Monday, February 6, 2012

Margin of Victory in a Tournament Setting - Why Not

The NOVA's 40k formatting as a general rule ignores the impacts of Margin of Victory, and therefore the impact of Battle Points.

For us, it is marginally relevant to Best Overall (i.e. less than 5% of your score), but even that will likely change and diminish going forward.  It also impacts seeding within same win-loss record for the Rounds 5-8 rebracket, but does not otherwise impact your score.

Why doesn't the NOVA use Margin of Victory as a meaningful mechanic, though?  This discussion came up in comments on YTTH, when someone wrote the following ...


Hey MVB and Neil,Over here in this part of the country, we use the margin of victory to determine who won the best. That way it is just like in the Ardboyz. So a person can get from a total loss (massacred) to a total victory (massacre).Why do you think the margin of victory is so bad? I find it really great for a small store that has small tournaments. Do you think it bad for tournaments with less then 20 people? Or do you just think it is bad, period, for everyone everywhere?I would love to hear more of your input about this. The margin of victory has worked pretty well so far, but if there is a better way of running a tournament then doing that, then it could very well be worth considering for the group I play with.Thanks dudes!

I gave a brief reply there, but decided to go into greater rambly depth here.
First, let's acknowledge what the 40k tournament scene is and is not.
  • There is no barrier to tournament entry
  • There's no meaningful national ranking system (Rankings HQ is meritless, b/c it does not rank people along standard tournament formats or sizes, and weights # of events participated in over % of possible score regardless of # of events, and a million other issues that could be a separate blog post)
  • There's no standardized tournament format 
  • There's no known way to skill bracket, and there are no 40k professionals
  • Due to the lack of professionalization and ranking, all first round pairings are random (or in some rare cases determined by subjective evaluations of army composition)

So theoreticals to highlight why this presents a severe problem for Margin of Victory:
Major tournament, Round 1
Table 1 - Player A (highly skilled) barely beats randomly paired opponent Player B (also highly skilled)
Table 2 - Player C (highly skilled) annihilates randomly paired opponent Player D (his first tournament 40k game ever)
In a simple bracketed tournament approach,Players A and C advance, and continue to play at wherever they would as part of a bracket, against other winners. If they continue to win, they will eventually face each other and determine who is best for that Tournament.

Pause and interjection - Tournaments NEVER DETERMINE THE BEST PLAYER They ONLY determine the best player for that tournamentA million variables from random pairings and match-ups to the issues in someone's personal life or amount of sleep all factor into how each person will play, especially in a game with dice as a determinant factor. LEAGUES and SEASONS are used to determine the best, and even those ONLY determine the best for that league/season. The goal, therefore, should then be to fairly determine precisely that ... who is the best for the tournament.

Let's go back to Player A and Player C, and put them in a Margin of Victory event. 
Player A is considered inferior to Player C at this point in time. Player C's score is higher, because he scored a Massive Victory against a beginner. Player A scored a Marginal Victory against an equivalent opponent and therefore scored much lower in a Margin of Victory format.
Question Series:How many professional sports use Margin of Victory to determine the winner of their tournaments? Is the Super Bowl determined by Battle Points scored in Rounds 1-3 of the NFL playoffs?Is the NCAA basketball tournament halted after Round 3 and the Championship awarded to the team with the most points over 3 rounds?How many people are happy with the way NCAA Football determines its title game? At least instead of arbitrarily awarding the Championship to the highest BCS-rated team, they have a play-off between the top two that's oriented around simple Win-Loss. Imagine if only a bare victory in triple overtime there was ruled as "too close to call margin-wise" after the fact.

The downsides to Margin of Victory are amplified by fewer rounds, and by more participants.
Fewer rounds means more undefeated players at the end from which Margin of Victory events assign a de facto-arbitrary Best General. More participants means a higher likelihood of some skilled players (unintentionally) dodging tough opposition while other skilled players (unfortunately) randomly face only peers.
The biggest issue people bring up is, "What about a small event in a smaller store? You don't have time to pare things down to a single undefeated for Best General. What if you have 16 players and only the time/resources for 3 rounds?"
My recommendation?
At the end of Round 3, split your Best General prize two ways between the two undefeated players, and award a Best Overall that's 50% win rate and 50% appearance score (the NOVA calls this award Renaissance Man, and it is our highest value prize).
Now you aren't telling one of the undefeated players their wins weren't good enough (even if they were against extremely tough opponents), and you are still awarding a BEST Overall (so, don't fret, there's still a single Winner if not having one bothers you). This is elegance at its best - not only do you reward a pristine performance equally across players who accomplish it, you reward Best for the person who is the best present at the entire Hobby, not just the best Gamer. If multiple players are still undefeated when your event ends, let's also face the facts - you have NOT yet determined the best Gamer at the event anyway.
Tournaments should strive for fairness above all else. The best way to enhance sportsmanship and the feelings people experience in the aftermath of the event is to provide the fairest environment possible for them to compete and participate in the fashion they personally desire to.
When you consider random pairing, variable player skill, and the many other complications already mentioned, it is inherently unfair to assign value to the amount someone wins or loses a game by.
The responsibility of a TO in my humble opinion is to do all we can to eliminate or further marginalize unfairness.
 There are some things about the game of Warhammer 40,000 you cannot and should not change, because doing so would inherently change the game itself. You cannot take dice out of it, the random rolls for first turn, the variability in army and match-up that occurs, the inherently random nature of first round pairings and the like. Adding unfairness via player-constructed (not GW-designed) tournament systems only causes more players to be upset by the results of an event, and it further discourages good sportsmanship as a result.

 - Mike

As a final note, it is always a risk for a TO to be perceived as competing with or criticizing events run differently from their own. While those I directly interact with across the tournament circuit know better, I feel always a need to reiterate: the NOVA is extremely supportive of variety and depth in the circuit. While I'm flattered by and love seeing other events run as "NOVA style," I also attend and enjoy participating in tournaments run in many other ways, including Margin of Victory. These posts are meant to share with readers why we do as we do; there are certainly many reasons a TO would use Margin of Victory, and as with any opinion in such a field - both sides are quite often right. By rejecting Margin of Victory as appropriate for the environment we seek to create at NOVA, we are not simultaneously rejecting the value of Margin of Victory in other events, formats, and games.

16 comments:

  1. So why not turn the problem on its head? Basically go from the assumption that the players with the lowest Margins of Victory are the better players?

    Basically if a player wants to game this, then they risk losing, which yields no Margin of Victory. And if they have the skill to soft-play a scrub, then that deserves some recognition.

    The Best General is then the player that won the hardest games, in terms of the narrowest Margins of Victory, with tie-breakers going to the player that lost the most material (points) pulling out the win.

    ReplyDelete
  2. You could do that, but you wouldn't eliminate the problem. It is just as likely that two scrubs play each other and one of them hamfistedly eeks out a win. :p

    The problem is in assigning value - negative OR positive - to any margin whatsoever.

    It would be a lolzy side-event to do, however, along the same fun turn-issues-on-their-heads lines as mirror matching.

    ReplyDelete
  3. I think it does eliminate the problem you've identified, which is that players with thinner margins of victory caused by playing opponents of equal skill will have their games under-valued, while players with thicker margins of victory caused by uneven match-ups will have their games over-valued.

    I'm also inclined to disagree with the assessment that it is just as likely that a scrubs will accidentally win at the same rate that a skilled players will, particularly where the players are matched by margins of victory. Over the course of eight games one might imagine that the tournament pairing format would separate the random scrubs from the skilled players.

    You might as well worry that the scrubs will accidentally win all eight games, as worry that they might consistently pull out enough, and slim enough, margins of victory to win Best General.

    ReplyDelete
  4. I think the more games you play, and/or the less players there are, the less the issue arises.

    It is still a POTENTIAL issue, no matter the quantity, and I think bigger than just "oh not likely" even over 8 rounds, but as per the comment in the article itself ... those two factors are very relevant to the, ahem, margin of the issue.

    If you used margin of victory purely, ignored w/l, and played 12 rounds with say 4 players ... you'd get a really meaningful result; heck, you'd really have more of a mini-league than a tournament, though.

    So I'm not stating Margin of Victory is unequivocally BAD, so much as sharing and enumerating the rationales for not using it in the NOVA style.

    ReplyDelete
  5. I think you have identified the biggest issues I have with "battle points"/margin of victory tournaments.

    It rewards people who have easier games, or crush their opponent, while penalizing people who play against other skilled players. This is a system that only really works when all players are of relatively equal skill.

    I think the idea of turning on its head (while interesting) does not really solve anything because if I am stomping someone all I need to do to get a smaller margin (unless we are playing Kill points or VPs), is make sure I only win by say 1 objective. Something that still favors players playing weaker players. THe other issue I have with these types of events at times is that "battle points" trumps "win-loss ratio" where a player can win all his/her games, and lose to a player who has lost a game but has more larger wins. This last part is easier to avoid, but often the mantra is that "we don't want anyone to be out of the running after just one loss"

    I think that if you want to use "battle points" to determine a winner (not split the prize, though that is also a good idea), you would need to include some kind of "strength of schedule" metric based on overall tournament performance of your opponents. That way while you may have players gain points by through massive victories over weaker players, marginal victories over better players (as determined by the current tourney) are rewarded. How one would successfully do this is still beyond me at this point (though I have played around with some ideas) I like it better than a typical battle point system. Where Players A(Highly skilled), B(Highly skilled), C(Highly skilled), D(weak/less skilled), E(weak/less skilled), F(New Player), G(Highly skilled), and H(middle of the road) play each other, but that there are other people in the tournament (with only 8 people it is easy to determine 1 undefeated winner)

    So lets say

    A plays B A wins marginally (10 points)
    C plays D C wins massively (20 points)
    E plays F E wins Massively(20 points)
    G plays H G wins Marginally (10 points)

    So lets say round 2
    A plays G (A wins 10 point marginal)
    C plays E (C massive win against 20 points)

    NOw we have a player who through luck of matchups is playing against 2 weaker players in a row, while player A plays 2 difficult games in a Row.

    Now we have a situation where player A has only 20 points, and Player C has 40, which means that if there is only 1 more round player A needs to get a Massive victory, and player C only needs to not lose. NOw these are wildy bad battle points, but it illustrates a problem with the "Margin of Victory system"

    ReplyDelete
  6. One reason the inverse isn't true is that it follows the assumption that large margin of victory = large difference in player skill. Unfortunately, this isn't the case. I have absolutely annihilated/slapped around players of equal or greater skill (and had the same done to me at times by players of 'lesser' skill) due to dice, mission conditions that favor one build over another, or just a matchup (this being the most common).

    Here's an example:

    I'll beat (almost always) any DE player so badly it'll look like they were just learning the game when I use my GK's. Is it because I'm a better player? Probably not, it's just that DE have a horrid matchup versus a mech/hench style GK's. So when you go tourney wide with "smaller margin of victory trumps larger margin of victory because smaller margin means they had tougher opponents" you have just as many problems as the situation you were trying to fix.

    Not to mention close games are usually due to an equivalence in skill level, not really a "we're both really good". Two bad players face off in a nova-style system with pure win-loss, and you'll probably have a closer margin. How does the metric work in this case?

    ReplyDelete
  7. @Andrew G - In that case, the two bad players meet, and one wins - going on to play another winner, and likely losing, evening things out.

    While DE vs GK may be a more uneven match-up than average, and your particular expertise in winning as GK vs DE may be exceptional, a balanced list tends not to get tabled by another balanced list under normal circumstances.

    Presuming that happens though, the following occurs;
    In NoVa, the DE player goes on to play a player he can beat, and comes back up to his skill level, or he loses, and drops to his skill level, as appropriate.

    In a Margin of Victory event, getting beaten so badly puts him out of the running, but you winning one so heavily won't do half as much for you as players lucky enough to get several they can massacre.

    @Mike - thoroughly enjoyable article. Honestly, I'd like to see you do an article on RHQ's 'failings' too, but I understand you being recalcitrant to criticise hard-working people...

    ReplyDelete
  8. @TKE

    I'm unsure as to whether we're agreeing or not on margin of victory being a fairly bad metric (and my post was more to point out the flaws of earlier posters idea to just flip-flop the way it's used, instead of large margin = good, have small margin = good).

    I think just reversing it amounts to swapping one set of problems for another, and that is the flaw I see. Rather than trying to make a flawed approach work, just get rid of it. (My personal opinion)

    @my de example/your comment
    I'll admit that the DE games I've played have typically been against skimmer-heavy builds, mind you, I'm hard pressed to find DE builds that aren't skimmer heavy. (this includes your standard venom spam with ravagers/etc, raider spam, mixtures of both, some using reavers, some using wwp tricks). The only variant I haven't faced that i think could be a different story, would be a more beast-mob heavy variant. My GK build I'd call a fairly balanced/all comers list, at the last GT I took it to I faced 6 different codexes in 6 different games. DE in this example just suffer paper tank syndrome, and suddenly psybacks are tank hunting monsters, and psyfle dreads might as well be destroyers of worlds.

    My point wasn't to say "look at me I beat de!" but more to try and point out that there are random, uncontrollable factors (such as match up) which can and do significantly influence margin of victory, outside of pure player skill. I used the DE example because it was the first that popped to mind as just horrendously bad.

    ReplyDelete
  9. "Tournaments NEVER DETERMINE THE BEST PLAYER They ONLY determine the best player for that tournament."

    Can this be a stamp handed out to everyone on the Internet please?

    ReplyDelete
  10. @MVB,

    We are trying something a little different at the Indy Open that I would like you to take a look at. It is a Win/Loss only event. But we have added bracket points to help determine the top player in each game bracket. The bracket points work as such:

    Win on primary-6 bracket points
    Win on secondary-5 bracket points
    Win on Tertiary-4 bracket points

    Tie breaker- both players get 3 bracket points

    Lose on tertiary-3 bracket points
    Lose on secondary-2 bracket points
    Lose on Primary-1 bracket point

    The idea here is that there are good and bad wins, and good and bad losses. As the skill level get's closer to their own, it will be harder to win on a primary mission. The concept is also meant to encourage players to play till the end of each game, because holding your opponent to a win on the tertiary means you earn almost as many bracket points as he does. These bracket points are also used as part of the Ren-man scoring.

    Just wanted to get your thoughts or thoughts of others. The link is http://theindyopen.com/blog/missions/

    ReplyDelete
    Replies
    1. We did something similar with the Bugeater GT last year and will do so again this year where we looked at success/failure on the tiered objectives to determine a "strength of victory" so to speak to help determine the spread within a particular result bracket (2-0, 1-1, 0-2, etc.).

      Since we had enough rounds to find one undefeated at the end, it avoided the problem that Mike talked about, but allowed us to more accurately pair the matches in the six games.

      Delete
  11. Hey MVB,


    This is colonelcolmcorbec again. I could not get the url to work so I picked "anymous". Thanks a lot for this. When I asked that question on steleks blog, I did not think someone would reply in such detail.


    You really thouroughly explained why the margin of victory is bad. I understand. With tournaments up at my location, there is a small number of people (very small) and the tourneys are never more then 4 games, sometimes 3.


    Your explanation helped a lot. I just do not know if for a 4 round tournament, having W/L only and then trying to see if someone gets all 4 wins with so few people will work. See, sometimes with our tourneys, there are so few people that people of lower skill fight people of higher late in the tourney (like round 3 or 4) even if they had done horrible so far.



    Thanks again!

    ReplyDelete
  12. Hey Colm again,


    And by the way, I think that up here, the 40K tourneys are going to use this system. It seems so much better then the ardboyz system. Thanks again dude. I cant wait to come to the Nova Open this year!

    ReplyDelete
  13. Mike
    I have been running our AWC league with a couple of other people and have been using a win loss draw system. We have 3 victory conditions for each mission. Usually one is some sort of kp variation one is an objective based and the last will be something a little more different like get victory points to the center of the table of get more units into your opponenet's deployment zone etc...
    If you achieve more victory conditions than your for you win. If it is tied it goes to victory points. If it is still tied than it is a draw. In the standings you get 10 points for a loss, 20 points for a draw and 30 for a win. Usually there are three round tourneys due to time constraints. So you can score 90 battle points max. The best general award is pulled from the three and oh people using number of objectives for tiebreaker. The battle points is added to the painting score and sportsmanship score to produce an overall winner. That way you can still lose or tie a game and have a shot at overall based on excelling at other areas. Has seemed to work out well for our season.

    ReplyDelete
  14. That's in a lot of impactful ways similar to what NOVA does, save for the fact that when we have multiple undefeateds, all of them are awarded something (even if a higher award is then drawn from their sum total).

    ReplyDelete