Alternate battle system

[LPhillips]

I see. So victory for the side with 90% or better chances is assured, but nothing else is. There can be a 3% chance (according to the system you're using) of only one victor being left (let's say you attack a strong stack with two Light Infantry and kill all but 1 unit), and that can actually occur randomly.

That is not what KGB originally described. He said that the outcomes in the 10% range were removed, leaving the middle 80%. My understanding of statistics isn't top-notch, but I think that's a gross misstatement. As described, the range of battles removed as "x" would be stated:
0%>x<10%

And only in favor of the victor. So you will see someone get a highly unlikely outcome such as annihilating a stack he barely had a chance of beating with few or no losses. I'd much rather see the system implemented as KGB described with 10% at each end of the scale invalidated. Anything in the range of likelihood from 0% to 10% would be thrown out. That would be much more balanced. In fact, it would remove almost all balance issues. It would make the likelihood of losing a battle where you had 88% chance of victory (according to the simulator) approximately 2*5/4=2.5%. Gone would be battles where you have a reasonable expectation of 3-5 of your troops remaining and you're left with a hero alone. Sure, the 10% rule protected me from losing the 92% chance battle I just fought against Negern's hero, but it does not protect me (or him; making me pay for the victory is important too) from BS results well under 10% likelihood.

Outcomes of even the smallest possible likelihood are currently still allowed; the only absolute factor is victory for the side with 90% chance taken from trial battles. I was going to ask why the battle simulator would return results like 51.7% chance of victory in an even fight, but that is obvious now given that it's a trial-and-error system. Thanks for the info. Please consider actually removing all outcomes under 10% likelihood. It needs some tinkering, but I would suggest measuring cumulative likelihood for HP losses exceeding said threshold and HP losses below it. So you would not see outrageous outcomes for either side in any battle.

[KGB]

LPhillips,

And only in favor of the victor. So you will see someone get a highly unlikely outcome such as annihilating a stack he barely had a chance of beating with few or no losses. I'd much rather see the system implemented as KGB described with 10% at each end of the scale invalidated. Anything in the range of likelihood from 0% to 10% would be thrown out. That would be much more balanced. In fact, it would remove almost all balance issues. It would make the likelihood of losing a battle where you had 88% chance of victory (according to the simulator) approximately 2*5/4=2.5%. Gone would be battles where you have a reasonable expectation of 3-5 of your troops remaining and you're left with a hero alone. Sure, the 10% rule protected me from losing the 92% chance battle I just fought against Negern's hero, but it does not protect me (or him; making me pay for the victory is important too) from BS results well under 10% likelihood.

Thanks for the info. Please consider actually removing all outcomes under 10% likelihood.

You can't do what you describe. That is remove any particular outcome that's <10%. The reasons for that are:

1) It's possible you'll remove outcomes in the middle. For example take a stack that consists of Dragon/Bat/Hv Inf fighting 4 Hv Infantry. The strengths look like 10,3,5 vs 3,3,3. What happens is that the bat's chance of winning is <10% simply because it's MUCH more likely you'll win on the 5 strength Hv Inf or 10 Str Dragon. So you would never have a possible win on the bat. Only on the Hv Inf or Dragon thus leaving a hole in the middle of your outcomes. That makes no sense because it says the bat is dead meat in this battle no matter what.

2) In large battles (8v8 or 8vs8+ in a city) it's possible NO outcome is as high as 10% for the defender. Thus it would be impossible for the defender to win a battle they otherwise have a solid chance to win or are even favored if there are enough defenders. For example consider a Dragon+7 bats vs 8 Hv Inf. You get 10,3,3,3,3,3,3,3 vs 8 3's. My simulator says the only outcomes here bigger than 10% are Dragon (50), Dragon +1 bat (13), Dragon +2 bats (10). Even though the Defender has a cumulative 15% chance to win this battle. Even if you think the Dragon should win all the time, why should he get to keep 2 bats when it's clear the odds says they should be lost to the 8 defenders since they are equal matches and there are 8 of them vs 7 bats.

3) Fight order would totally ruin this concept. Take that same battle above. Put the Dragon first and the ONLY outcome >10% is the Dragon (25) because it's so likely the dragon wins on his own that the other units are all in the 8-9% range. Thus the Dragon stack would have to win this battle taking NO losses.

I don't think any of these things are what is desired/wanted.

So what's done is to remove the outside extreme 10%. Thus you never get screwed by losing too many men or win losing too few.

So the actual game code for battle 2 turns out to allow Dragon, Dragon+1bat,Dragon+2bats,1 Hv Inf (defenders only win).
So the actual game code for battle 3 turns out to allow Dragon+7bats,Dragon+6bats,Dragon+5bats,Dragon+4bats,Dragon+3bats,Dragon+2bats,Dragon+1bat,1 Hv Inf (defenders only win).

Thus the effect of putting the Dragon first is you can minimize your losses at the risk of your Dragon. That's exactly what you'd expect to get from changing your fight order.

KGB

[LPhillips]

I would suggest measuring cumulative likelihood for HP losses exceeding said threshold and HP losses below it. So you would not see outrageous outcomes for either side in any battle.

You missed the whole point of my suggestion. I suggested actually removing the cumulative 10% at each end, not specific outcomes of 10% likelihood or less. You keep saying that is what is done now, but it's not. The only thing that is done is protecting victory for someone whose chances are 90% or better. The actual outcomes removed may be smaller than 1%, and always only at one end of the scale. 99% of outcomes are still possible, and 100% of outcomes are still possible in all battles where one side does not have a >90% chance of victory. An attacker can kill a stack that should net him 4-5 losses and not take any, because outcomes are not restricted at the other extreme. I think extreme outcomes at both ends of the scale (cumulatively speaking) need to be removed. Thus, the actual 10 percent at each end of the scale.

So, for example, your outcomes range from:
Attacker wins 81% of the time.
Attacker wins, units remaining
1 @3%, 2 @6%, 3 @14%, 4 @53%, 5 @11%, 6 @2.2%, 7 @0.78%, 8 @0.02%
Defender wins, units remaining
1 @6%, 2 @12%, 3 @1%, 4 @3%, 5 @2%, 6 @0.05%, 7 @0.03%, 8 @0.01%

(defender's % don't mesh with the attacker's. I was lazy, don't worry about it).

As I go through the practical math of this exercise, I see that things look quite complicated, but they really aren't. It is still worth discussion. Perhaps it is possible to achieve the results KGB keeps describing where the 10% extremes at each end are eliminated and the middle 80% remain. A practical application here would be to remove all outcomes where the defender has more than 2 units remaining, as the total probability of better outcomes for the defender is 9.9%. That's one end of the scale. Then you remove all outcomes where the attacker has more than 5 units remaining, as the total probability of better outcomes for the attacker is 3%, but 5 is not removed because then the total probability would be 14%.

This method would achieve KGB's description of the results (which, I repeat, does not describe the results of the current system, which as I said removes between 0 and 10% of possible outcomes, and only in favor of a victor with >90% chances). It still allows for the employment of the current battle system in full, and the 90% victor protection (which fits in perfectly along the same line of reasoning). The only complication I see is that in logical continuity, guaranteed victories should perhaps have the 10% chopped off of each end of the victor's scale, as those are then the only possible outcomes. So in the example above (if the attacker's total probability of victory was 91% instead) he could not survive with less than 3 units or more than 5.

[KGB]

LPhillips,

This method would achieve KGB's description of the results (which, I repeat, does not describe the results of the current system, which as I said removes between 0 and 10% of possible outcomes, and only in favor of a victor with >90% chances). It still allows for the employment of the current battle system in full, and the 90% victor protection (which fits in perfectly along the same line of reasoning). The only complication I see is that in logical continuity, guaranteed victories should perhaps have the 10% chopped off of each end of the victor's scale, as those are then the only possible outcomes. So in the example above (if the attacker's total probability of victory was 91% instead) he could not survive with less than 3 units or more than 5.

Where did you get the idea that the current battle system *isn't* implementing what I said (removing a cumulative 10% at either end). Piranha specifically spent a lot of time implementing exactly that logic. I found one bug (which he fixed long ago) but otherwise have seen nothing else to indicate it's broken.

In the example you posted

So, for example, your outcomes range from:
Attacker wins 81% of the time.
Attacker wins, units remaining
1 @3%, 2 @6%, 3 @14%, 4 @53%, 5 @11%, 6 @2.2%, 7 @0.78%, 8 @0.02%
Defender wins, units remaining
1 @6%, 2 @12%, 3 @1%, 4 @3%, 5 @2%, 6 @0.05%, 7 @0.03%, 8 @0.01%

The possible outcomes are Attacker wins with 1,2,3,4 or 5 men or the Defender wins with 1 or 2 men. The cumulative 10% (or as close to 10% as possible) at the ends are indeed being removed. Sorry if we've misled you into thinking that's not what's implemented.

If you can find a battle that isn't doing this we'd be really interested in seeing it.

KGB

[LPhillips]

Pirhanna said:
"The battle is run 1000 times to find out what the chances are for each side to win. Then the real battle is run and if its less than 10% chance for one side to win, and they do win the real battle will be run again until the side with 10% to win loses."

If the cumulative 10% is being removed, that's great. But that isn't what he said. I'll take your word for it.

I'm not completely pleased with some results that occur. But I can't say that, without a battle simulator, I could even begin to test whether they are within the correct parameters.

[KGB]

LPhillips,

Just poor wording on his part. Believe me the original thread discussed it in great detail (6+ pages long). Then we both wrote separate simulators and compared numbers.

If you see anything strange, just write down the units involved and the bonus's for each side. Takes only a minute for me to verify the results. There could still be bugs lurking.

KGB

[LPhillips]

Thank you for the explanation and the offer. I'll certainly start checking out "unusual" battle results. Sometimes I grossly misjudge the odds, and find myself somewhat confused. For instance: that battle where I just took on Negern and found myself with over 90% chance to win, I was expecting to be barely above 60% but felt like taking a risk (I could easily afford to at that particular moment).

There's a bit of beginner's zeal in my posts to be sure. Nothing wrong with waiting until I gain experience with the system before making snap judgments. This said in retrospect, of course.

[KGB]

LPhillips,

I imagine you'd have hated Beta2 before the algorithm was put in place. I know I did because there were many times you'd get a result that was only 1-2% likely that always seemed to happen a key moments in the game (such as a single Lt Infantry in a city, str 3, defeating 4 wizards of strength 5 I slipped past my opponent to ravage their back cities). Drove me nuts which led to the concept of dropping a cumulative 10% on each end.

I still misjudge odds from time to time as well. Normally I'm pretty good at guessing the win% but where I miss out is guessing what units are in the 80% to be killed. For example in the dragon/bat vs Hv Inf example I mentioned above, with the 85% likelihood to win it seems like it should be a slaughter with you winning with a lot of men left. But in reality it's the dragon that is skewing the results because it's strength is so much greater than all the other units. Those type of battles, where there are 1-2 units with really high strength compared to the rest are the easiest to misjudge in terms of units that should be killed.

KGB