amplification step is calculated, there's another step (number 5) that applies a small random factor. While this factor is small (2%), it could absolutely overpower the absolutely tiny difference in base suitability.
Is the TV amplification too strong? Maybe.
That leads to another issue though: I am for technical reasons currently unable to make changes to the scheduler. Essentially, my dev environment for the Java EE back-end that runs it isn't in a working state due to changes in versions of software and libraries during my switch from a Mac to a Windows PC a while back.
Now, this is something I am currently (as in I've spent the last few days on it) in the process of resolving (with a complete rewrite of both the R gamefinder and the B scheduler in Nodejs rather than JavaEE). Naturally, this isn't as smooth as I would like. Again, version upgrades of libraries I used before is causing me to rewrite some base platform code I had already functioning. However, as opposed to my unwillingness to spend hours on end on enabling me to compile and deploy the Java EE back-end I have no such problem with the nodejs stuff which feels much leaner to work with in general.
To summarize a way too long post: I think the blackbox scheduler works generally well and I honestly believe it's a matter of certain people simply reviewing their overall minmax strategy and realize it's not the best way to play these days.
Sorry for the wall of text...  |
|
|
|
|
| |
All Star Bowl!
Search
