Frankenstein
Joined: Aug 02, 2003
|
 Posted:
Sep 02, 2003 - 19:05 |
 
|
First of all, I'd like to thank Christer, Klipp, Grumble, Skyjunkie etc. for the terrific thing they've accomplished.
That said, there are, from my experience, major flaws in the current STR system:
- Block, the arguably best and most important skill in the game, is highly underrated (i.e. rated just like any other ordinary skill)
- Stunties with ordinary skills are overrated.
- Khemri are underrated. This might contradict with the specific statistics and is based on the experiences I've made with my Khemri-team.
- Wood Elf Linos are overrated. Especially in comparison with their Pro-brethren. I think this is due to the facts, that neither MA+1 nor AG 4 on the entire team have such a great impact as the current formula suggests.
- I have observed that Halflings with an identical skill and attribute line (except for their lower AV) as Goblins are rated higher. If I have understood the formula correctly, this phenomenon should not occur and might indicate a bug in the formula, unless there is an extra-formula for flings.
Even if statistics indicate that the str-formula predicts win-probabilities properly, I'd still go with Churchill. I assume that many coaches (including me) can't exploit the strengths of their teams and the weaknesses of their opponent's team to a full extent, and not everyone out there plays to win (and coaches who concentrate on eliminating other teams rather than on winning won't play Wood Elves, for example).
I'd like to see if the same statistics based on LRB-rating (without mng's and cash) would produce different results. I very much expect them to be as accurate as the str-formula.
Furthermore, I'd also like to see a 3rd factor added: race vs race at a specific rating. Some races are stronger than other races (classic mismatch: Dwarves vs. Amazons at TR 100). Since the differences are dependent on the TR of the involved teams, this might be too difficult to implement though. |
|
|
sceadeau
Joined: Aug 02, 2003
|
| Posted:
Sep 02, 2003 - 19:56 |
|
Unfortunately, I believe most of you are missing the actual point of the strength system. It is not for you to find people to play against. The fact that you use it as such shows how little we respect bloodbowl's built in strength system, the Team Rating.
The strength formula is strictly meant as a way of determining who should win for the coach/team rank formulas. It's basically a fumbbl only thing, its SOLE PURPOSE is to come up with a number that can be plugged into other formulas to determine who the winner should be.
Coach rating 130 with a team of str 150 is not equal to a coach of rating 200 with a team of str 150 on the field. The coach rating, since it is tied directly into the strength system, shows how well you understand and can exploit the strength system loopholes.
One of which may be how many players you happen to have on your team.
I've proposed something more dramatic than most of you, and that would be to eliminate the STR number from our eyes completely. Not just how it's calculated, but the entire number. Let the fumbbl site figure out how well people are doing based on it's own internal statistics, and let everyone else just go out there and play.
Honestly, nowhere has Christer ever said we should use STR to determine who we should play (str 100 zon vs dwarf is the perfect example of a game i wouldn't recommend outside of forced encounters). Grumbledook is right in saying just look at your opponent's roster and figure out if they're someone you'd wish to play against.
It's a tool, not a crutch. Let it go and get back to playing the game. |
|
|
EvolveToAnarchism
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 03:42 |
|
I'll preface my comments with these wonderfully wise words from Mr-Klipp.
| Quote: | Mr-Klipp:
Str is a measure of how likely that team is to win based on players, stats, skills, ect. It does not factor in coach ability. Coaching ability is reflected in ranking. |
Now to the main issue of this thread.
| Quote: | Bohm:
I think you are missing the point Christer, What roos is saying is that players in excces of 11 cont fully aginst the str ration even thoug they are of limited use....... |
| Quote: | Korhil:
STR may not account for Reserves as you'd like to see it - lowering their value...
But when you consider the alternative, TR, Reservers count full value there too. Keep that in mind |
.
Korhil presents a false dichotomy. It's not an either or situation. Christer has stated that the STR system is open to tweaking. And if it wants to do the ambitious task that Klipp proposed above, it probably still needs some tweaking. I believe that Bohm is onto something.
Evidence for Bohm's point: Imagine a 16 player dwarven squad. Will players #15 and #16 ever hit the pitch? I highly doubt it. Do they inflate the team's STR without increasing equivalently increasing the team's chances to win?
Problem with tweaking it for excess players: Imagine a 16 player halfling squad. Will players #15 and #16 ever hit the pitch? Hell yes! Do they inflate the team's STR without increasing equivalently increasing the team's chances to win?
Gut feeling: Something needs to be tweaked. I rarely check my opponents team before playing a match. Upon joining a game, I've frequently been quite suprised that the STR says the teams are an even match-up (thus I've unfortunately begun checking rosters some of the time in the few open games I still play). If I had a DP on my roster, it might be a little closer but without, I'm left with the feeling that I'm seriously outclassed by the opposing team. But alas I'll play the game, and since we are within 15 STR it can't show up as an anomaly on the STR vs TR stats page.
Another STR issue: I finally caved in and started a C.H.A.O.S. Pact team. I started with 11 Marauders and 4 RRs. It's the equivalent of a norse team that started with 11 Linemen and 4 RRs except that the Norse's RRs cost 60K instead of 90K for the Pact. The problem is that the EXACT same rosters have two different STRs because the pact RR is more expensive. If the STR objective is as Klipp stated above, it fails in this scenario. Which poses the following question: Is there any merit to RRs adding differing amounts to team STR depending on the team race?
An old STR issues: Has anyone done a data analysis on some of the TR100 games looking for systemic racial flaws? Or asked to do an even more ambitious thorough statistical analysis of all the data? Personally I'm curious as to how muh an improvement the current STR formula is over Mirascael suggestion.
Sorry for posing so many questions without providing many answers.
As Always,
Evolve To Anarchism
C.H.A.O.S. Pact |
|
|
cataphract
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 03:57 |
|
I agree, with warhammer and roos... the strength, as was said earlier
"Str is a measure of how likely that team is to win based on players, stats, skills, ect. "
indeed the status of reserve also needs to be factored in! in the same way MNGs don't add to team strength so should reserve players only add part of their full value since they will not spend the full game on the pitch.
By calculating strength from the top 11 players and then adding partial strength for the oher players we should gain a better understanding of "how likely that team is to win"
I use this (theoretical) example
a wood elf team with thirteen players and no skill rolls (unlikely i know but work with me) has a team strength of X
Anoter wood elf team with 11 players, and a few skill rolls (some blocks, dodges, or whatever)
also has a strength X
The advantage of the ON FIELD skills is more than the advantage of OFF FIELD players... the second team is IMO more "likely to win" and thus the team str should reflect that. |
_________________
"the eunuch should not take pride in its chastity" |
|
BadMrMojo
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 04:15 |
|
| Evo wrote: | | Another STR issue: I finally caved in and started a C.H.A.O.S. Pact team. I started with 11 Marauders and 4 RRs. It's the equivalent of a norse team that started with 11 Linemen and 4 RRs except that the Norse's RRs cost 60K instead of 90K for the Pact. The problem is that the EXACT same rosters have two different STRs because the pact RR is more expensive. If the STR objective is as Klipp stated above, it fails in this scenario. Which poses the following question: Is there any merit to RRs adding differing amounts to team STR depending on the team race? |
Well, the Norse team has a slightly better chance of winning its first match and therefore should have a slightly higher TS than the Pact team. Unless the Norse have 120k sitting in the bank at creation, they must have additional FF or coaching staff and an appropriately (slightly) larger chance of winning. |
_________________
Ta-Ouch! of BloodBowl
Condensed Guide for Newbies |
|
Britnoth
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 04:51 |
|
Why not use the average value of all available players? That would encourage specialist reserves..... |
|
|
EvolveToAnarchism
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 06:05 |
|
| Quote: | BMM:
Well, the Norse team has a slightly better chance of winning its first match and therefore should have a slightly higher TS than the Pact team. Unless the Norse have 120k sitting in the bank at creation, they must have additional FF or coaching staff and an appropriately (slightly) larger chance of winning. |
If the example didn't make it clear. I am assuming the norse has banked or otherwise lost 120K. As I obviously didn't make clear enough when I said they have the EXACT same roster. According to the present STR calculations with RRs adding different amounts to STR, the Chaos Pact roster would be favoured to win when playing against the exact same team that just happened to be Norse.
As Always,
Evolve To Clarity |
|
|
Mr-Klipp
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 06:12 |
|
Any system such as this is only going to be able to give you an overall average. There will be some teams that need their rerolls more than others, but the only way to factor in this cost would be to somehow figure out how usefull a reroll is to the team as it currently stands, and I don't think you will come up with a forumula for that. The high reroll cost of pact teams is a special case, as it is a balancing factor agianst there ability to field players of multiple races. If you have some idea of how to better calculate the value of rerolls for a given team I'd love to hear it. |
_________________
Looking to get your minis painted? Look no further.
The Finishing Touch |
|
Britnoth
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 08:40 |
|
Well you can make a rough estimate at least. 
Big guys don't benefit, so make each big guy reduce the value of rerolls by say 10%.
Skills replace rerolls often, so make pass, catch, dodge etc reduce the value by say 1% each? |
|
|
Mr-Klipp
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 09:34 |
|
| Britnoth wrote: | Well you can make a rough estimate at least.
Big guys don't benefit, so make each big guy reduce the value of rerolls by say 10%.
Skills replace rerolls often, so make pass, catch, dodge etc reduce the value by say 1% each? |
Neither of these is truly applicable, as if you don't use the reroll on these players you will have it available for another action instead, or a block, or some other needed roll. In the end, I feel that there are far too many possibilities to attempt to consider them all. The cost difference between rerolls on different teams already takes this into account, giving teams with higher stats and more skills cheaper rerolls. This should be a good enough generalization for our purposes. |
_________________
Looking to get your minis painted? Look no further.
The Finishing Touch |
|
Britnoth
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 11:36 |
|
| Mr-Klipp wrote: | | Neither of these is truly applicable, as if you don't use the reroll on these players you will have it available for another action instead, or a block, or some other needed roll. In the end, I feel that there are far too many possibilities to attempt to consider them all. The cost difference between rerolls on different teams already takes this into account, giving teams with higher stats and more skills cheaper rerolls. This should be a good enough generalization for our purposes. |
Available yes, but not always used and if it is it is on a less important player. In a stunty team your big guys are worth several normal players... (at least) and having rerolls for when your Wild Animal is forced to block and rolls skulls saves your entire sides movement for that turn.
Also - if you are in a position to score you can expect the player to attempt just that, cutting out any blocking etc unless its strictly necessary. Every action from picking up the ball, passing it, catching it, dodging past players and using GFI can benefit from skills which replace a reroll. Non-bashing teams benefit less from them as far as I have seen.
I don't understand why stronger teams at the start should have cheaper rerolls too, maybe I haven't been playing long enough.
PS. If it is a 'good enough generalization' why are people talking about it isnt? |
|
|
EvolveToAnarchism
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 12:08 |
|
| Quote: | | Mr Klipp: The cost difference between rerolls on different teams already takes this into account, giving teams with higher stats and more skills cheaper rerolls. This should be a good enough generalization for our purposes. |
I just don't think this cuts it as a justification for weighing the RRs as differently because of a team's starting races.
The example I sited of my 11 marauder team shows you that differently weighted RRs skews the STR formuala from it's intended purpose. You got two EXACTLY the same rosters whose STR is different by 12.
With all the variations in skill choices and injury effects it's possible to have two teams of different races with the exact same roster but with differently valued RRs. I know it's not likely that you'll get too many EXACTLY the same rosters of different races, but it's likely that there are some pretty equivalent rosters where this type of problem arises.
I'd like to think that all the games I've played has helped me judge the relative strengths of different rosters. But my little thematic pact team doesn't even need any use of questionable judgement. I think it's obvious that my original roster was given too high a STR. Whether it carries over when the team is at a higher TR is open to debate which relies heavily on the whole issue of RRs.
Re: RRs. There's the question of the justification of the cost of the different teams RRs. What are the potential reasons for team RRs costing different amounts? I believe there are several potential explanations:
1. RRs are costed to balance out TR 100 rosters.
2. RRs are costed to balance out team development.
3. RRs are more effective for different teams.
4. Jervis followed his numerologists advice on how much they should cost.
(this list is by no means exhaustive)
Whether RRs contribution to STR should vary, largely depends on if #3 is true. I'm not sure if this is the case. I haven't seen any very coherent arguments its truth. Especially on developped teams. Any takers on my challenge to justify #3? Be sure to justify it all TR levels.
Any rebutal to my concerns with benchwarmers relative value to STR?
Trust one of the few coaches who opts completely out of the whole ranking business to find issues with the STR formula. But I guess that happens when you build thematic teams. My critique of the STR system is primarily an attempt to give something back to the FUMBBL community in an area where I hope my skills are useful.
As Always,
Evolve To Anarchism |
|
|
Mr-Klipp
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 13:12 |
|
| Britnoth wrote: |
I don't understand why stronger teams at the start should have cheaper rerolls too, maybe I haven't been playing long enough.
|
Because they *are* stronger, and thus less likely to *need* them. Rerolls are costed by how usefull they will be to a team. A team with better stats and more skills needs the rerolls less, and therefore they will give them less benefit, and therefore they are cheaper.
| Britnoth wrote: |
PS. If it is a 'good enough generalization' why are people talking about it isnt? |
No matter what you do, there will be someone who disagrees with you. The fact that you can find someone who disagrees with an idea does not in itself make it wrong. |
_________________
Looking to get your minis painted? Look no further.
The Finishing Touch |
|
Mr-Klipp
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 13:22 |
|
| EvolveToAnarchism wrote: | The example I sited of my 11 marauder team shows you that differently weighted RRs skews the STR formuala from it's intended purpose. You got two EXACTLY the same rosters whose STR is different by 12.
|
This is a special case that could only ever occur with a pact team, a team, I will add, that is not yet tested and balanced to the rest of the teams. Not to mention, this will only occur with a very specific and limited variation of certain pact teams. One very rare occurance does not prove a complete system flawed.
| EvolveToAnarchism wrote: |
With all the variations in skill choices and injury effects it's possible to have two teams of different races with the exact same roster but with differently valued RRs. I know it's not likely that you'll get too many EXACTLY the same rosters of different races, but it's likely that there are some pretty equivalent rosters where this type of problem arises.
|
The chances of this with any other roster but the pact teams are more than a million to one. The only time this could ever occur is if *every* player on a team got the same stat bonuses, and some rather rare stat bonuses at that. The teams that are the closest in stats have for the most part the same reroll costs. Again, one possible rare example does not disprove the overall system.
| EvolveToAnarchism wrote: |
Re: RRs. There's the question of the justification of the cost of the different teams RRs. What are the potential reasons for team RRs costing different amounts? I believe there are several potential explanations:
1. RRs are costed to balance out TR 100 rosters.
2. RRs are costed to balance out team development.
3. RRs are more effective for different teams.
4. Jervis followed his numerologists advice on how much they should cost.
(this list is by no means exhaustive)
Whether RRs contribution to STR should vary, largely depends on if #3 is true. I'm not sure if this is the case. I haven't seen any very coherent arguments its truth. Especially on developped teams. Any takers on my challenge to justify #3? Be sure to justify it all TR levels.
|
As I've discussed with you before, the teams with lower reroll costs have in general one or more of the following. More skills, better skill access, and higher stats. These factors make the team less likely to need a reroll on any given action, regardless of TR. As an example, AG 4 with drastically reduce the number of ball related failures you suffer in a game. When you combine this with full access to agility skills, and a thrower with pass access, you have a team that will have less need for rerolls. Contrast this to a chaos team with no pass access, no agility access, and no ag4, and you should be able to quickly see that barring some very good luck with doubles rolls, the chaos team will have a greater need of it's rerolls to perform these sort of functions no matter the teams TR. |
_________________
Looking to get your minis painted? Look no further.
The Finishing Touch |
|
sceadeau
Joined: Aug 02, 2003
|
| Posted:
Sep 03, 2003 - 13:24 |
|
Evo:
When I first worked with Christer on the STR formula, I took those exact numbers into account. We found the average number of useful players (av 9 was 11, av 8 --> 12, av 7 ---> 13, av 6 ---> 14) and started to reduce the average str of the team after that a small bit. If you had less than these number of players, your str would be reduced dramatically. If you had more, it would be reduced.
As we all know, a 16 man dwarf team is a lot like a 13 man dwarf team, except they have 3 more dwarves with no skills who sit in the box the entire game. And a 9 man halfling team is most likely won't win anything.
However, I take it that this led to some abuse of the ranking system, and it had to be changed. As it stands now, the high AG teams are still winning more than they should mathematically win. That either means the AG coaches are better than other coaches, or the math is off. We think it's the latter 
The formula is known to be wrong even now, however...the problem comes from finding better numbers to use...
I do think there is another way, a better way...however...it would probably make most of the players unhappy. It would be a player by player analysis of matchups AFTER they play...so there would be no STR number at all, but instead a comparitive team v team number. So it could look at newbie dwarves and newbies 'zons and figure out that that's a mismatch.
The problem with this system, of course, is that you won't get a STR number at all. STR would be calculated internally for ranking purposes, only. That would mean coaches would have to go back to just comparing rosters when they want to play....
Although, from what I've seen...almost everyone does this anyway. They use str to propose matchups, and then both go look at rosters to make sure they want to actually play it. So this wouldn't really affect much.
The STR formula will HAVE to change once we get a handicap table in javabowl. At the very least, it will have to become a function of TR, which means new numbers and a reworked system. |
|
|
|
|
| |
Gnomes are trash
Search
