Hitonagashi
Joined: Apr 09, 2006
|
 Posted:
Dec 15, 2011 - 17:39 |
 
|
| VoodooMike wrote: |
Indeed, if that is NOT the case then it vindicates their decision to take the risk to score and improve their players, since the only reason to play defensively to prevent potential injuries is you worry they will negatively impact future games.
|
Not true.
Each time I score, I concede the initiative back to the opposing coach. Each time, I give him a chance to control my players (by forcing me to attack his cage to score), and to inflict damage, potentially winning by more. While I have the ball, I have the tempo and the initiative, and he is responding to me.
It is more certain to win 2-0 than to win 3-0, and you can minimise the effect of luck far more effectively. Indeed, in the box am, improving your team is completely irrelevant. You are matched by TV, so if I do or do not score, I'll still get an even match next game, hence, why even give him a chance at getting back into the game? |
|
|
JanMattys
Joined: Feb 29, 2004
|
| Posted:
Dec 15, 2011 - 17:43 |
|
| Hitonagashi wrote: | | VoodooMike wrote: |
Indeed, if that is NOT the case then it vindicates their decision to take the risk to score and improve their players, since the only reason to play defensively to prevent potential injuries is you worry they will negatively impact future games.
|
Not true.
Each time I score, I concede the initiative back to the opposing coach. Each time, I give him a chance to control my players (by forcing me to attack his cage to score), and to inflict damage, potentially winning by more. While I have the ball, I have the tempo and the initiative, and he is responding to me.
It is more certain to win 2-0 than to win 3-0, and you can minimise the effect of luck far more effectively. Indeed, in the box am, improving your team is completely irrelevant. You are matched by TV, so if I do or do not score, I'll still get an even match next game, hence, why even give him a chance at getting back into the game? |
| JanMattys wrote: | The key word, here, is "control". Often, scoring doesn't give you control over the game. It just gives you a significant advantage, but it also gives the initiative back to your opponent. As weird as it sounds, scoring isn't something you do "mindlessly" in BloodBowl.
Blocking, ball possession, and positioning, on the other hand, DO give you control over the game and its momentum.
In my (limited) experience, the better coaches aren't the ones who outscore the hell out of you... the better coaches are the ones who make you feel like they are in complete control and you're stuck in their net. Even if they are only leading 1-0 and camping near your endzone for 14 turns straight. |
Sorry for quoting myself, and thanks to Hitonagashi for putting it out much better than I did. |
_________________
 |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 15, 2011 - 18:00 |
|
| Hitonagashi wrote: | | It is more certain to win 2-0 than to win 3-0, and you can minimise the effect of luck far more effectively. Indeed, in the box am, improving your team is completely irrelevant. You are matched by TV, so if I do or do not score, I'll still get an even match next game, hence, why even give him a chance at getting back into the game? |
And a better coach than you could do that and more, could they not? The objection I'm seeing is that a coach that wins 2-0 is potentially as skilled as a coach that wins 3-0, but takes more risks. I can concede that that is potentially true!
However, taking risks means letting luck into the mix. That's fine if we're talking about a single match, but across multiple matches the coach that takes more risks will be more subject to luck's effect, and that will be reflected in larger fluctuations in rating which will affect their overall rating as a coach, while the conservative coach will not have those fluctuations.
This goes back to what we said about RandomOracle - he has a large number of games where he consistently performs better than the average and because of that his rating is high and is less likely to be affected by a bad game.
Rating systems may generate numbers on a per-match basis, but you have to put that into context - ratings are aggregated information across a large number of matches, and any inconsistent performance aspects will come out in the wash because of that. If you consistently win 2-0, then your rating will be stable and likely higher than the guy who has some 3-0 wins, a bunch of draws, and a bunch of losses. If some guy consistently wins 3-0, then yeah, I'm still convinced he's better than you are, all this talk of strategic conservatism notwithstanding. |
|
|
easilyamused
Joined: Jun 06, 2008
|
| Posted:
Dec 15, 2011 - 18:09 |
|
I have no B data......
How disappointing  |
_________________
 |
|
dode74
Joined: Aug 14, 2009
|
| Posted:
Dec 15, 2011 - 18:31 |
|
| Hitonagashi wrote: | | VoodooMike wrote: |
Indeed, if that is NOT the case then it vindicates their decision to take the risk to score and improve their players, since the only reason to play defensively to prevent potential injuries is you worry they will negatively impact future games.
|
Not true.
Each time I score, I concede the initiative back to the opposing coach. Each time, I give him a chance to control my players (by forcing me to attack his cage to score), and to inflict damage, potentially winning by more. While I have the ball, I have the tempo and the initiative, and he is responding to me.
It is more certain to win 2-0 than to win 3-0, and you can minimise the effect of luck far more effectively. Indeed, in the box am, improving your team is completely irrelevant. You are matched by TV, so if I do or do not score, I'll still get an even match next game, hence, why even give him a chance at getting back into the game? |
While your point is taken in general, it's this specific case we're talking of: http://fumbbl.com/index.php?name=PNphpBB2&file=viewtopic&p=490873#490873 |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 15, 2011 - 18:49 |
|
| easilyamused wrote: | I have no B data......
How disappointing Razz |
You probably made a typo, or had spaces before or after your name. It worked fine when I typed your name in. Or hey, here's a direct link to your rating. |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 16, 2011 - 14:21 |
|
Ok, so I've updated the script to take TV into account when calculating the rating. Ratings are broken down into up to 4 groupings: 0-1350, 1350-1700, 1700-2050, 2050+ .. if there are next to no matches for a given pairing, that category isn't used (the lower category with more information is used).
The 350 ranges were chosen because they were approximately the SD of the distribution of the matches.
I doubt you're going to see people changing their relative positions, however. Everyone's RandomOracle predictions didn't pan out... his rating actually went up. |
|
|
Hitonagashi
Joined: Apr 09, 2006
|
| Posted:
Dec 16, 2011 - 14:41 |
|
| VoodooMike wrote: |
Rating systems may generate numbers on a per-match basis, but you have to put that into context - ratings are aggregated information across a large number of matches, and any inconsistent performance aspects will come out in the wash because of that. If you consistently win 2-0, then your rating will be stable and likely higher than the guy who has some 3-0 wins, a bunch of draws, and a bunch of losses. If some guy consistently wins 3-0, then yeah, I'm still convinced he's better than you are, all this talk of strategic conservatism notwithstanding. |
Touche.
Yep, I agree with the logic completely. The only complaint I have left is that we aren't being incentivised with the current system to win 3-0, but that's probably sour grapes on my part  . The good coaches (except Malmir for some reason) certainly appear to be performing well under your ranking system.
Does your system take into account relative coaching skill? (I.E. if I beat RandomOracle 3-0, I'm a better coach than if I hammer randomnewb101 3-0) |
|
|
JimmyFantastic
Joined: Feb 06, 2007
|
| Posted:
Dec 16, 2011 - 14:46 |
|
What about xboxlive trueskill ranking? does that have any merit? |
_________________
Pull down the veil - actively bad for the hobby! |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 16, 2011 - 15:07 |
|
| Hitonagashi wrote: | | Yep, I agree with the logic completely. The only complaint I have left is that we aren't being incentivised with the current system to win 3-0, but that's probably sour grapes on my part Wink. The good coaches (except Malmir for some reason) certainly appear to be performing well under your ranking system. |
Well, there are two sides to Blood Bowl - there's your coaching ability on the field, and your ability to manage the team between matches. Concepts like sweet spotting are the managing side of things.. its where you can maximize your chance of success on the field by knowing how to build your team, and how to stay at the best range for your team, and so on. That doesn't mean you're good at playing the game, of course, we just assume that people get better at both aspects over time.
This rating is about on-the-field playing skill. Someone who is really good at the off-the-field stuff will not easily be able to raise their rating just through management. Only by doing well ON the field will their rating be high.
| Hitonagashi wrote: | | Does your system take into account relative coaching skill? (I.E. if I beat RandomOracle 3-0, I'm a better coach than if I hammer randomnewb101 3-0) |
No, it can't do that at the moment, and unless the system was deeply integrated with the operation of the site over a fairly long term, it never could. The reason for this is twofold - first, right now it doesn't know what the other coach's ratings are as it calculates your rating, in order to alter the calculation in any way. Second, your rating right now was not necessarily your rating at the time of the match any more than the other coach's rating right now is the same as it was, so it would require historical rating data to apply to the modern calculation. |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Dec 16, 2011 - 16:23 |
|
| VoodooMike wrote: |
Well, there are two sides to Blood Bowl - there's your coaching ability on the field, and your ability to manage the team between matches. Concepts like sweet spotting are the managing side of things.. its where you can maximize your chance of success on the field by knowing how to build your team, and how to stay at the best range for your team, and so on. That doesn't mean you're good at playing the game, of course, we just assume that people get better at both aspects over time.
This rating is about on-the-field playing skill. Someone who is really good at the off-the-field stuff will not easily be able to raise their rating just through management. Only by doing well ON the field will their rating be high. |
If we are measuring coaching skill should we be requiring a coach to use a range of teams?
DukeTyrion was Sprint Champion over several categories last time out.
How much of Random's success is down to having the time (and skill) to build a great team?
| VoodooMike wrote: |
No, it can't do that at the moment, and unless the system was deeply integrated with the operation of the site over a fairly long term, it never could. The reason for this is twofold - first, right now it doesn't know what the other coach's ratings are as it calculates your rating, in order to alter the calculation in any way. Second, your rating right now was not necessarily your rating at the time of the match any more than the other coach's rating right now is the same as it was, so it would require historical rating data to apply to the modern calculation. |
I would have thought that the rating of the opponents would be much more important than any racial matchups or TDs scored. |
_________________
O[L]C 2016 Swiss! - April ---- All Star Bowl - Teams of Stars - 2 more teams needed |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 16, 2011 - 22:28 |
|
| koadah wrote: | | If we are measuring coaching skill should we be requiring a coach to use a range of teams? |
Not necessarily. It depends on your definition, I suppose. The rating shows you how well they do, relative to the average, across all games played with all teams they have active. Different races have different play styles, so it's almost to be expected that most people will be better with some races than others. If you wanted to claim you were good with all of them then sure, you'd need to play equal numbers of games with all different races.
| koadah wrote: | | How much of Random's success is down to having the time (and skill) to build a great team? |
You tell me! All I can say is that on a per-match basis he performs quite a bit better than average, regardless of the TV level, and consistently so. Everyone seems to be looking for a "yeah, but..." in regards to randomoracle - I've seen people doing so even before this rating - but maybe he's just that good <shrug>.
| koadah wrote: | | I would have thought that the rating of the opponents would be much more important than any racial matchups or TDs scored. |
It's circular. Think about it... how do you calculate the other guy's rating in order to take it into account when calculating your rating, if his rating has to take into account your rating? Apply that issue to every coach of every team of every match you've played.
That said, skill is great, but there is a definite imbalance between races, and if we're using averaged data then that means we're getting data from both good and bad coaches, so the skill level will balance out to some degree, making the TD difference information valid regardless. |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Dec 17, 2011 - 00:29 |
|
| VoodooMike wrote: | | Everyone seems to be looking for a "yeah, but..." in regards to randomoracle - I've seen people doing so even before this rating - but maybe he's just that good <shrug>. |
If he wins the Fumbbl Cup people will just whine even louder about C-POMB.
To be a true Box Legend you must win the Human Premiership!
(Shameless )
Come on KFoged. You too!
| VoodooMike wrote: |
| koadah wrote: | | I would have thought that the rating of the opponents would be much more important than any racial matchups or TDs scored. |
It's circular. Think about it... how do you calculate the other guy's rating in order to take it into account when calculating your rating, if his rating has to take into account your rating? Apply that issue to every coach of every team of every match you've played. |
I see that for a short season which but for an on going rating you can pick a zero date like Box switch over day. Treat everyone as even on that day and work forward from then. |
_________________
O[L]C 2016 Swiss! - April ---- All Star Bowl - Teams of Stars - 2 more teams needed |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 17, 2011 - 02:10 |
|
| koadah wrote: | | If he wins the Fumbbl Cup people will just whine even louder about C-POMB. |
Sour grapes are... sour.
| koadah wrote: | | I see that for a short season which but for an on going rating you can pick a zero date like Box switch over day. Treat everyone as even on that day and work forward from then. |
What it comes down to is unless you store the rating a coach had at the time a game was played, you're not going to be incorporating that into modern ratings. Doing that sort of thing is beyond the scope of this experiment. We don't even have anything approaching easy access to modern information - the historical information is based on the days-long data scrape from a while back.
In a perfect world, Christer would give us, or some of us, direct query access to the data to work with. As it stands, while much of the data can theoretically be gotten at via the api, it is a ridiculously slow process to snag it, and then an additionally slow process to put it together in a useful way. We work with what we've got. |
|
|
|
|
| |
Gnome Roster - how a...
Search
