SGOTM 06 - Smurkz

[LowtherCastle]

If you know that jungle growth can't happen once the farm is built, then I don't understand why you say that it's a problem throughout the game.

I said sooner or later, not throughout. I'm confident you know the difference.

 

[ChrisShaffer]

But "later" isn't relevant in the context of this discussion, is it?

 

[LowtherCastle]

But "later" isn't relevant in the context of this discussion, is it?

You guys are funny. Of course, I'm referring to sooner or later in the early game, when it counts. I'm not just blabbing away here. Of course jungle creep doesn't happen every early game, but SGs don't happen every month either, do they?

Okay, let me just go back to my point, which is that I prefer to avoid the risk of jungle creeping onto a resource that I'm counting on in the early game, if at all feasible. Instead of trying to (fruitlessly?) re-explain my point, permit me to quote akots and Balbes:​

Remember, teams who play it safe don't win - they finish in the top half. The winner is somebody who took a risk and it paid off. (Of course the bottom half are those who gambled and lost.)

Top half sounds OK to me. I don't think that any gamble is worth a risk of losing the game. As I said previously, I try always to play the starting moves risk-free and not rely on chance and luck here. Let's try to discuss this on a rational basis and tomorrow when the heads are more clear and fog of opportunity can be replaced with a clear picture of the future and desire to gamble does calm down a little bit.

In my book, risk-free means just that: risk-free. Not almost risk-free.​

Remember, teams who play it safe don't win - they finish in the top half. The winner is somebody who took a risk and it paid off. (Of course the bottom half are those who gambled and lost.)

Let's try to discuss this on a rational basis

But this is on a completely rational basis I'm not preaching along the lines of "better lucky than good". What I mean to say one must be good and lucky in order to be gunning for the 1st place, which I believe we're trying to do. Suppose there's 20 teams and 10 of them chose to pass up the opportnity. Of the other 10 who gambled, 2 won and 8 lost. Now you have 2 teams doing great, 10 doing so-so, and 8 probably out of the race. You need to be one of the 2 successful teams. Personally, I hate taking this kind of risks, but there's no way around that, or the other guy who didn't hesitate is going to pull ahead. To sum it up, you must be willing to jump at opportunities, you must be lucky enough to have it pay off every time, and you must play skillfully in between.

OK, here we disagree. This is a non-reload game and we cannot know whether we succeed or not. Also, I don't think you have to be lucky to win this competition. I do belive that to win, we have to play like a team and discuss every move and try to convince the other team members that what is planned is a good thing to do and not just someting that we have to do in order to win. Believe me, I have played a lot of SGOTMs, sometimes won, sometimes lost but never gambled on something which cannot be properly controlled.

To be clear, akots is discussing not hoping for 'good' luck and I'm suggesting that you guys were hoping for not 'bad' luck,* but the point remains the same. Evidently, I side with akots and you guys side with Balbes on this point.

*Assuming any of you guys even considered the very small possibility of jungle creep, which is not evident from your thread.​

Bottom line, you guys played what I consider a risky strategy and you got snake-bit. I play most of my GOTMs risky and get snake-bit most of the time....

 

[unkle]

Hey guys, I do not think we disagree *that* much right ? Only think that really matters is to know when you start thinking you are gambling... Gambling is sending an early settler without too much exploration for instance. Or moving away from starting point *without* info,...

To me, you are not gambling when chance is ~1%. That's all But I do see your point. If risk-free is really risk-free, then you cannot do much unless protected by several warriors (bears do happen,...), etc. That's all. And I'll definitely play the same again.

 

[ShannonCT]

Yes, I realize that. ONe thing I don't know, though, is whether jungle creep is prevented by a worker actively building a farm. Do you know? Because I think you guys were just two or three turns from starting the farm, weren't you?

Jungle growth is blocked by simply starting a productivity-enhancing improvement on a tile. So the probability of jungle creep on the rice tile should be zero after your first worker is built.

If the probability of jungle creep on the rice was indeed 1%, it would be foolish to not utilize a good starting spot for fear of this risk. You can't win a tournament playing with the fear of very unlikely events.

At Xteam, we were stung heavily by another unlikely event. Our initial capture of Carthage (Hannibal's only city) was thwarted despite being >98% to take it. Had we not had that bad luck, we would have been the first team to clear our starting continent, and would have gotten to Astro a good bit sooner.

I can think of at least one other team that had a painful roll of the die (RMB).

 

[Niklas]

@LC:
I agree with akots too! Balbes is arguing that you should take a gamble that is expected to pay off 2 times out of 10. I would never argue that you should take such risks in a game like this.

It is totally impossible to play a risk-free game. You always run the risk of losing a fight even if your odds are brilliant. The team that brings in 3 attackers against 3 defenders and hope to win, when each attack is 50-50, is reckless and gambling. The team that brings 5-6 attackers against 3 defenders are probably close to the mark in terms of optimal outcome over the whole game. The team that brings 15 attackers against 3 defenders to get the chance of losing below 1% (I didn't do the maths) are no less stupid than the ones who brought only 3, and they will not win the game.

In our situation, we needed the rice to stay clear for <25 turns, which (by maths others have presented) means 1% risk over the whole game. The team that discards a great site for the capitol because of a 1% risk it might go awry are IMO just as stupid as the ones bringing 15 units in my example above. I can agree that there are other good replacement sites that may have been equally good, that's an orthogonal discussion. But to discard a site because of that 1% risk, no, I don't buy that argument, and like everyone else in our team has expressed, we would have done the same again and again.

You are right that we did not consider jungle creep when we decided where to settle. This is not because we were stupid, or not knowing about it, but simply because it was such an unlikely thing that it didn't even cross our minds. At least for me that's true, and I'm pretty sure that goes for the rest as well. And if in the next game we end up with a resource next to a jungle, and with this game fresh in mind, I'm sure we would still settle next to it and take that 1% risk.

To answer some of your comments in your earlier post: I totally agree that a game completely without randomness would be rather boring. The question is in what guise that randomness comes. There is good randomness, and there is bad randomness, I doubt even you would say otherwise. Clearly Firaxis have, as I argued before, realized this and taken measures to steer most of the randomness into the good category. I claim that they have failed in the cases I have listed.

But to be specific about your listed examples; the actions of the player are never never completely random, unless you literally flip a coin. The direction you send your scout, where you place your towns even before knowing resource locations, all such things can to a very large extent be affected by the player. They are perfect examples of the kind of good randomness (or not really randomness) that I think the game should contain. The actions of the AI are also not completely random, indeed if the AIs were programmed to act randomly we would have a boring game. We need predictability, and we need to be able to make educated choices based on that.

The case you list with a team popping a key tech from a hut, that's bad randomness too. And I think the GOTM staff have learnt something from that, since we never see huts close to the start any more. That's a very good move by the staff, and I'm asking something similar to be done for the cases of randomness I've listed here.

To repeat myself: a random event that is too unlikely to be accounted for, totally unaffectable, and has a large impact on the game, can IMO not be considered anything but a design flaw.

I'm also not arguing this for the sake of our game here and now (regardless of what I heatedly said when it happened). We played to the end, got the end date we got, and that's it. I'm arguing for the sake of future games, where I want as fair a competition as possible. I don't care if it happens to our own game or to some other team's, I strongly dislike it regardless. I wasn't particularly happy about popping gems in my capitol's BFC in BOTM02 for instance, that just felt cheesy. I truly think this is a design flaw in the game, and one that could and should be remedied by the GOTMs, just like SGLs were removed for Civ3 GOTM.


@ShannonT:
As I noted in my original post, I would argue that some randomness is "good" from a design perspective, while some is not good. We had pretty lousy luck failing to capture Athens, losing 2 elephants who each had >60% chance, without doing a single hp damage. But that kind of randomness is still something we can affect ourselves, so it's not truly completely random. We can choose to bring in one more unit to be safer. Or we can play it slightly less safe, as we both did, knowing that amortized over the course of the whole game we will likely win time by it rather than lose, even if we lose one crucial fight. And over the course of a whole game, all teams are bound to have some bad battle luck. We're not upset about our battle luck at Athens - c'est la vie. What we're upset with is a single, totally random event (totally meaning we could have done nothing to affect it), with ~1% probability (meaning it would be stupid to try to account for it), and with a direct impact of ~15 turns (David's estimate, I think it's more) on our end date.

 

[ShannonCT]

@ShannonT:
As I noted in my original post, I would argue that some randomness is "good" from a design perspective, while some is not good. We had pretty lousy luck failing to capture Athens, losing 2 elephants who each had >60% chance, without doing a single hp damage. But that kind of randomness is still something we can affect ourselves, so it's not truly completely random. We can choose to bring in one more unit to be safer. Or we can play it slightly less safe, as we both did, knowing that amortized over the course of the whole game we will likely win time by it rather than lose, even if we lose one crucial fight. And over the course of a whole game, all teams are bound to have some bad battle luck. We're not upset about our battle luck at Athens - c'est la vie. What we're upset with is a single, totally random event (totally meaning we could have done nothing to affect it), with ~1% probability (meaning it would be stupid to try to account for it), and with a direct impact of ~15 turns (David's estimate, I think it's more) on our end date.

It turns out, I've discovered from testing, that you could have prevented jungle creep on the rice tile by placing your warrior there. (I have no idea how to verify this from the code itself, but my testing has convinced me that it's true.)

Is that case, the "bad randomness" you suffered was completely preventable. The tradeoff, of course, would be not being able to use your warrior for exploration. But it's now a tradeoff like one would make if one brought an extra unit to make the probability of winning a battle 99.99% instead of 99%.

 

[Niklas]

It turns out, I've discovered from testing, that you could have prevented jungle creep on the rice tile by placing your warrior there. (I have no idea how to verify this from the code itself, but my testing has convinced me that it's true.)

How much testing? I could almost swear this was not the case, since I have strong memories of having a fog-busting warrior on a hill suddenly "grow a wall of trees". I could certainly be wrong though. Checking in the code would be nice, but I've already tried to dig there and failed.

Is that case, the "bad randomness" you suffered was completely preventable. The tradeoff, of course, would be not being able to use your warrior for exploration. But it's now a tradeoff like one would make if one brought an extra unit to make the probability of winning a battle 99.99% instead of 99%.

I agree. If it is indeed the case that placing the warrior there would have prevented the spread, then it changes things. It's no longer unpreventable (unless it happens before the warrior could get there of course). But still, I don't see any team that truly would do that. The cost of lost exploration is pretty dire to pay for that 99.99 instead of just 99. Sort of like the team bringing 15 units when 5-6 will do. For that reason I still think jungle creep is "bad" from the design POV, and I still want to see it remedied in some way. But your argument is certainly valid.

 

[ShannonCT]

How much testing? I could almost swear this was not the case, since I have strong memories of having a fog-busting warrior on a hill suddenly "grow a wall of trees". I could certainly be wrong though. Checking in the code would be nice, but I've already tried to dig there and failed.

I placed three cities in the jungle, cleared some jungle so that 2/3 of the land in the BFC was jungle and 1/3 was grassland. That's ~40 jungle tiles and ~20 grassland in the BFCs. I then set down 6 warriors on grassland tiles. Then I hit enter 180 times. 10 of the grassland tiles without warriors turned to jungle. None of the grassland tiles with warriors turned to jungle. Based on that, it's highly unlikely that jungles can grow under the feet of warriors. You're welcome to test more and try to prove me wrong.

The game is probably coded differently for forests, since forest growth is considered a good thing in most situations. But it seems that the designers designed the game to allow players to eliminate the kind of bad luck you experienced. It's just one of the hundreds of little game features that not many people know about (including me until just now).

I agree. If it is indeed the case that placing the warrior there would have prevented the spread, then it changes things. It's no longer unpreventable (unless it happens before the warrior could get there of course). But still, I don't see any team that truly would do that. The cost of lost exploration is pretty dire to pay for that 99.99 instead of just 99. Sort of like the team bringing 15 units when 5-6 will do. For that reason I still think jungle creep is "bad" from the design POV, and I still want to see it remedied in some way. But your argument is certainly valid.

You could have moved your warrior onto the rice on the same turn that you founded London, so the creep was completely preventable. And keeping the warrior in your borders was a decent option anyway with it being an Always War game. Since you guys were planning to go Hunting->Archery after Agriculture, you weren't going to use your warrior to search for the nearest AI to rush. Granted, it's still nice to get exploring as early as possible, but you could have scouted a good second city location before your settler was finished if you had delayed scouting until after your worker had started on the rice.

 

[DaviddesJ]

Based on that, it's highly unlikely that jungles can grow under the feet of warriors. You're welcome to test more and try to prove me wrong.

Note, you can make such testing easier by increasing the jungle growth rate in Civ4.ini.

The game is probably coded differently for forests

I would give very long odds against it.

 

[Niklas]

I placed three cities in the jungle, cleared some jungle so that 2/3 of the land in the BFC was jungle and 1/3 was grassland. That's ~40 jungle tiles and ~20 grassland in the BFCs. I then set down 6 warriors on grassland tiles. Then I hit enter 180 times. 10 of the grassland tiles without warriors turned to jungle. None of the grassland tiles with warriors turned to jungle. Based on that, it's highly unlikely that jungles can grow under the feet of warriors. You're welcome to test more and try to prove me wrong.

That's certainly a rather thorough test, and I will agree that you likely have it right. Jungles don't grow on trampled soil. That's very good to know.

The game is probably coded differently for forests, since forest growth is considered a good thing in most situations. But it seems that the designers designed the game to allow players to eliminate the kind of bad luck you experienced. It's just one of the hundreds of little game features that not many people know about (including me until just now).

I gotta agree with David that it's unlikely forest spread is coded differently. It's more likely that I remember wrong, wouldn't be the first time. But I agree with your conclusion that the designers have given us a way out, kudos to them for thinking that far.

You could have moved your warrior onto the rice on the same turn that you founded London, so the creep was completely preventable. And keeping the warrior in your borders was a decent option anyway with it being an Always War game. Since you guys were planning to go Hunting->Archery after Agriculture, you weren't going to use your warrior to search for the nearest AI to rush. Granted, it's still nice to get exploring as early as possible, but you could have scouted a good second city location before your settler was finished if you had delayed scouting until after your worker had started on the rice.

All true for this game, but in the general case? I agree though that this makes jungle creep a lot less dangerous, to the point where most of the design badness of this random event goes away. Most, but certainly not all. I no longer feel so strongly about it, though I'm still leaning towards nerfing it, as I said in my last post.

EDIT: I also suddenly remember the staff's seeming preference for removing the warrior for Challengers. Suddently that has a lot more implications...

I still don't see any redeeming features with popping a mine resource out of the blue though. What do the rest of you think about that?

 

[DaviddesJ]

I agree with your conclusion that the designers have given us a way out, kudos to them for thinking that far.

I'm sure they didn't code it this way in order to create a strategy of parking units on tiles to avoid jungle growth. They just wanted to avoid the awkward case where a jungle grows on a tile while a worker is improving it, or about to improve it. It's easier and simpler to check whether a unit is there, than to check whether an improvement is currently being built. That's what I'd code (which is why I suggested this as the likely implementation in the first place).

 

[DaviddesJ]

I still don't see any redeeming features with popping a mine resource out of the blue though. What do the rest of you think about that?

I like randomness. As I've said many times before, 99% of Civ4 games are not "competitive", and so I think the designers should give essentially zero weight to such concerns. But, even aside from that, I personally think the random element is just fine in the "competitive" games as well.

I think you could set the probability of discovering mine resources (and jungle growth) to zero in the GOTM mod, if you wanted.

This randomness that you're worried about is all way less than the random generation of leaders in Civ3, by the way.

 

[TDK]

I'm sure they didn't code it this way in order to create a strategy of parking units on tiles to avoid jungle growth. They just wanted to avoid the awkward case where a jungle grows on a tile while a worker is improving it, or about to improve it. It's easier and simpler to check whether a unit is there, than to check whether an improvement is currently being built. That's what I'd code (which is why I suggested this as the likely implementation in the first place).

In SGOTM2 I had this happen. A worker was just about to complete a rice paddy, but I needed to complete a road on an adjacent tile, and in that single turn a jungle crept onto the rice. So it's not just about having turns put into am improvement, you actually have to have a unit there.
I was under the impression that forest/jungle growth was more likely in the early turns of the game, due to more years passing per turn. Is this wrong?
My experience in SGOTM2 and my belief that jungle growth is more likely in the early turns(right or wrong), would make me reluctant to settle that spot. That being said, in this game, I didn't even see the possibility of doing so.
I agree that it was the best spot to settle if we disregard jungle creep.

Good game Smurkz!

TDK

 

[unkle]

I never thought jungle/forest growth would be linked to number of years rather than number of turns.
I tend to agree that we see it much more early in the game. Which could be due to:
* jungle growth rate is based on years
* we clear forest jungle so regrowth is less likely (missing seeds somehow)
* this is something that matters us only in the early years (so our perception is biased)

So since I estimated the ~1%, I may be wrong on the number (would doubt it would be much bigger though).

In all case your SGOTM2 experience is something very interesting to me...

 

[Gyathaar]

To save you from some work...

This is the code in BtS (didnt check if it has changed much from warlords)

Spoiler :

void CvPlot::doFeature()
{
	PROFILE("CvPlot::doFeature()")

	CvCity* pCity;
	CvPlot* pLoopPlot;
	CvWString szBuffer;
	int iProbability;
	int iI, iJ;

	if (getFeatureType() != NO_FEATURE)
	{
		iProbability = GC.getFeatureInfo(getFeatureType()).getDisappearanceProbability();

		if (iProbability > 0)
		{
			if (GC.getGameINLINE().getSorenRandNum(10000, "Feature Disappearance") < iProbability)
			{
				setFeatureType(NO_FEATURE);
			}
		}
	}
	else
	{
		if (!isUnit())
		{
			if (getImprovementType() == NO_IMPROVEMENT)
			{
				for (iI = 0; iI < GC.getNumFeatureInfos(); ++iI)
				{
					if (canHaveFeature((FeatureTypes)iI))
					{
						if ((getBonusType() == NO_BONUS) || (GC.getBonusInfo(getBonusType()).isFeature(iI)))
						{
							iProbability = 0;

							for (iJ = 0; iJ < NUM_CARDINALDIRECTION_TYPES; iJ++)
							{
								pLoopPlot = plotCardinalDirection(getX_INLINE(), getY_INLINE(), ((CardinalDirectionTypes)iJ));

								if (pLoopPlot != NULL)
								{
									if (pLoopPlot->getFeatureType() == ((FeatureTypes)iI))
									{
										if (pLoopPlot->getImprovementType() == NO_IMPROVEMENT)
										{
											iProbability += GC.getFeatureInfo((FeatureTypes)iI).getGrowthProbability();
										}
										else
										{
											iProbability += GC.getImprovementInfo(pLoopPlot->getImprovementType()).getFeatureGrowthProbability();
										}
									}
								}
							}

							iProbability *= std::max(0, (GC.getFEATURE_GROWTH_MODIFIER() + 100));
							iProbability /= 100;

							if (isRoute())
							{
								iProbability *= std::max(0, (GC.getROUTE_FEATURE_GROWTH_MODIFIER() + 100));
								iProbability /= 100;
							}

							if (iProbability > 0)
							{
								if (GC.getGameINLINE().getSorenRandNum(10000, "Feature Growth") < iProbability)
								{
									setFeatureType((FeatureTypes)iI);

									pCity = GC.getMapINLINE().findCity(getX_INLINE(), getY_INLINE(), getOwnerINLINE(), NO_TEAM, false);

									if (pCity != NULL)
									{
										// Tell the owner of this city.
										szBuffer = gDLL->getText("TXT_KEY_MISC_FEATURE_GROWN_NEAR_CITY", GC.getFeatureInfo((FeatureTypes) iI).getTextKeyWide(), pCity->getNameKey());
										gDLL->getInterfaceIFace()->addMessage(getOwnerINLINE(), false, GC.getEVENT_MESSAGE_TIME(), szBuffer, "AS2D_FEATUREGROWTH", MESSAGE_TYPE_INFO, GC.getFeatureInfo((FeatureTypes) iI).getButton(), (ColorTypes)GC.getInfoTypeForString("COLOR_WHITE"), getX_INLINE(), getY_INLINE(), true, true);
									}

									break;
								}
							}
						}
					}
				}
			}
		}
	}
}

this is the important section:

	if (!isUnit())
		{
			if (getImprovementType() == NO_IMPROVEMENT)

so as long as there is no unit on the plot, a feature can grow there if there are no finished improvement of the plot

 

[LowtherCastle]

Okay, but what about the likelihood on any given turn that the jungle grows on that plot and does it have naything to do with jungle on adjacent or cornering tiles?

When I did a test with 160 rice tiles and hitting enter for 100 turns, with jungle on two adjacent tiles and all four cornering tiles, I got a 1/15 chance of jungle growing on the rice in 20 turns.

 

[Gyathaar]

jungle has a 16% growth probability...
the chance increase for each of the tiles it borders in north, south, east and west (not northwest, etc)
after this the chance is raised by another 25% of the original chance (and then dropped to half if the tile has a road)
This chance is measured in 1/100 percent..

chance of jungle growing in any turn surrounded by n jungle tiles (n = 0,1,2,3 or 4) and no road or units on tile:

probability = 16 * n *1.25 * 0.01%

so when surrounded by 4 jungle tiles :
16*4 * 1.25 * 0.01% = 0.8% per turn

surrounded by 2 jungle tiles :
16*2 * 1.25 * 0.01% = 0.4% per turn

chance of jungle growing in a tile surrounded by 2 other jungle tiles (assuming no jungle growth in surrounding tiles) over 17 turns:

(1-(1 - 0.004)^17) * 100% = 6.59%


chance of jungle growing in a tile surrounded by 4 other jungle tiles over 20 turns:

(1-(1 - 0.008)^20) * 100% = 14.8% chance (or 1 in 6.7 )

(unless I misread something in the code)

 

[DaviddesJ]

That's a lot higher than I thought, but it seems consistent with the code.

By that calculation, we had ~10% chance of jungle growing on the rice, before our worker got there. I think this is still a chance well worth taking, but certainly, in hindsight, it seems worth garrisoning the tile with the warrior.

 

[LowtherCastle]

Well, it surely shows the complexity of assessing the optimal starting position and research path. Is your starting position better if you have to sacrifice early exploration to protect the rice? Interesting question.