Subject: A Game of Chancelast
Pages: 1

Messages / 1 to 17 of 17

As I've mentioned in the boards in the past, GT's attack algorithm applies a small random factor to the strength of units. The attack value dips lower than the defend value, so defending has a slight advantage in the random adjustment.

Though the random factor can sometimes change the outcome of attacks (more so on the higher end of strengths, just based on percentages), in the end, it means the attack system is mostly predictable. Plotting the results of a large sampling, here are the results of an attack of strength 3 (top-left), attack strength 7 (top-right), defend strength 3 (bottom-left), and defend strength 7 (bottom-right):

user image
This means, of course, that a unit of strength 3 could never overtake a unit of strength 7 (of the same type):

user image
I'll soon be creating a test world that will use an alternate attack algorithm. Rather than a slight random variation, it's going to adjust the strength value using a normal distribution (a natural bell curve), still adjusting slightly for an advantage for defending over attacking:

user image
So, this means it will be entirely possible (though still less likely) for a unit of strength 3 to overtake a unit of strength 7:

user image
Stronger units will still overall be more likely to win, but this will definitely add a lot more chance and unpredictability to the battles. The battles will depend more on how well your particular units perform in their battles that day than just who can build the strongest units.

To some degree, I'm sure one can argue that it even adds a bit more realism to the units. Down the road, I could even see adjusting the game so that each unit now has an experience level, where new units have wide bell curves and battle victories increase experience and tighten up the bell curve a bit.

Once the alternate Game Cycle is ready to go, I'll likely be participating in the first test world to some degree so I can see how well it works firsthand. (I also need to record various videos to use on a new Global Triumph blog I'll be launching soon, so participating in a test world would be a perfect opportunity.)
Happy Easter to everyone!!!

I built a page in order to develop and test the chance attack algorithm in the Game Cycle.

I decided to add some explanatory text and open it up to TMS users. So, if any of you would like to see, test, and/or provide feedback, please feel free to take a look!

(link removed; page disabled)

Right now, I'm leaning towards "Average" for the method and "Postprocess" for the bias. If you have any feedback, please feel free to share it here or send me a note.
I've created a Development Sandbox world that has the new "Gaussian Attack Algorithm" mode enabled.

A few notes:

- I set this up as a development world rather than a normal more with alternate settings because it's possible the algorithm or settings will be adjusted as the world progresses.

- I'll have a country in this Development Sandbox so I can run tests and see how it functions.

- In addition to the Gaussian curve adjustment, the new mode also adjusts how armor level is determined when a defense turret shares a sector with a land unit. (The normal algorithm used the turret's armor level, whereas the new mode uses combined armor levels. For example, a defense turret with a tank yields an armor level of 7 rather than 4.)

- The deviation of the Gaussian curve used to decide an attack strength varies per unit. If anyone is curious, here are the attack power and deviation values.

Bases: 1000 (strength) / 30

Defense Turret: 70 / 70

Construction Truck: 30 / 30
Infantry: 20 / 40
Jeep: 40 / 40
Tank: 90 / 90

Transport: 50 / 50
Carrier: 90 / 90
Missile Frigate: 90 / 90
Warship: 90 / 90
Warship Fire: 120 / 90

Jet: 40 / 50
Missile: 300 / 100
Bomb: 2000 / 100

Again, these values may change as the game progresses (based on feedback).

Please note that I'm considering makes this algorithm be the official attack algorithm for Global Triumph, making the game less predictable, perhaps more realistic, and hopefully more interesting. So, player feedback will be very important!
I'll help out.
Thanks, Rick and Xyhelm!
Yea!  Looks mathy.  Let's see what happens.
Here is an example of what could happen when an infantry of strength 9 attacks a tank of strength 5:

user image
Though the thank is still more likely to get a value that's higher than the infantry, if the tank happens to get low and the infantry happens to get high, it's possible that the infantry will win. (It's certainly possible that a tank operator could have an off day, or that the infantry are fighting well and being particularly innovative!) Or, even if the infantry doesn't win, it could do more damage than it normally would.

Of course, the opposite is also true, in that the tank could get an even higher value and the infantry just perform really badly in that battle, where the infantry 9 doesn't even put a dent in the tank unit!
In the Development Sandbox, here is what I have noticed so far....

Under the normal combat rules, when 2 jets attack a level 1 infantry, the infantry dies and no jets die.

Under the new rules, I noticed that a minority of the time, the level 1 infantry will beat the 2 jets and both jets are lost.  I might understand this when there are 3 infantry (which might have been the case but I don't think so).  Loosing $1600 worth of jets to $100-$300 worth of infantry is REALLY discouraging--even unfair.  I feel like I lost more money in jets attacking low level infantry than they spent on those low level infantry.

So far, since the random factor is much higher with these combat rules, infantry is the way to go.  If it's possible for level 9 infantry to pound a level 9 tank really hard (and I know the opposite is true too), one might as well pay the lower price for infantry.  In short, when risk is high, invest little.  Seriously.
Well done, sir.  I resign.

I approve of the algorithm except for one big thing I didn't like.  So I don't think it should apply to air units.  I put details in my previous post.
Sorry, Xyhelm! I somehow completely missed your post about the jets!

You're right, I found using them to be too random. Honestly, I'm not sure why I set such a high deviation for them, especially since I dropped the deviation for the other air units so they wouldn't be as random. I'm thinking it may have actually been an error! I'm going to change that 50 to be a 10 (and missiles to be 50).

I wish I had seen your previous post sooner, so we could have tested the adjustment. I may just reset the Dev Sandbox and give it another test.

Overall, I think I do prefer the Gaussian Attack Algorithm mode to the regular mode. The game generally played the same, so it's not too extreme a change. But, there were definitely times when attacks didn't play out as expected, which I think made it more interesting and less predictable.

I like that infantry play a bigger role in the Gaussian algorithm, too. You can spend less and take more of a gamble, but going against stronger units is still difficult to overcome.

So, I think I may switch that to be the standard attack algorithm for the game. (Perhaps I'll do that after another test run, though. Haven't decided yet.)
The Development Sandbox has been reset.

The jet and missile deviations have been adjusted, so this will be another test of the new attack algorithm.
Coolness.  We are in agreement.  I think the new algorithm will be great moving forward.

I love crunching the numbers and knowing all the rules behind how combat works.  I'm confused on how armor is applied in combat.  I hope someone will make a new spreadsheet like this one.

Life happens.  So I'm not able to join you in the new sandbox.  Laterz.
I'd never seen that spreadsheet before! Pretty cool that that exists! I wonder how many players know about it and use it for reference.

It does support switching to this new algorithm, though. Definitely less predictable!

The defending object's armor plays a factor in reducing the attacking object's strength. I don't think I ever got into specifics because I didn't necessarily want players to have it all down to an exact science. haha

I may release a more detailed and specific breakdown after the switch to the Gaussian algorithm, though, since that variation makes things less predictable anyway.
I'm going to have to review and make some more adjustments. I forgot I was supposed to be testing the air attacks more than anything, so I set a mass attack of strength 1 jets against level 1 infantry and nearly every one of them lost. That seems fairly unrealistic. haha

The Gaussian algorithm may be too extreme for air attacks, since their damage is cumulative.

For example, if a tank is charging through a line of enemies, each attach is considered distinct. So, if it starts at strength 3, and then is 2.5 after the first attack, it's saved as 3 for the next attack. If it's 1.9, then it's saved as 2. With air units, though, it treats that flight there (and back, for jets) as a single attack, so all of the fire they take on the way wear down the strength.

With the Gaussian algorithm, if any one of those flyover attacks swings badly for the air unit, the damage is done!

I turned it off in the new world for now until I get a chance to review and make adjustments.
I am going to voice my opinion against this new algorithm.  After seeing intense combat in WWI I can say from my perspective it makes the game unplayable.

The algorithm is to unpredictable, and with the added randomness of who moves first, combat is too much like rolling dice blindfolded. Combat will just come down to whoever gets "lucky" during a cycle.

I enjoyed this game because of the once a day cycle, the variety of maps, and the predictability of combat.

When you are spreading yourself thin it helps to know that three Level 9 Infantry will beat one Level 9 jeep. Every time.

Once the randomness starts and not only do you lose all three infantry but the Jeep is still a Level 4 the game becomes unplayable.

This is just my opinion based on my experiences. But I will not be playing any additional maps that use the new algorithm.
I have to say, I'm with Hanibel here.
I'm more of a fan of the game Diplomacy than I am of Risk.
I also feel like the new algorithm makes it most cost effective to just build masses of infantry, as thanks to the element of chance, it seems like the same dollar value worth of infantry nearly always beats the same dollar value of jeeps or tanks.

I get the appeal of the new algorithm, but I hope that going forward some maps are played using the original algorithm.
Hanibel wrote on :
When you are spreading yourself thin it helps to know that three Level 9 Infantry will beat one Level 9 jeep. Every time.
Realistically speaking, though, three infantry shouldn't beat one keep every single time, no matter what. I feel like the fighting was too predictable.

I think that also makes the game almost entirely about sector growth. Whoever expands the fastest from the beginning gets the most money and then gets the strongest units.

That's still the case with the Gaussian algorithm, but just slightly less so. In my testing, Gaussian still ended up with mostly similar outcomes, except that now there were occasional upsets. I personally liked that. For example, sometimes I'd be sure I'd get a base, but the enemy ended up keeping it. It seemed more realistic to war, in that sometimes things go wrong and you have to adjust. (You can still dump extra money in to ensure a win, of course.)

All that being said, this is the first large scale war using it. The old attack algorithm was extremely predictable, but maybe it still led to a more enjoyable game. I was going to have the survey at the end to get feedback.

The results of the current survey were 4 keep, 5 adjust, 6 revert, and 2 no preference. So, for now, I think I'm more inclined to try adjusting the deviation to see if that makes it more enjoyable to play (while still allowing for less predictability).
Page of 1
«Previous Page|Next Page»

Message Board