Ed. Scott Emery yobbo@shell.portal.com
Contents of this issue:Letter FROM the editor: Penultimate issue
Pontification: Pontificus Tacticus
Feature: How the Draft Works
Feature: Enhanced Replacement Strategy
Letter FROM the editor: Penultimate issue
This is the Penultimate Issue of TOW as a mildly regular series. TOW has satisfied its original mission, To help beginning managers get acquainted with SPARF and provide a sounding board for new strategies. Now, with a little reading and head-scratching, a beginning manager can avoid most of the classic and team-wrecking mistakes that have plagued their fellows. In the final issue I will write another article trying to tie stuff together. If there is something that you have been burning to say in this forum now is the time! The final issue should come out after the end of the season.
There will be archival issues of *something* coming out to document changes in SPARF and if the changes are significant, maybe TOW will rise again!
Scott Emery
Soon to be ex-TOW editorPS Thanks to Evan Harris of the Uluru Dingoes for providing the feature articles this time out!
Pontification: Pontificus Tacticus
I decided to take a look at imbalanced training vs. balanced training as the PPP is tearing up Silver, yet everyone in gold has an imbalanced team structure. I modifed SPP to handle rosters and generate a single number when both of the player's stats are known. I then built a fake team by taking a balanced player and created unbalanced players by moving around skill in lots of ten. I then matched them up against the balanced players that they were generated from and the chart below is what I came up with.
Note that in the below fake field, Every letter is at the top of its range. L = 10, M = 20, G = 30, V = 40, T = 50, O = 60, E = 70.
TOP: Top Varys
BOTTOM: Bottom Even
ts LFP ts FF ts RFP
be RBP be FB be LBP
62 55 g 60 53 g 58 53 g
LFP T G G L m FF O V V M m RFP E T T G m
RBP G G G G w FB V V V V w LBP T T T T w
37 56 p 40 54 p 41 53 p
ts LHF ts CHF ts RHF
be RHB be CHB be LHB
57 57 g 55 55 g 54 54 g
LHF V G V L m CHF T V T M m RHF O T O G m
RHB G G G G w CHB V V V V w LHB T T T T w
42 60 p 44 57 p 45 55 p
ts LW ts C ts RW
be RW be C be LW
62 50 60 50 58 50
LW T G M M m C O V G G m RW E T V V m
RW G G G G m C V V V V m LW T T T T m
37 50 40 50 41 50
ts LHB ts CHB ts RHB
be RHF be CHF be LHF
50 35 g 50 40 g 50 42 g
LHB G G L T m CHB V V M O m RHB T T G E m
RHF G G G G w CHF V V V V w LHF T T T T w
50 40 p 50 42 p 50 44 p
ts LBP ts FB ts RBP
be RFP be FF be LFP
40 40 g 42 42 g 44 44 g
LBP M G M T m FB G V G O m RBP V T V E m
RFP G G G G w FF V V V V w LFP T T T T w
60 42 p 57 44 p 55 45 p
ts ROV ts RKM ts RKR
57 40 g 55 42 g 54 44 g
ROV V G L V m RKM T V M T m RKR O T G O m
w w w
p p p
be ROV be RKM be RKR
42 42 g 44 44 g 45 45 g
ROV G G G G m RKM V V V V m RKR T T T T m
w w w
p p p
te IC1 te IC2
IC1 V V V V IC2 V V V V
be IC1 be IC2
IC1 V V V V IC2 V V V V
I generated this beautiful field with "exact" numbers, just to see if there are any suprises in the mix. Mark turns out to be more dominant than expected, and there doesn't appear to be any reason to use defense. !???! Now, I don't know any championship team that hasn't had some outrageous defense on it, so what gives here? I racked my brain, uh, "Well, maybe defense affects kick?". It was more than I could figure with the information I had so I turned to Mel. Boy, was I in for a shock.
Let us neatly sidestep a complicated issue (that we didn't discuss) and deal with the ball after it has been caught. If there was a mark, then there are no influences on the ball other than the kick score of the player and the distance from the goal. The formula one and a half times the kick score minus the number of SPARF tiles (about 1 1/2 meters) from the goal gives your chance to score. After a mark, the player sizes up all of the options available and takes the most attractive of them which is usually a shot on goal, if the player can kick that far.
So, what if the player didn't mark? Well the one rule in ARF that I know is "you gotta do *something* with that ball". The first thing that the simulator does is figure out what the player *can* do. The simulator positions the defence and the offense. It then eliminates impossible plays (you can't punch to the right if Wade Overhiyr is in your face over there). The chance of success is calculated for the list of possible plays, then the list is randomized and assigned monotonically increasing penalties based on the formula provided in TOW. Once the modified chance of success has been calculated it is multiplied by a "base desirability" for the action and this results in the modified desirability of the action. The offending player then takes the action with the highest modified desirability with the modified chance to succeed. Since a shot on goal has a higher "base desirability" than a pass backward, players will occasionally shoot even though they have a better chance of success passing backward. It sort of makes sense, doesn't it?
Now, on to the serious pontification:
So, where does that leave us with the field as laid out above? The left factor is an indicator of the chance to mark. When your mark is higher, then it is higher. If you assign mark skill to other skills and you lose your chance to take that unencumbered kick. My guess is that the mark chance is around .1% * mark factor, based on the number of marks per game that I see and my guess of the number of markable transactions that occur. The right factor is the outcome of the formula provided in TOW. I suspect that it factors in proportionally to the penalty described above. The most obvious (base chance * factor ^ n) is a little harsh.
Let me leave you with one interesting tactic. Because our SPARF players are stupid (Ya gotta love them!) we can sometimes take advantage of them. During training (hard-coded in the simulator) they are told that the best thing that they can do is kick on goal (the "base desirability" for scoring is high). If some kickless forward is sitting up on the half, because only the big boys get to play around the goal, try matching up a weak defender with him. The weak offender won't have his tendency to shoot on goal squelched and he will take feeble shots that should have been passed forward. Believe it or Don't!
Feature: How the Draft Works
[Ed. note: If anyone doubts Evan's qualifications to write about the draft, he wrote the software that implements it.]
As there has been some confusion over the way the draft works, I thought a quick note on the operation of the current bid processing program works might be useful.
After sucking in the lot, bidding and (more recently) banking information, the first step is to sort the bids by the price. This was been a stumbling block for some, since it wasn't mentioned in the initial draft announcement.
The top bid is now examined. It is tested to see whether it is valid. It may be invalid because the lot may be sold out, the condition (if specified) may not be true, or the team may have insufficient funds for the bid. If it is valid, each valid bid at the same price for the same lot by different managers is collected.
If the number of these bids is less than or equal to the number of players in the lot, each bid is successful, otherwise the lot is tied and each of the bids is removed.
Remember that a team can bid for more than one player in a lot. Initially, only one player per team is considered. By only counting the number of teams and not the number of players bid for, some form of fairness is achieved. Consider a lot with 3 players which has 3 teams bid the same price for the players, two teams want one player and the third wants all three. Each team will be awarded one player, rather than the bidding being tied because one team wanted all three.
After one player is awarded to each team, the second bids from each team is considered, as a fresh invocation examining the top bid. Thus, if two teams had bid the same price and both wanted all three players, they would have got one each initially, and then tied on the third player.
This process is repeated until all bids have been examined and either awarded, tied or rejected.
Hmmm. I hope (optimistically?) that that was clear. Any questions?
If anyone has any suggestions on how the process might be improved to permit the sort of thing that some have wanted to do, please let Mel know. I am safe in the knowledge that if it's not sufficiently easy to do, I won't be implementing it. 8-)
As a starting point, perhaps priorities could be added to bids such that a bid is not considered until higher priority (but lower price) bids have been processed. Whether this is feasible (or useful) requires more thought than I'm prepared to give it now.
Feature: Enhanced Replacement Strategy
Enhanced Replacement Strategy -----------------------------
Injuries in recent weeks have reignited my interest in an enhanced replacement strategy. Currently, we are limited to a 2F line, which I will assume everyone knows about.
I think I posted something like this to the kibitz list when it was a list for suggesting changes, a long time ago. It was an enhanced version of mail I sent Mel, and I've enhanced it again on this pass.
Basically, the idea is to have multiple "replacement candidate" lines, and a single "precedence/priority" line. When a player is injured, his immediate replacement from each candidate line is taken and the one with the highest precedence is used.
For example, consider these candidate/precedence lines.
Cand: RKM ROV RKR C RW LW LHF Cand: FB CHB RBP LBP LHB RHB IC2 Cand: CHF LHF LFP RFP RHF FF Prec: RKM ROV RKR C RW CHF LW LHF LFP RFP RHF FF FB CHB RBP LBP LHB RHB
Each line has IC1 and IC2 implicitly appended to the end of it.
Now, if the RKR is injured, the C moves to RKR, the RW moves to C, the LW moves to RW, ..., up to the IC1 moving to LHF.
LHF is on two lists, what happens if he is injured? Well, we collect the candidates from the first and third lines, they are IC1 and LFP. Since LFP has the higher precedence, this player is moved to LHF and LFP is then replaced by the same mechanism.
Loops are not possible due to the Prec line. What an implementation would do each time a player is injured is consider each position on the Prec line in turn.
In pseudocode:
foreach position ( Prec: ) if starting_player[position] is not injured player[position] = starting_player[position] else collect candidate players who are uninjured and not already positioned from Cand: lists player[position] = highest candidate in Prec:
Random Comments
Each position string is a little misleading. In some cases we mean 'the position in which we are placing the player', in others we mean 'the player who was initially selected in position X'. We do not mean 'the player currently in position X'.
Each position would need to be in the Prec: line, unless some default ordering of positions were used.
What if more than 2 players are injured? That's a little difficult since I don't know what currently happens. In "real life", the least injured comes back on, is placed in the forward pocket, and basically does nothing. My preference is to have a virtual IC3, IC4, ... which are filled by injured players in order of seriousness of injury (or likelihood of reinjury). They then get placed after IC2, as normal.
Rawstats would have to be enhanced, but perhaps this whole thing is a job for viewsim.