Wreck

I just got wine in the mail! Woohoo! I wondered how one ships relatively fragile things like glass bottles; now I know. The packaging is styrofoam, cleverly molded the shape of wine bottles. All bottles are intact! Mmm, I think it is steak tonight, and a lovely red. A cabernet sauvignon, methinks. But should it be Devon Hill? Le Riche? L'Avenir? Who knows; not me. Ah, the adventure of new wine! Once again my thanks to you Charl. I expect to be much more thankful tonight once that wine hits the palate! [ July 29, 2002, 02:17 PM: Message edited by: Wreck ]

I will want to revisit my choice for Ghost's too, probably.

I just ordered mine. Thanks for the alert!

Steve, get back to work! Seriously I think the silent majority are happy to wait, or at least, not unhappy. A little grumbling is expected; ignore it. Nothing to see here. These aren't the droids you're looking for. Move along.

Will mow for CMBB!

I'm done! What's holding all you slowpokes up?

I have been buying stuff. No problems until I get to Major Taktik's scenario. Clearly the intent here is to disallow the player things other than tanks. I assume this includes disallowing buying any "support" other than the permitted light AT weapons. But there is slight ambiguity. Many companies (if bought whole) come with mortars and/or MGs. Is it permissible to keep those? Also, as I read it the limits on arty mean that you can buy, at maximum, 2 observers, neither of which can be larger than 81mm. So you could buy 75mm observers if you wanted, yes?

Woops -- edited to remove. Sorry, Nordic guys. [ June 28, 2002, 12:03 AM: Message edited by: Wreck ]

I guess that Kreigstadt has been finished by everyone. Is that so? Anyway, that would be logical since it is small and short and the other two are long.

I have now finished two of three games, and the remaining one (vs redwolf) is on turn 34 of 40, so it will be done soon - probably this weekend. Any other games finished? Or can I taunt you all for being slow and lazy?

The only setting I care about limiting is arty. Like others, I don't care to get in an arty duel. Imagine: my naval observers against your 300mm rocket horde (or whatever) in an ME.

redwolf: We have limits. I have signed up for a tourney thinking 5 games (6 players, 1 game each), and now it is 7. That's fine. Mirrored battles require 14 games. That's too much. As for scenarios: we have 5 for ROWII. I don't see why people think scenarios are incomparable with QB-style games. But if people really think that (and I strongly disagree), then another solution is to get two more scenarios. As for me, I think that there are MANY aspects to what makes a truly titanic CM player. One is to be able to dominate in QB MEs, as is seen at TH. Another is dominate in QB attacks. Yet another is to do well in scenarios, that is, with forces you don't pick yourself and interesting terrain. And one more, even rarer skill, is to do well in a fog of war about the scenario itself, where you don't even know if it is balanced. This is what I learned in ROW and Nordic Wannabees. I think it was valuable. So that's four things I think we should emphasize in Titans. My proposed games, and scoring system, reflect that, along with the time limits we all have. I have not seen a detailed proposal from anyone else, beyond "7 QBs". But even with that: what about scenarios? What about imbalanced games? If you don't think these are important parts of CM mastery, well, say so. Let's be upfront about our biases. As for the whining issue: if we play just 7 games and declare a winner, yes, someone will probably whine. I won't, though. (Well maybe a little about that lucky gun-damage...) If the winner can't take a little whining in stride, then he's not much of a winner. And if the non-winners can't get over their loss, even if it was unfair in retrospect, I am sorry for them. Ultimately, it is still a game and I hope we can all keep that in mind.

One other point on the luck of the draw and the hypothesized "variable player skill in inbalanced games" effect. If the effect is large enough, then yes it is a problem. But if it is small it is not. The reason is simply that random factors in the games themselves will swamp it. There are many points in a game where the game result can measurably move based on a simple "die roll". For example a tank-on-tank confrontation: first kill wins big. Random factors are numerous, and expected to even out over time. Think of luck of the draw as one of them. Only if it is very large -- that is, larger than the effect of several in-game results -- are we ever likely to notice it. I don't think it is large. Actually the more I think about it, I don't think the effect exists at all with Integral Nabla. But surely if it does, it's small.

Regarding the "square law" and fairness. First, you were using the square law only *analogically*, at least originally. Your analysis of Trafalgar, the domain was in number of ships, the force available on both side. You then analogized the square law effect in military force found there to a different thing: a proposed "variable player skill in imbalanced scenarios" effect on CM play. The idea was that, if between two average players, a scenario averages 60-40 in result, then between two really great players the average result would be more skewed. Perhaps 75-25. That, in itself, is no problem for Nablas, as has been pointed out. But then you add in the variable player skill. Perhaps there is one player among us, head and shoulders above the others. If you happen to draw him in an imbalanced scenario, then you may get a disproportionately bad effect. Of course, the scoring system still takes care of it, by and large. But if you throw in a certain variability due to game mechanics, then perhaps playing a really good player is disproportionately bad in an imbalanced scenario. This is all very tenuous reasoning to me. I am not even sure such a "variable player strength in imbalanced scenarios" effect exists. Analogizing to military force proves nothing. If it does exist, I doubt it will cause any serious problems when scored with Nablas. (Though I will concede that it would cause problems with simple scoring by sums -- but then I am against that simple scoring method.) And if I wrong twice, and the problem actually exists *and* is not fully addressed by Integer Nablas, then I am willing to submit to the luck of the draw. It is not like anyone here really knows who the 800 pound gorilla is, if there is one.

On this we will just have to disagree. I don't think average score is better or worse than lots of other ways one can aggregate scores. Most of the other ways, including Nabla scoring and Integer Nabla, have the feature that the a lower average scorer can beat a higher one. In all cases, it is rare. Higher score clearly correlates with winningness. But the correlation is not absolute.

Regarding the hypothetic "square effect" of playing variable quality opponents in imbalanced scenarios. I see what Fionn is saying. I don't know that the effect is square (or nonlinear at all), but what if it is? First off, we can say that the tourney is fair going into it, as long as the game assignments are random. No player has a higher chance to be assigned to a "bad" player in any particular game, than any other player. After the assignments are out, then we might say that it is unfair. And that's bad, I guess. But at least at the highest level, the tourney as proposed is fair. Second, I note that the uneven-players-non-linear effect (if it exists) means more than just not mixing even and uneven scenarios. Rather, it means that all scenarios must have the exact same balance. A 50-50 scenario doesn't mix with a 60-40, but then neither does a 70-30. Only a set of scenarios *all* 60-40 can be fair. (That's using Nablas. Without a relative scoring system, all scenarios must be 50-50.) So if we believe in this effect, and we are not satisfied with accepting the luck of the draw in assignments, we must give up the ROWII scenarios. (Indeed, ROWII itself might be seen as suspect.) Third point. Even limiting ourselves to 7 quickbattles, we still must have them be all the same balance. Practically speaking, the only balance we are remotely likely to be able to establish for 7 battles is "even" -- 50-50. So much for uneven battles. And it gets worse: the chance that all ten of us (8 players, and TB155 and WineCape, who definitely have a say), are likely to agree on balance that is *actually* 50-50 for all battles is pretty slim. So, the only real fair way to do it would be to let the individual pairs of players negotiate what they think are fair battles, using I-cut-you-choose type principles. That would work, of course. But it limits the scope of battles quite a bit. It makes negotiating terms a key aspect of the tourney, which I don't think it should be. And worst, it makes all the battle incomparable to each other. So there won't be a lot of cross-comparison possible. My take on things: I believe the uneven-players-non-linear effect is plausible, but certainly not proven. If it does exist, I doubt it is as sharp as a square law. Perhaps n*log(n). If it is sufficiently small then the effect of player skill should outweigh it. But even if it is not small, I am willing to take my chances on luck of the draw.

Indentally, Fionn is on to a feature (which may be a strength or weakness, depending on your POV) of the Integer Nabla system with the 80-scorer example. Integer Nabla emphasizes small differences in score, turning them into something larger. A guy getting one less on score in a battle than another guy has underscored him by a small percentage -- about 1.5% on average. (I.e., if Fionn scores 60-40 in a battle and I score 59-41, I have scored 1.6% fewer points.) This is a small difference. When we plug the results into Integer Nabla, then Fionn will score at least one more point than I do. Since the total awarded for a game is 12 points (using 3-2-1-0), that means he gets a substantially higher fraction of total score awarded. Is that a bug, or a feature? I think it is a feature.

Generally correct. It is more complicated than you pose, because there are two other games. You are assuming that the other players all scored lower than the 80% guy, i.e., the scores look like this: 90-10, 80-20, 40-60, 35-65 80-20, 79-21, 40-60, 35-65 80-20, 79-21, 40-60, 35-65 In which case, yes, the guy who scored the 90 does get one less point. I don't see that as a problem, BTW. The idea that the 80-scorer is inferior reflects the idea that summed scores is some higher way true. Summed scores is OK as a system for even battles, but I see nothing special about it. I don't see that either player is superior from the data above. Note that the results might also be this: 90-10, 85-15, 85-15, 80-20 80-20, 79-21, 40-60, 35-65 80-20, 79-21, 40-60, 35-65 In which case, using Integer Nabla the guy scoring 90 outscores the steady 80 by one point. The context is important. [ June 10, 2002, 11:16 AM: Message edited by: Wreck ]

It is true that I agree that "unbalanced scenarios and balanced QBs are [not] directly comparable". But that understates it. I don't think any two different games/scenarios are comparable. Thought experiment: I scored 60-40 in a certain scenario or quickbattle. Am I better than someone who scored 75-25 on some other scenario or quickbattle? To compare any series of games, you have to fudge. Indeed, using the numbers that CM reports is already fudging. Does anyone think that two battles each ending 75-25 are the same? No, of course not. The game-reported scores will be used in awarding points for all games -- scenarios and QBs. The scoring for the QBs will not be decided on who scores the most points. They will be decided on which attacker scores the most points (and conversely, which defender).

Fionn: I appreciate that you "get" Nabla. Good! Saves me endless typing. I do not believe there is any particular scoring system that is related to "types" of games. Integer Nabla is the best system for what we are talking about doing. But it would also be the best (IMO) for 7 vanilla QBs. And it would also be best for 7 scenarios. So, I disagree when you say we are playing half the games "under one system" and half under another. We are playing all under one system: Integer Nabla. (At least, that's what I propose to do.) All of the games -- including QBs -- may be imbalanced. Some will be imbalanced intentionally. But they all might be imbalanced unintentionally. And that includes QBs. Attack/defense is going to be imbalanced unless we negotiate the params to death, something I am not eager to do among eight players (it is hard enough with two). And even then, we might get it wrong! Or, Treeburst might generate a map, completely innocently, in such a way as to make the scenario unfair. There are many ways imbalance could happen in a scenario that we (or a majority of us) think is even. So I take exception to your statement that "the... balanced QBs does NOT require NABLA scoring but a simple additive process". Only if they are truly even can they avoid needing Nabla. And we cannot know if they are even. We will probably have a good idea, yes. So they are not likely to be more imbalanced than say 60-40 or 55-45. But even that is too much. Meanwhile, using Nabla with one or more truly even battle(s) does not hurt. There is only one real downside to using Integer Nabla. And that is, that if you have some truly even (50-50) scenarios, then you can compare all the players to each other, not just all the attackers to other attackers and all the defenders to other defenders. You can extract slightly more information per battle about who is best. I see the advantage, but I think the practical world considerations outweigh it completely. If we have even one battle that is really 55-45 but we think is balanced, for whatever reason, then Nabla is superior. So, there are a couple of major downsides if we use non-Nabla. There is little downside to using it. That's my view on it.

Swamp: I know what you are saying about playing all QB masters. That is, in fact, a means in which the system could be unfair (in a sense). Like, I hope not to meet you there. But the bottom line is that we must have 7 battles in order for everyone to go head to head at least once with everyone else. (This is not absolutely necessary, but it *is* nice.) This might be a good reason to avoid MEs, which I would expect TH players to dominate. But then again, I think mopping up at MEs is part of what makes a good CM player. I would think it a poorer contest to only do QBs (or QB-style), or to only do scenarios. I think mastery of all types of battles is part of what we think of when we talk about titans of CM. Incidentally, for those worried about imbalanced scenarios, there is not likely to be *that* much of it. There will be at least one of the new scenarios that is imbalanced; possibly others. We don't know and they are intentionally not telling, to keep the spice in things. But given that these are scenarios designed and tested by B&T, I expect several should be very nicely balanced already. And where there is imbalance, I expect it is not too great -- i.e., a 60-40 median sort of battle, not a 90-10. Imbalanced QBs are possible and I think we should play some (i.e. normal old attack/defend is, IMO, imbalanced). But this is subject to negotiation.

Not true. Each of us plays each scenario. And while I think you are getting how Nabla's works, you are still not understanding how "integer Nabla" works. There is no uniform advantage to be gained. If it is true -- and I am not sure it is -- that even a small imbalance will snowball, it will be true for all four games played of the particular scenario. Under integer Nabla, the four "victors" (guys with the imbalance favoring them) have 6 points to split among themselves. 3-2-1-0. The four "losers" also have 6 points. So the "loser" who does best at preventing the snowball effect is, in fact, a winner -- 3 points. The "victor" who does worst at exploiting his advantage will get zero points. In short, it DOES NOT MATTER if a particular scenario is balanced or not. What matters is, that it is the SAME battle played by all 4 pairs of players. If you want me to run through how it works, I am happy to. With examples. No. There is no difference between a balanced battle and an imbalanced one. All that matters is the *relationships* between the scores of the four players on each side of a battle. This is, in fact, a huge strength of the system. Say that TB155 happens to generate a map that is not fair for an defense, say. We all play on it, and all four defenders discover an unassailable position, and they all "win" the scenario. Is that a problem for us? No. They are rated against each other (the other defenders that is), not against the attackers (per se). Of course, the points the attackers are getting are related to those the defenders get. Whichever defender gets 3 points, his attacker gets 0. The defender getting 2 points, his attacker gets 1. Etc. Again: using the integer Nabla scoring system, we don't need balanced scenarios. If some are balanced, fine, no problem. If not, no problem. We don't need to worry about it, and we don't have to design the scenarios to be balanced.

Regarding what the players should know about the map, again I say: keep it QBish. We should know the size of the map, the terrain/hilliness/etc, the weather, the date. Basically, we should be able to use the editor to autogenerate maps so we know what we are dealing with. The referee should then autogenerate maps until he gets one that is more or less OK, then maybe tweak it in relatively small ways that don't invalidate the overall character of the map, but perhaps add some tactical challenge. So, don't stick in a new river, but maybe remove a too-well-placed building, say. Or add one hill in someone's setup to even out a hill on the other side. Or move the flags to fairer locations. That sort of thing.

Attack:defend points are 3:2, in QBs. And though it is true that we might have any point ratio because of the referee-buying, I think staying with 3:2 is appropriate since QBing is an important part of what I would consider CM mastery, while playing with weird point ratios is not. And we are already going to get some weird ratios anyway, in the four ROWII scenarios. Similarly, I think I would prefer to limit player buying to full units; that is no cherry-picking within companies, battalions, etc. I see no huge problem with this, except that it is very non-typical of PBEM or ladder play. At least, in my experience. If anybody thinks it is important, let him voice the thought! Another reason to play 3:2 and quickbattleish would be as an empirical test of the proposition that attack/defend is easier for the attacker. Lots of good players think that, but I think the evidence is not really convincing. This tourney can help to take out the effect of player skill, which is one of the main confounding factors. So I would suggest one German attack, one German defense, and one ME. Deathdealer: 2 attack/defends, 1 ME is what I meant.

Hey, the idea of giving integer points based on relative order surely does not originate with me. Someone invented it for the ROW final; and it surely must predate that, right? Anyway, I agree with using 3-2-1-0 points. It's similar to 2-1-0 as I propounded earlier, but adding one more level because we went up from 6 players to 8. If Fionn (or anyone else) is still uncomfortable with its workings/fairness, please let's get it out in the open and discuss it. I believe it is a good and fair system, and am happy to discuss how it works. One advantage to using 3-2-1-0 over Nabla's, is that we will not have to wait for the end of ROWII to compute score. With 8 presumably very dedicated players, we are likely to finish before all 72 of the ROWII fellows. Regarding neutral-flag handling: if both redwolf and Fionn think proportional adjustment is right, it's right with me. So I will vote for that if we need to vote. Regarding the three choose-your-own scenarios: I am all for choose your own, but I would prefer if we could agree to "fix" as many aspects of the battles as possible rather than have them all negotiated. This makes them more comparable to each other. It also means the effect of terrain on the battle will be fairer. If nothing else, I think we should fix the type of each battle. I suggest two attacks and one ME. The terrain obviously matters a lot with this. Ghost: ROWII has its own thread, easily found on the front page of this forum.