The 90% rule

[ams16]

I've had my doubts about the 90% rule for a while now. I understand the reasoning behind it - losing a crucial battle through dumb luck really sucks. But I also think that chance is part of the game, and that's that.

However, I also have realized that losing a crucial battle early in the game is MUCH worse than losing a battle in the mid- or end-game.

So I have a new proposal:
On turns 1-2, you auto-win 70% battles.
On turns 3-4, you auto-win 75% battles.
On turns 5-6, you auto-win 80% battles.
On turns 7-9, you auto-win 85% battles.
On turns 10-14, you auto-win 90% battles.
On turns 15-20, you auto-win 95% battles.
Afterwards, there is no auto-win.

I'm sure you might want to modify the numbers. Also, the turns should probably be different on different sized maps. Perhaps there could be a version vs. neutrals and version vs. players.

Just a thought. Any ideas?

[Moonknight]

It's all relative, if we lower the percentages at the beginning of the game, then players will take more chances and there'll be just as many unlucky losses (as in there'll be a 69% loss on the first turn).

Once there is no auto-win, players will take very few unnecessary chances.

I don't like the changing by turn b/c then I have to keep track of, "what turn is it"?

Seems like it'd be confusing for newbies as well.

Just my thoughts...

[Garesch]

The worst situation to lose a >90% battle is when it's your carefully-nurtured hero stack or your only hero. As such, what about making auto-win a hero ability - applying only to stacks that have a hero in them. This provides a thematically appropriate explanation for why normal probabilities are being over-ridden, since a hero's leadership abilities would be expected to prevent lopsided low-chance losses.

That's the basic idea. If you wanted to make it more restrictive, you could have the 90% rule only apply to stacks containing a player's only hero or all of the player's heroes - but not apply if the player has any hero elsewhere on the map. Alternatively, if you want to get fancy, you could make it a variable hero ability that can be trained to a certain maximum (e.g. paladins and valkeries start at 90% and can spend points up to a best value of 80%, while death knights start at 92% and can spend points to a best value of 84%)...

[LichKing]

Once I suggested something like Garesch idea. It'd be good, I think, if every unit had a stat called 'intelligence' or 'tactics', that affects the auto-win threshold. For example:

Intelligence 0 (0% auto-win): fire elemental, sandworm, crow, eagle, scorpion, spider, sea serpent, kraken, elephant, mammoth
,, 1 (auto-win at 95%): gryphon, orc, wolf rider, yeti, minotaur, catapult/ram, ghost
,, 2 (auto-win at 90%): scout, infantry, cavalry, dwarf, giant, pikeman
,, 3 (auto-win at 85%): elf, wizard, crusader, demon, medusa, unicorn, archon, green dragon, pegasus
,, 4 (auto-win at 80%): devil, grand archon, red dragon

Heroes could start at 90% and improve with skill (or ability) points up to 75%.

There could be various alternatives:

1. each unit conserves its survival chance after an auto-win depending on its intelligence
2. the highest intelligence is used (but there would be more auto-wins for stacks with red dragons etc)

Alternative 1 would be more realistic but maybe too complicated to do, while alternative 2 would be easier and not too different from now, because already stacks with STR>80 (if I'm not wrong) have an auto-win at 85%, and for STR>400 auto-win is at 80%.

In the alternative 1, it could be possible, for example, that in a battle hero+spider vs light infantry (town wall 0), the spider dies (it's impossible right now), but if the battle is lost (also the hero dies) the battle is rerolled.

[KGB]

Adjusting the auto-win percentage is a very very delicate thing that I don't recommend.

For example, consider a stack of 8 Giants vs 8 Giants. Right now that is a 50/50 battle. But would you believe that simply changing 1 Giant for a Crusader/L1 Paladin changes the win % all the way to 85%. Just a measly +3 (or +4) strength difference. So at 85% that would suddenly be an auto-win. To me that battle doesn't seem remotely one sided enough to guarantee one player an auto-win.

Reaching 90% is much much easier than players think if you aren't risking a ton of 2v1, 3v2 type battles hoping to succeed with 75% win chance.

If anything, I'd say the auto win percentage should be higher than 90% for anything but hero stacks (maybe 95%)

KGB

[LichKing]

I'd be for a higher chance of auto-win. It would make more sense if, in case of auto-win and rerolled battle (that is, if the strong side loses against the odds), instead of actually rerolling the whole battle (with the high chance of a much better outcome for the winning side) the winner loses all units except one. At that point you could lower the auto-win threshold without fearing of having "counterfeited" battle results and you would solve both problems.

[kenc80]

i think 90 is good. If we start changing it depending on the situation, stack or what turn it is, it makes the math that much harder to compute for beginners & average players.

[KGB]

LichKing,

It would make more sense if, in case of auto-win and rerolled battle (that is, if the strong side loses against the odds), instead of actually rerolling the whole battle (with the high chance of a much better outcome for the winning side) the winner loses all units except one.

I understand what you want to do. But what you proposed isn't what you want. For example consider 8 dragons vs 1 bat. Imagine the bat somehow killed 2 dragons which isn't allowed under the 90% rule (this causes a re-roll). Your proposal would then see the winning side reduced to 1 dragon! I don't think that's quite what you had in mind.

Also, there is not a better chance of an outcome for either side on a re-roll. To understand this, consider 8 Lt Infantry vs 8 Lt Infantry. The odds of this battle turn out to (a random sample from my simulator which uses 10000 combats):

Side A: 1, 32, 103, 300, 672, 1095, 1343, 1436
Side B: 1, 25, 109, 308, 717, 1114, 1317, 1429

You read these numbers left to right as the number of men surviving when you win starting with 8 men and going down to 1. So for side A, 1/10000 times you win with 8 men left, 32/1000 you win with 7 men, 103/10000 with 6 men etc.

When the 90% rule is applied to this combat each side loses the extreme 10% starting on the left (most men 8) and working to the right. That means both sides lose the 1,30, 100 and 300 numbers and what remains is one side or the MUST WIN with between 1-4 men.

So if the combat returns one side winning with say 6 men left, it gets re-rolled. When it does, there is NO advantage to being either side in this battle. This is true of all battles, even ones that are 90/10 instead of 50/50 because the reason a battle gets rejected could be because the winner did too much damage just as easily as it can be the loser doing too much damage.

However, I do understand that some players don't like the re-rolls. If I was implementing it, I would not have re-rolled, instead I would have returned the closest legitimate result instead (so in the case of one side winning with 6 men, I would have simply returned them winning with 4). That would have preserved which side won in battles like this. However doing that would mean we'd no longer be able to review the detailed battle numbers (the 'i' icon on the battle screen) since they would show 6 deaths instead of 4 (as long as you didn't care that the numbers didn't match the results and just wanted to see the straight percentages).

KGB

[LichKing]

I understand what you want to do. But what you proposed isn't what you want. For example consider 8 dragons vs 1 bat. Imagine the bat somehow killed 2 dragons which isn't allowed under the 90% rule (this causes a re-roll). Your proposal would then see the winning side reduced to 1 dragon! I don't think that's quite what you had in mind.

Why that? The battle is rerolled when odds are >90% and the strongest side loses. 8 dragons will never lose against a crow. An auto-win with 1 dragon left would mean that the crow has killed all 8 dragons against the odds.

About the better chance: units with ambush can give much better outcomes if there's an auto-win. Say you have pala+medusa against anything and an autowin (slightly more than 90%). If the battle is lost and rerolled, the medusa will have a second chance for her critical (likely to succeed). Same for orcs/wolves/ghosts against not too strong enemies. In those cases is even better not to overcome the 90% threshold by much, that is to attack with less units, to have an increased chance of battle reroll and critical reroll. The advantage is greater in battles involving few units. The rule 'winning stack with 1 unit left' in case of auto-win would take place *only* if the battle is lost against the odds. So, if you lose a battle in the autowin range, the battle wouldn't be rerolled, you would win but with only 1 unit left.

Rereading I understood what you meant btw. So the battle isn't rerolled when the battle is lost, but whenever it is out of the middle 80% range. It doesn't change much. You would have:

1. worst case (battle lost): no reroll, 1 unit left.
2. 2nd worst case (battle won, but in the first 10% fraction, that is an unlikely and unfavorable outcome): no reroll, you keep the units in that 10% fraction.
3. best case (last 10% fraction, the most favorable one): no changes, as it is currently working (the most favorable 10% is rerolled).

[KGB]

LichKing,

About the better chance: units with ambush can give much better outcomes if there's an auto-win. Say you have pala+medusa against anything and an autowin (slightly more than 90%). If the battle is lost and rerolled, the medusa will have a second chance for her critical (likely to succeed). Same for orcs/wolves/ghosts against not too strong enemies. In those cases is even better not to overcome the 90% threshold by much, that is to attack with less units, to have an increased chance of battle reroll and critical reroll. The advantage is greater in battles involving few units.

Nope. The ambush chance is factored in as part of your overall winning chance (this is why 1000 simulations are done to calculate the actual winning chance). So it matters not whether you ambush the first or second time or not at all.

Rereading I understood what you meant btw. So the battle isn't rerolled when the battle is lost, but whenever it is out of the middle 80% range.

Correct. This is means that 50% of the time on a re-roll you will get a better result (initial battle went against you too much) and 50% of the time you will get a worse result (initial battle roll went for you too much). This is why it's fair for both sides.

This is why the bat killing 2 dragons gets re-rolled and why you can't just drop winner down to 1 man.

It doesn't change much. You would have:
1. worst case (battle lost): no reroll, 1 unit left.
2. 2nd worst case (battle won, but in the first 10% fraction, that is an unlikely and unfavorable outcome): no reroll, you keep the units in that 10% fraction.
3. best case (last 10% fraction, the most favorable one): no changes, as it is currently working (the most favorable 10% is rerolled).

So basically you just want to take my 2 cases (your 2 and 3 case) and add a 3rd case specifically for a fully lost battle that instead of keeping 10% only keeps 1 man? In the other 2 cases, men get restored to life that should not have died (case 2) or get killed that should have died (case 3). There is no re-roll at all. Why bother at all with case 1 as it just adds slightly more programming.

The question of course becomes what to show in the battle screen. For example lets say the roll turns out that 1 bat kills 2 dragons when it should kill 0. Are you going to show 2 dragons dying but the player still has 8 units left? That would be really confusing I think. Or are you going to fudge what's shown in the battle screen to reflect the final outcome (no dragons die, 8 dragons left). If you do that, then there the full replay (where you use the i icon to see the actual rolls and percentages for each unit including ambush rolls) can no longer be available since it won't make any sense because the replay is going to show units dying that don't actually die in the battle screen or the game or it will show units left that get killed after the battle because too many survived (case 3).

KGB