18 coaches online • Server time: 06:26
Forum Chat
Log in
Recent Forum Topics goto Post Theory-craft Leaguegoto Post On-spot substitution...goto Post Juggernaut as counte...
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
The_Murker



Joined: Jan 30, 2011

Post   Posted: Oct 18, 2017 - 04:13 Reply with quote Back to top

Fame. CR isn't fame. CR is skill. Skill at winning. What ever tactics win the most vs. real competition should gain that coach the most CR.

If there is no real competition, then a coach should not gain or lose CR. So I agree with you.. "the ones that play it safe and only play vs 140s coaches should not get the same CR reward as those who took the "unnecessary" risk to play high CR coaches and won. " I agree. They should get almost no CR at all.

But nor should they lose any.

I also agree with you again.. " coaches who choose to always play it safe with tier 1 teams against tier 2,3 teams should not get the same CR reward as those who took the "unnecessary" risk and played a tier 2 team against tier 1 and won." I agree. Further, any CR gain OR loss when one of the teams involved is a low tier team should be reduced by a factor of 2 or more. If one coach is just 'horsing around' no meaningful CR should change hands.

Tactics, however.. if one coach uses a tactic that wins often, and the other uses a tactic that wins fantastically 'some' of the time, the frequent winner is the more skilled player, imo. It isn't about a reward, or fame. It is suppose to be an indication of which coach is more likly to win an encounter.

_________________
Image
Join the wait-list. Watch the action. Leave the Empire. Come to Bretonnia!
Arktoris



Joined: Feb 16, 2004

Post   Posted: Oct 18, 2017 - 05:41 Reply with quote Back to top

The_Murker wrote:
Tactics, however.. if one coach uses a tactic that wins often, and the other uses a tactic that wins fantastically 'some' of the time, the frequent winner is the more skilled player, imo. It isn't about a reward, or fame. It is suppose to be an indication of which coach is more likly to win an encounter.


The coach more likely to win the encounter is the one that knows how to score more. For once the score is 2-0, the grinder coach is now panicking. For he has no idea what to do now.

_________________
Hail to Manowar! The latest charioteer to DIE for bloodbowl! - Slain, by Ghor Oggaz
Arktoris



Joined: Feb 16, 2004

Post   Posted: Oct 18, 2017 - 06:02 Reply with quote Back to top

The_Murker wrote:
Fame. CR isn't fame. CR is skill. Skill at winning.


I agree.

If you play 10 games vs 140s coaches and win 6 times that's nice.

But if I play 10 games vs 155+ coaches and win 6 times, I've demonstrated higher skill than you and should receive a bigger CR boost.

If you play 10 games with dwarves vs goblins and win 6 times, that's nice.

But if I play 10 games with goblins vs dwarves and win 6 times, I've demonstrated higher skill than you and should receive the bigger CR boost.

If you play 10 games and win 6 times 2-1, that's nice.

But if I play those same 10 games and win 6 times 3-1, I've demonstrated higher skill than you...and should receive the bigger CR boost.

Bigger the challenge...bigger the reward. That's what CR has always been about.

Your skill determines how great of a challenge you are able to successfully handle.

_________________
Hail to Manowar! The latest charioteer to DIE for bloodbowl! - Slain, by Ghor Oggaz


Last edited by Arktoris on Oct 18, 2017 - 10:38; edited 1 time in total
thoralf



Joined: Mar 06, 2008

Post   Posted: Oct 18, 2017 - 06:09 Reply with quote Back to top

Arktoris wrote:
Bigger the challenge...bigger the reward. That's what CR has always been about.


Yet the overall CR is not the division CR is not the racial CR.

Same challenge, different rewards.

_________________
There is always Sneaky Git.
Grod



Joined: Sep 30, 2003

Post   Posted: Oct 18, 2017 - 06:31 Reply with quote Back to top

A comment about the new formula. As I understand, it takes into account 3 possible outcomes in terms of the result

* A win by 2 or more Tds
* A win by exactly 1 Td
* A draw.

The idea I understand is to reward coaches who don't stall and try to achieve extra TDs. But wouldn't it have the exact opposite effect?

A common situation in a game is to be winning by 1 touchdown and in the second half, being in a position to either score straight away, or stall until turn 16 to get to 2-0. I generally would score straight away, confident my worst case scenario is 2-1, with a win being a win that's fine. Under the new formula, if I score straight away to go for 3-0 or higher, there is zero upside to my CR. However I can now lose some potential CR if my opponent does manage to make it 2-1. So now, if I am concerned about CR, I would always stall in such a situation. In fact, once you are sure of being at least 2 touch downs clear, you should absolutely stall in all situations. Or am I reading this wrong?

I understand Christer this is a work in progress. You will never make everyone happy! I always find it interesting developing new algorithms.

You are probably finding that its hard to debug and fix given how long it takes to evaluate the results. Unfortunately there is no way around the trial and error process. However, can I humbly suggest one of two approaches (which you might already be doing):

1. Just do the calculation for say the first (or even last) 100K matches, and see if you are happy with the results before doing the rest. Should take the testing time down to the 10-20 mins range. At least this way you will know if something is obviously wrong before committing to spend most of the day to calculate it.

2. Work on speeding up the calculation.
There is only ~2.3M games played, which is generally a trivial calculation. Even in a scripting language it should only take a few minutes (and in a language like C++ sub-second). My guess is that read/write speed to database is the issue, and you are running the update code in a loop, writing the result each time to database. You will need a different script for testing that pulls in all the matches and stores everything in memory, runs the calculation sequentially and then writes the result at the very end?

_________________
I am so clever that sometimes I don't understand a single word of what I am saying.

Oscar Wilde


Last edited by Grod on Oct 18, 2017 - 09:08; edited 3 times in total
Seventyone



Joined: Dec 02, 2010

Post   Posted: Oct 18, 2017 - 07:24 Reply with quote Back to top

Hi Christer

I know I am a poor coach but i am interested in how you work out CR. I thought that there was some factor that weighted tournament games higher than "normal" box or ranked games. Is this the case?

_________________
[img w=400]https://fumbbl.com/i/493475[/img]
Sp00keh



Joined: Dec 06, 2011

Post   Posted: Oct 18, 2017 - 09:08 Reply with quote Back to top

@happygrue,
Ok if you win most of your games by a slim margin, eg 2-1
But some games you lose, and when you lose you give away huge TDs, you could have a positive win rate, high CR, and (slightly) negative TD differential.

Arguably this is good coaching. You've bagged the wins safely and not risked your team for more TDs that weren't needed
You've given up when it's a loss and not risked your team to prevent TDs that don't matter

I don't see how this is anything to do with minmaxing, generally. Maybe the example guy you found minmaxes but the 2 behaviours don't directly link, necessarily
Sp00keh



Joined: Dec 06, 2011

Post   Posted: Oct 18, 2017 - 09:45 Reply with quote Back to top

I noticed JackassRampant was actually right on page1 and P does affect the relative difference in CR change for minor win Vs major win

Minor win is worth between 90% and 50% the CR change of a Major win, depending if you were extremely unlikely or extremely likely to win the game, respectively

I've updated my posts to show this
Tripleskull



Joined: Oct 12, 2003

Post   Posted: Oct 18, 2017 - 13:06 Reply with quote Back to top

Arktoris wrote:
But if I can consistently beat PeteW 3-0 game after game...


Well guess what. You can't. And your coaching skills are not even the problem here because it can't be done because PeteW would beat you most of the time if you tried. Even if you where an elite coach.
MattDakka



Joined: Oct 09, 2007

Post   Posted: Oct 18, 2017 - 14:21 Reply with quote Back to top

The goal of a BB game is winning by having the most TDs at the end of the match. Winning 10-0 or winning 1-0 doesn't make any difference for the purpose of winning and being successful.
A high TD difference might be related either to higher coach's skill (but not necessarily) or a more entertaining game (but CR is not meant to calculate the level of entertainment) or a noobstomp or a dicing or one turner.
Also, some slow teams like Khemri, Dwarfs, Nurgle, Orcs are not generally able, by design, to score a lot of TDs without risking to tie or lose the game (I won 4-1 with Khemri 2 days ago but it was due to lucky dice and my Legend Thrower).
They need to control the pace of the game to improve their chance to win.
JackassRampant



Joined: Feb 26, 2011

Post   Posted: Oct 18, 2017 - 14:41 Reply with quote Back to top

Sp00keh wrote:
I noticed JackassRampant was actually right on page1 and P does affect the relative difference in CR change for minor win Vs major win

Minor win is worth between 90% and 50% the CR change of a Major win, depending if you were extremely unlikely or extremely likely to win the game, respectively

I've updated my posts to show this
That's actually a good thing. If a lower ranked coach blows out a higher ranked coach, there's probably dice or matchup issues going on. It strikes me though that S should be greater than 0.9 (or less than 0.1) if you want the 2-point win factor to truly be "minor". Like .93/.07 or something.

_________________
Veni, Vidi, Risi
tussock



Joined: May 29, 2011

Post   Posted: Oct 18, 2017 - 14:51 Reply with quote Back to top

Scoring lots absolutely helps you be a successful coach, because your team builds go quicker and give you better teams with less games. Same as rolling a lot of blocking cas helps you be a successful coach, in that your killers end up superstars really quickly and that is also good for winning more games.

You also get to practice playing D a lot, and playing D well, turning the ball over quickly, taking advantage of every slight fault in positioning, having players in position to recover and score quickly such that the other team can't stop you, that makes you a better coach.

Now, it's all somewhat risky, in that you will lose a few more games because of the odd early score that ends up steamrolling your team from the next kickoff, but winning endless numbers of relatively easy games by small margins so your min-max team doesn't grow out of it's comfort zone is sort of the issue at hand, eh.

In that the goal of fumbbl is basically building teams and learning to play them well for the majors, like the one about to be drawn. Very Happy

_________________
ImageImage
Sp00keh



Joined: Dec 06, 2011

Post   Posted: Oct 18, 2017 - 15:39 Reply with quote Back to top

@Tussock "the goal of fumbbl.. [tournaments]" maybe, but that's offtopic

@JackassRampant yea I'm undecided. I think a-win-is-a-win... But also it penalises grindy cherry pickers
happygrue



Joined: Oct 15, 2010

Post   Posted: Oct 18, 2017 - 15:45 Reply with quote Back to top

Sp00keh wrote:
@happygrue,
Ok if you win most of your games by a slim margin, eg 2-1
But some games you lose, and when you lose you give away huge TDs, you could have a positive win rate, high CR, and (slightly) negative TD differential.

Arguably this is good coaching. You've bagged the wins safely and not risked your team for more TDs that weren't needed
You've given up when it's a loss and not risked your team to prevent TDs that don't matter

I don't see how this is anything to do with minmaxing, generally. Maybe the example guy you found minmaxes but the 2 behaviours don't directly link, necessarily


Fair enough on both points. Plenty that TD differential might *suggest* but It's an assumption based on other assumptions. My thinking goes that coaches who win a lot tend to play in ways that prevent the other coach from scoring. Aggregate that over a lot of games and it seems like a significant thing to measure. But without diving into the data to sort that out, it's just my opinion. Very Happy

_________________
Come join us in #metabox, the Discord channel for HLP, ARR, and E.L.F. in the box!
Image
happygrue



Joined: Oct 15, 2010

Post   Posted: Oct 18, 2017 - 16:14 Reply with quote Back to top

@Christer: I'm seeing something strange with the ranking development graphs. I notice that some of my individual race ranking are quite high, while other race rankings with very similar records are suspiciously low. When I check out the ranking development graphs, I'm seeing some of the race CR graphs "look accurate" in that the endpoint looks like where I am now. Others (specifically the surprising ones) are ending at a very different point than my current rating. Is this a fluke due to work in progress or is something else going on?

_________________
Come join us in #metabox, the Discord channel for HLP, ARR, and E.L.F. in the box!
Image
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic