Convincing the Player of Automated Losses in Strategy Games

Under the category of "systems with no randomization", you have Chess. I'd hardly call that a game without interesting strategy.

On the other hand, systems of high randomization include any competitive variation of poker (the ones with rounds of betting, and/or the option to discard from your hand to draw new cards), which is a game about statistical analysis of your own position, psychological analysis of other players' position, and spreading intentional disinformation about your position to the other players. I don't enjoy this flavour of strategy, but others do.

Both of these enjoy the "you saw the whole thing played out" feeling, but in this case this is the entire game. In the case of Civilization, battles are microcosm events in a vastly larger game; the event is being treated that way.

If units were mini-AIs with a small set of statistics, instead of just statistics that are rolled against eachother, then maybe instead of a "roll, roll... victory to player 2" battle, the game could drop the units into a little battleground and have them fight out the battle themselves in automated, tactical fashion. Units would have very specific behaviour patterns, and these would give way to the results, instead of the results being wrapped into tidy numbers.

Then you could watch the battles unfold, but you might still think they're unfair or fixed, if you see the stupidity of your foot soldiers give way to failed battle after failed battle, shaking your fist and swearing that if you were in command you could do a better job.

I was immediately reminded of this thread. In that one, I couldn't come up with a good excuse to expand a simple test beyond the announcement of the outcome, but in the context of your problem, it seems very handy. If each "fight" consisted of ten hidden dice rolls instead of one, and they each had little names, then you'd be able to use the outcome either as a net result for furthering the game or broken down into a play-by-play that can be presented to the player if a more detailed report at their request.

So when your little men charge in there, they embark on a branching "engagement tree" of tests and chances. For an example, your cavalry gallop across a field toward some archers, and instead of taking their buffs from terrain and bonuses from the matchup into account and then choosing a random number, the game runs three quick checks, each with two possible outcomes, for a total of eight possible results. First, the enemy commander recognizes your formation and anticipates your maneuver(50% chance of this happening). Second, your troops pass over a natural spring that renders the earth swampy and unsuited to horsemen, which could not have been foreseen(10%). Finally, the archers luck out and their first volley takes out your field commander(20%), hurting morale and breaking the chain of command, allowing them to mop up a bogged down, leaderless rabble of mounted pinatas.

That's a nightmare scenario, it'll maybe happen in one out of fifty battles, and if even one of the tests had gone your way, you'd have whupped their bow-bending asses, but murphy stepped in and littered your path with slippery banana peels, so you lose this round.

Would being able to see that quick three-step chronology of the defeat be enough to assuage your rage and confusion?

I'm personally far more bothered by randomness than illogical unit stats in games that automate challenges. The random nature of combat should only be there to simulate little mundane details that the game doesn't want to deal with. It shouldn't have as much influence as it seems to in games like Civ 4.

I think there are better ways to add blurriness to combat odds. For example, you could give the army troops several states (such as morale) that influence combat, and allow the player's actions to influence them. Make the player aware of the states, but don't let them know which (or how much) of their actions influence them. Have unit animations or dialog hint at the level of the states, and never display them as numbers.

This type of system gives the player direct control, makes the units feel more life-like, and preserves the blurry edge of combat. My army lost because they were tired, sad, sick, or hungry. Not because the computer rolled artificial dice.

Quote:

Original post by Wavinator Or would it be better to use a system with very little randomization, so that when you apply a decision you have better assurances of the outcome? (This could lead to less interesting strategy)

Why do you think removing randomness would lead to less interesting strategies?

Randomness is only interesting if you can know, manipulate the odds and employ risk management strategies. Spreading out the risk, rather than sticking all your eggs in one basket and getting unstuck by a freak roll.

If the game is too unpredictable, players cannot reasonably plan anything - which kind of puts an end to any kind of strategy, interesting or otherwise. At least the newer Civ games use a hitpoint system that seems to use multiple trials to determine the outcome of a battle, giving greater predictability. While the odd freak result like your example is still possible, at least you don't get tanks being annihilated by spearmen quite so often.

Iron Chef Carnage, I like your idea. In it you have just as much control over the outcome, but because there is feedback, it is not as frustrating. I think it is the feedback (lack there of) that causes the frustration.

It is similar in idea to the PnP D&D 4th ed "Skill Challenge" concept. When a player wants their character to perform some non combat task, say negotiating for the release of a prisoner, then this would be handled by a skill challenge.

A skill Challenge is made up of several skill tests.

A Skill Test is run by a player rolling a 20 sided die and adding there skill level to it. If this equal to or higher than a target number (either pre-set or the result of some other skill roll).

The Skill Challenge builds on this by having the player make several skill checks in succession. However, there is one extra rule.

The player must make a certain number of successes before they make a certain number of failures.

Using the example of a game like Civ4:

You have a unit of Cavalry and a unit of Archers in combat.

Each success a side gets reduces the HP of the enemy unit by 1 (each unit has 3 HP).

Cavalry have a net +10 against Archers (this equates to around a 50% extra chance of success using a 20 sided die as a randomiser).

In each test it would be for a different line of the "story" of the battle:

So in the first roll, the Cavalry rolled a 5, giving them a score of 15 against the Archers. However, the Archers roll a 16 and so beat the Cavalry.

Now in the battle report, this might say: They archers controlled the ruined fort and the first charge by the Cavalry was held off by them.

In the next test, the results might have been: a 20 for the Cavalry (10 + 5), and a 17 for the Archers.

The battle report might state: The Cavalry caught a small patrol of Archers in an ambush.

The next test might have been: 13 for the Calvary (3 + 10) and a 19 for the Archers.

The battle report: A night attack in the Cavalry camp by the Archers caught them off guard.

The next test could be: 17 for the Cavalry (7 + 10) and a 20 for the Archers

And the Battle Report: The Cavalry commander, desperate to destroy the archers, made a final assault on the Archer's encampment, but fell to their superior numbers.

The reports would be constructed from templates selected and modified by variables (who won or lost the check, how much they lost the check by, the types of units involved, and other factors that you want to add in).

As an example a template might be:
The [winner] caught a [intensity] patrol of [losers] in an ambush.

[intensity] would be based on the amount the winners won by (it does not have to reflect the amount of damage done, but it could if you wanted it to). So a win of 5 or less might indicate a "small" patrol. A win of between 5 and 10 would indicate a "moderate" patrol, a win of more could mean a "large" patrol.

This template could have been selected because the previous check resulted in the current winners being the losers, or maybe because a variable on the terrain square indicated that this was a viable option. Or it could have been selected completely at random. You might have even had several templates available and one of these was chosen at random.

Iron chef's idea is good and looks like it's relatively easy to implement.

An even simpler feedback mechanism would be if the game said "Ye gods! The German archers managed to defeat the Russian cavalry on an open field! A miracle!". In other words it would confirm to you that you should really have won.

This message can be customized for unit, terrain, outcome, offense vs. defence, etc. So the battle above might get a "Cavalry charges into prepared archer ambush!". If one cavalry unit defeats another after an even fight it might be "Cavalry forces enemy cavalry to retreat after heavy losses." or something like that.

So yeah, it's a feedback thing I think. I'd like two things: First some kind of guesstimate from the game what I could expect from an engagement and secondly some debriefing like the "Ye gods"-comment above. Something that says "your decision was correct, but you where outta luck this time".


Also I wonder, Wavinator, do you feel robbed because you feel the cavalry should win according "what history tells you" or because you feel the cavalry should win according to your understanding of the game rules? There is a subtle difference there.

Thanks for the great responses.

Quote:

Original post by Iron Chef Carnage Would being able to see that quick three-step chronology of the defeat be enough to assuage your rage and confusion?

Bingo! That's the answer for me. I get annoyed because it doesn't seem plausible, and it doesn't seem plausible because I didn't see the rolls. But your solution and Edtharan's comments about skill challenge sound even more entertaining than seeing the rolls.

One problem might be coming up with enough varied text, but you might be able to mix it with pictures or icons, which would have the benefit of speedy play. And the situations themselves would have to be intelligible, so that the player isn't crying out, "What were they doing on patrol???? They should have been fortified!" or whatever.

Quote:

Original post by Kekko So yeah, it's a feedback thing I think. I'd like two things: First some kind of guesstimate from the game what I could expect from an engagement and secondly some debriefing like the "Ye gods"-comment above. Something that says "your decision was correct, but you where outta luck this time".

That captures the essence of it for me, I think. It would validate my choice.

I also think that giving some kind of estimate is critical to this sort of situation. The flipside of wondering how my guys died has often been wondering what the hell it takes to kill a certain unit. Just as in modern combat, I'd love to see some sort of general's comment along the lines of what's required for victory in a given battle.

If a branching narrative is used, you could expose this to players. You could say, "We'll win if..." and throw in the most obvious branching points.

Quote:

Also I wonder, Wavinator, do you feel robbed because you feel the cavalry should win according "what history tells you" or because you feel the cavalry should win according to your understanding of the game rules? There is a subtle difference there.

True. I think history is a big factor but not the biggest. I've played really abstract strategy games and been put off by not believing (for lack of a better word) the results even when they were represented as a single number over a map of a country. I'll get the same feeling in a Risk clone ("I have an 8 in Australia, how the heck did all of these surrounding 2's defeat me?")

Personally I would only use dice in a "break glass and use two die" kind of emergency, but I hate dice.

I'd just have lots of rules and variables.

Commanders/Generals would have skill levels which would be compared to one another in battle.

Units would have have tests like can those arrows penetrate that armour.

An awareness value for whether the unit has been flanked or caught by surprise.

Which would go into a distance value for how many arrows the archers are able to fire before the knights reach them.

Something like that anyway not sure whether it would work, you would play by using superior generals, superior units, and try to gain intel on your enemy and position your army in an optimal position, and not get flanked, surrounded or outnumbered. Which if done perfectly by the player should guarantee a win unless the opposite army is almost as equal which in that case either both armies are annihilited or one is slightly defeated or a stalemate.

I am reminded a bit of the combat in Taipan. Here, you made very simple choices; Fight or Run. Also, you had Dump Cargo which increased the chance of escaping by Run at the cost of cargo. And you had a "hurry up" option which was Fight To The Death; it did nothing more than select fight every round of the battle for you until one side or the other won. Lastly, there was a random but small chance that Li Yuen's forces would drive off attackers if you had good standing with him.

This extremely simple system for a very early game could easily be improved upon to generate a fun, interesting combat system for a game. If a battle were modeled as a series of steps and the player had to choose what to do at each step, this would give a sense of control and interaction to an otherwise automatic result.

So, perhaps you can offer a basic strategy for your forces (draw, sweep, charge, feint) before a fight, with the right advantages you might be able to change strategy during the fight (defend, rally, advance), and deal with the end of the fight (pursue, retreat, massacre, mercy). Choices can modify the results of course, but you are still left with a largely automatic fight.

Quote:

I'd just have lots of rules and variables.

Now I agree that complex combat can be fun, the scale of the combat under discussion does not lend itself to such detail.

So although a detailed combat system as you described might be enjoyable in a smaller scale game, in a game like Civ, where the Squares on the board represent hundreds of kilometres (or more), such detail is not really feasible.

An easy way to simulate such detail existing, but without actually going into it, is to use a random generator.

At the coarse level of detail that Civ (and other games like it) have, this level of detail does not exist, so the only response is to use some other system and the easiest is that random number generator.

Quote:

Now I agree that complex combat can be fun, the scale of the combat under discussion does not lend itself to such detail.

When I said a lot of rules/variables I didnt mean the player would have to/or be aware of them all. They would just have to use common sense of how they think they would function, and just focus on manuevering and position the units how best they think they would work.

You wouldnt have to play it too differently than you would in a normal civ game, but perhaps, could be more aware that there is a consistent ruleset behind it and that dice will not often be rolled.

[Edited by - Calabi on October 8, 2008 10:44:25 AM]

Quote:

Original post by Calabi Quote: Now I agree that complex combat can be fun, the scale of the combat under discussion does not lend itself to such detail. When I said a lot of rules/variables I didnt mean the player would have to/or be aware of them all. They would just have to use common sense of how they think they would function, and just focus on manuevering and position the units how best they think they would work. You wouldnt have to play it too differently than you would in a normal civ game, but perhaps could be more aware that there is a more consistent ruleset behind it and that dice will not often be rolled.

This is exactly the type of gameplay I would prefer. Have every little player decision or adjustment define some subtle effect that boosts their units in some specific way. Simply hiding variables brings the game characters to life.