Slitherine

Xitax wrote:As to showing the span of results, iteration is a bad and costly model (in terms of cpu cycles), but a reasonable model would be to show the best and worse result within one standard deviation of the mean.

I agree with the iteration being a bad idea, it was a (too) primitive way to get mean and standard deviation of the combat results. I expect that there is also a nice & clean way to obtain the values using probability theory. The 'one standard deviation' you mentioned would get the values of a probability of 68,27 %, would this cover enough outlier?

Xitax wrote:However having said that I am against it because it will clutter the GUI and I don't want any more stuff in there.

It wouldn't need to clutter the GUI, as it could e.g. be implemented by using some sort of color indication for the prediction mouse cursor and showing e.g. the best and worst expected result for the player of a certain probability only in the 'combat details' dialog. But this would most likely cause heated discussions on how the differences of the three results translates into a color indication in a meaningful way.
Instead i would really prefer the combat prediction mouse cursor to just also include the best and worst expected outcome (e.g. a number without box above and below each of the two prediction boxes, the numbers above containg the best, the numbers below the worst outcome for the player and of a certain probability). In my opinion this would be an intuitive and clean interface .

Xitax wrote:I don't have a problem with knowing that the combat predictor is not 100% accurate and the numbers that we do get there are a good enough to give me a feeling for possible results as they are right now.

I also have no problem with the randomness, but i don't like having no indication of how 'stable' the predicted results are. And if i remember my PG days, for a newbie it was not only frustranting but even worse incomprehensible why the combat indicator showed a postive results, while in the real battle the troops would get slaughtered. And i am not talking about having forgotten the enemy unit being protected by artillery, but rather trying to understand how/in which amount the different factors (wheather, entrenchment, experience) influence a battle.

Hmm. What if you could open an additional window with more detailed information? Then you could include there a more detailed listing of probable outcomes. I'm just having a hard time imagining how this could all be shown on the screen in that limited area as you hover over the attack icon.

Another idea would be to put that info in the 'information' sidebar in all that empty space on the lower half.

Xitax wrote:Hmm. What if you could open an additional window with more detailed information? Then you could include there a more detailed listing of probable outcomes. I'm just having a hard time imagining how this could all be shown on the screen in that limited area as you hover over the attack icon.

Another idea would be to put that info in the 'information' sidebar in all that empty space on the lower half.

I suggest you check out how Wesnoth does this for. Excellent game, and they handled the issues with windows within windows.
Of course, they also have units with multiple kinds of attacks, which would be great for Panzer Corps (Attack with MG or attack with 88, where the two weapon systems have seperate ammunition counts) but now it's starting to get too complicated and we're losing the beautiful simplicity that is the hallmark of the series.

Ironically I was going to mention Wesnoth as an example but I wasn't sure anyone here would be familiar with it...

But now that you did, I was thinking of this kind of window when I wrote that. It's available to bring up if you want it:

Xitax wrote:Hmm. What if you could open an additional window with more detailed information? Then you could include there a more detailed listing of probable outcomes.

There is already the 'Combat Details' dialog, why not just extend it instead of adding another dialog ?

Xitax wrote:I'm just having a hard time imagining how this could all be shown on the screen in that limited area as you hover over the attack icon.

Would something like this really be so bad (besides the color choice for the lower numbers)?

• 

I would like having detailed prediction data available in a dialog window. But i am even more interested in having summary information available without having to open further windows. Else at least two additional mouse clicks (window open & close) would be required for every enemy unit i consider to attack. Having summary information available in the combat prediction mouse cursor allows for a way faster UI, as the prediction is updated by just moving the mouse over the enemy units i am interested in. And i am also of the opinion that the mouse cursor would be the logical place to put this information.

Xitax wrote:Another idea would be to put that info in the 'information' sidebar in all that empty space on the lower half.

An interesting idea.

Kerensky wrote: ...and we're losing the beautiful simplicity that is the hallmark of the series.

I disagree!

Simplicity means transparence what will come. But the results often are too far away from the numbers shown in the calculations. On both sides, too good and too bad, and surely not simple in this case.

I think the argument 1) goes in circles and 2) that many suggestions do indeed lose sight of the beautiful simplicity of the current game. Any solution requiring an extra dialog simply isn't accessible enough for a game of this kind.

So I'll simply reaffirm my conviction you can accomplish real improvement with relatively modest effort, just providing existing information in a more immediate way; making the application more user-friendly.

In other words, some of you are making this out to be much more difficult than it needs to be.

(Regarding the "why don't you believe the developer" question, let me remind you that my point was that he and I are discussing different issues)

Other than that, we'll simply have to wait and see what future versions bring. I am grateful I have been given this chance to bring up my points, and there really isn't anything else to add. My claims still stand. What remains is to see what (if, indeed, any) impact they have on future development!

Ok Molve, no issues, guess I misunderstood your reply to Iain where you said:
" Of course you can make the predictions more accurate without changing the underlying mechanics. "

Semantics maybe, but at that point in the thread it seemed some people want to make it more accurate which is not possible. Now the theme is this realization but wanting more data displays showing all the potential outcomes and what are the odds of such might be etc etc etc.
Reminds me of the line from the movie Naked Gun. "Nordbert has a 50 % chance of living, but theres only a 20% chance of that"

More data doesnt make the predictor more accurate, but imho gives overloads of information that really isnt needed to play the game. If the developers want to add in all this in , no sweat off my back, as long as I can turn it off and it doesnt effect my CPU, although I'd rather them be hard at work on more content, expansions etc.
cheers!

What I have noticed:

Do a Control click on unit you want to attack.

Those results predictions are different and are accurate or much closer.


Have not had the time to document it.

For example: Screen shows 3 to 1 but Control Click reveals: 1 to 1 or 0 to -1

It appears the screen shows basic combat data.

The Control Click includes entrenchment, experience, suppression, terrain, weather etc...

Thus the different prediction on combat odds.


It's a pain in the buttock to do a control click on a unit before every attack. Let alone move up another unit and do another control click to get odds.

Even when you do this the outcome varies, but not as wild as in using the on screen data.


Maybe, what I am noticing in pie in the sky!

Sur ewould be nice to see if any one else has checked this out or appear to notice the same thing?

Might need some specifics there. I did a test with 7 or 8 different types of attacks and all the cross-hairs numbers matched the Ctrl-click numbers, minus the suppression, of course.

On a (I think) related note:
I just listened to a recent Three Moves Ahead podcast where they had Brian Reynonlds (of Microprose/Firaxis fame) as a guest, and they were talking about Alpha Centauri.
At one point, Brian explained that they contemplated about save/reload "cheating" of players in Civ, and thus the programmers agreed on the brilliant move of just storing the random seed in the savefile, thus preventing players from utilizing this maneuver.
He then went on to explain how this idea totally bombed with the testers, one guy even replaying the same battle like 1500 times and (obviously) never winning, explaining how, against all odds, he'd be unable to win. The stupid idea of storing the seed in the savegame was silently dropped.
He then said he went from preventing ANY cheating to just providing the cheat option right in the main menu in Civ2. Let players have their fun!
Soren Johnson, also a guest, and also of Firaxis(Civ3) fame, went on to reveal that they ran into exactly the same trap in Civ3, where the situation was later resolved by proving an option in the advanced settings that'd reset the seed whenever a savegame was loaded.

Panzer Corps fell into that same trap. And the idea of storing the RNG seed in the savegame remains just as moronic as it ever was.
Like Brian explained in that podcast, it makes perfect sense from a programmer standpoint, but programmers of a particular game tick differently than (at least some of the) players.

As my own (limited) tests revealed, PG combat results were also very varied and you could fall victim to horrible dice rolls. BUT you could just say "screw this, ain't gonna happen!" and reload your latest quicksave.
I did this in many campaigns, especially earlier ones. At some point, I occasionally "Iron Man"'ed, but on the average, I guess I have to admit I reloaded when a 0/4 turned into a 7/0 more often than not.
And since I could just circumvent horrible combat results that way, they also haven't stuck in memory.
In PzC, there's no way out. And that's the worst thing.

So, I propose just making that ONE change, provide an option to reseed the RNG upon loading a savegame. I really don't see why this wasn't simply included in 1.01 already, it ought to be VERY VERY simple to implement and has NO further implications that need to be tested or balanced.
_____
rezaf

Ok, so playing the new DLC campaign has made me wary about the RNG once again.
I decided to keep a list of all combat results in a mission. Here's the result:

There were 91 combats initiated by me (actually a few more, but I didn't count rugged defense results).
Of those 91:
36 had the predicted result - mostly exactly, sometimes taking more losses, but inflicting more casualties in turn.
14 had a slightly better result (usually killing one more point of strength than predicted).
2 has a considerably better result.
36(!) had a slightly worse result - mostly taking losses when none were predicted or failing to kill that last point of strength of an enemy unit.
4 had a considerably worse (aka disastrous) result.

Slightly means 1 difference max, so if 1/3 was predicted and 2/3 was the result, it's slightly worse, if 1/5 was the result, it'd count as considerably better.

As you can see, at least in this mission (it was Lillehammer, btw.) the RNG clearly penalized the player.
Oh, the game was played on Colonel difficulty.

I'll see about keeping tabs in more missions, but the trend is already clear, and at almost 100 combats can't be pure coincidence, imo.
_____
rezaf

The problem is, that 100 examples are nowhere close enough for a trend observation or similar. I think this generator is fair, in the sense that is will produces random numbers for both participiants, not advantaging the one above the other.

The problem is not that results are not as predicted in general, rather the problem is that deviances are to big to frequently.

Ranta wrote:The problem is, that 100 examples are nowhere close enough for a trend observation or similar. I think this generator is fair, in the sense that is will produces random numbers for both participiants, not advantaging the one above the other.

Well. It was supposed to be a starting point.
I continued taking notes through Narvik. Here's the result:

163 attacks.
70 more or less as predicted.
14 slightly better.
6 considerably better.
40 slightly worse.
23 considerably worse.

Still not seeing a trend?
_____
rezaf

rezaf wrote: ...
As my own (limited) tests revealed, PG combat results were also very varied and you could fall victim to horrible dice rolls. BUT you could just say "screw this, ain't gonna happen!" and reload your latest quicksave.
I did this in many campaigns, especially earlier ones. At some point, I occasionally "Iron Man"'ed, but on the average, I guess I have to admit I reloaded when a 0/4 turned into a 7/0 more often than not.
And since I could just circumvent horrible combat results that way, they also haven't stuck in memory.
In PzC, there's no way out. And that's the worst thing.

So, I propose just making that ONE change, provide an option to reseed the RNG upon loading a savegame. I really don't see why this wasn't simply included in 1.01 already, it ought to be VERY VERY simple to implement and has NO further implications that need to be tested or balanced.
_____
rezaf

There's a workaround in PzC as well.
As "bad rolls" are always the same for a given turn, you can "spend" a roll with another unit (using Artillery/Bombers/Aux units). So in you can circumvent the results to a extent.

But I agree a simple reseed on load option would be simpler.

I have been saying this for months. The RNG needs to be toned down by 10%

Don't think so. Should be substantially more predictable.

if the predicted result is the expected outcome, cut down standard deviation

The first mission in the '40 DLC turned out a lot more balanced, for whatever reason.

140 battles total.
60 as predicted.
30 slightly better.
6 considerably better.
28 slightly worse.
16 considerably worse.

Still a large number of CW results, but other than that ...
_____
rezaf

Is there a place in PzC where I can check the statistics of damage caused vs expected and damage taken vs expected (similar to stats function in Wesnoth) for a scenario or campaing?