Slitherine

Xerkis wrote:

MartyWard wrote: It certainly isn't a problem when the extreme results work in my favor!

How would you know? When it error in your favor, you most likely didn’t attack. An opportunity lost.

Maybe that's why I get creamed, I don't pay a lot of attention to it. I think the phrase is leading with your chin

willgamer wrote:So I'm feeling just gobsmacked and stuck...

IMHO, Molve has identified a real problem;

The RNG is broken because it gens long sequences of numbers in a narrow range;

The devs have said the broken RNG might be hard to fix, and/or, has a low priority;

We don't even have a list of items in the first patch, much less an eta for the first patch itself;

I give up!

Wouldnt a long sequence of #'s in a narrow range give the expected results more often than extreme results?

Where is a link where the developers have said the RNG is broken?

TheGrayMouser wrote: Wouldnt a long sequence of #'s in a narrow range give the expected results more often than extreme results?

Where is a link where the developers have said the RNG is broken?

Depends what the bounds of the range are... might be high, might be low, might be middling.....

http://www.matrixgames.com/forums/tm.asp?m=2870266

willgamer wrote:

TheGrayMouser wrote: Wouldnt a long sequence of #'s in a narrow range give the expected results more often than extreme results?

Where is a link where the developers have said the RNG is broken?

Depends what the bounds of the range are... might be high, might be low, might be middling.....

http://www.matrixgames.com/forums/tm.asp?m=2870266

Ah I see, i made the false assumption that a string would be like a bell curve of sorts with the liklyhood of rolling a 100 a 99 or a 1, 2 much less likly than rolling say #'s with in 40-60 or something like that. Thanks for the link BTW.

Just a curiosuity question for those more mathametically inclined:
Not that the engine should or does do this but, what would be the effect , concern , problem or what not be if a truly random # was generated per "dice roll"
As it stands a unit of 10 sp's roll 10 percentile dice with , for example 60% to miss, 20% supress and 20% to kill for each roll. What would be the issue if each "roll" was truly a roll with the same proboblity or chance to roll a one or a 100 or a 53?
Is the issue the small sampling of #'s ( ie only 10 rolls) could give vastly divergent results in any single example? (of course it would average out over time)
Just curious , I have no idea how this works......

TheGrayMouser wrote:Molve , how can you "predict the unpredictable? The game cant , in the pre combat predictor , incorporate the RANDOM 0-+2 initiative boost either unit can get when combat is resolved. (Im not even so sure it utilizes or factors in the nominal difference of two units initiative)

Since each point of initiative over your opponent gives 10% (or is it 20%??) ist fire capaibilty of your strength points(which allows them to supress or kill w/o firing back), very quickly you can see there would /could be an exponential increase in possible outcomes that cant boil down to a 1-5 preidiction.
You could , i suppose give a high low prediction but would that really be usefull? Yes , we will either win at 5-1 or lose 1-5 with everything in between being possible.... That is not helpful at all.

Cleary for the % dice that roll if you hit miss or suppress are not just staight up , pure random #'s either, my guess some curve, mean , median etc is used , otherwise you would at least by now have heard of a player losing a 10 sp tiger to a miltia infantry unit that takes no losses. Improbable yet possible if just straight random dice were thrown.

Rugged defence is somewhat of a mystery, at least in terms of what % chance you will trigger it yet it tells you what % chance it might be triggered. That should be enough in assessing the risk in your decision to attck or not, is it not?

First off, a casual game like this needs sufficient information at the player's fingertips. Having to delve into a second information screen needs to always remain completely optional. That is, doing so may well give additional or deeper quality information that helps you decide on finer nuances; but major surprises (such as getting 50% losses on an attack) needs to be advertised clearly without it.

Put simply: as a designer you need to assume your details screens aren't used by your players. Then it's okay to reward those players that do with something extra. But the basic gameplay cannot - must not - assume players look at the details for any combat. Otherwise the game isn't casual, full stop.


Several times in this thread alone I'm reading replies that somehow assume things are impossible to change. Rest assured that everything I'm proposing can be accomplished.

For instance, if the game calculates the outcome first, and bases its "prediction" around the final result instead of having two separate procedures, then it isn't just simple to make sure there are no nasty surprises - it's outright trivial.

While the programmers will know this truth - nothing will have to be made known to the player. The game could still choose to deny the player the information that his attack is about to get Ruggedly Defended etc. (The player does sthen still control the risk of getting a RD, since he can always play safe and bomb all entrenchment out of the defender first, if he chooses to be timid)

It's just that the game needs to stop allowing itself to be taken hostage by the Random Number Generator. It's a game, and a casual one. The game should control the RNG, not the other way around.

This game is not a good place of giving "0-7" estimations and "5-3" results. This is not the place for ambushes resulting in "0-0" results. There should be clear consequences for your actions, allowing the player to be in control and in charge of what risks he's taking.

I'm sure there's a market for fully randomized gameplay, but first off: that's not the Panzer General gameplay we know; and second: let me repeat I'm in no way out to destroy your game - whatever changes are made, the current situation could always be retained as an option (perhaps an "Ironman" checkbox?).

Molve, I personally agree with everything you are saying except for one (very minor) point.
I don’t consider PzC as a “casual game”. Obviously you (and I’m sure many others) do think of it as such. But I’m also sure that others do not. I’m just simply saying that I don't think your classification of the game isn’t going to be much of a winning argument in this subject.

But, like I’ve said many times in this thread; the predictive odds do need to be a bit more closer. Should there be a surprise result for whatever reason? Yes, absolutely. But should it by as far apart as 0-6 and then results of 5-0? No, never.

And you have given plenty of valid ways to fix this issue. All very doable.

… at least this is the way I look at this subject.

Molve wrote:

TheGrayMouser wrote:Molve , how can you "predict the unpredictable? The game cant , in the pre combat predictor , incorporate the RANDOM 0-+2 initiative boost either unit can get when combat is resolved. (Im not even so sure it utilizes or factors in the nominal difference of two units initiative)

Since each point of initiative over your opponent gives 10% (or is it 20%??) ist fire capaibilty of your strength points(which allows them to supress or kill w/o firing back), very quickly you can see there would /could be an exponential increase in possible outcomes that cant boil down to a 1-5 preidiction.
You could , i suppose give a high low prediction but would that really be usefull? Yes , we will either win at 5-1 or lose 1-5 with everything in between being possible.... That is not helpful at all.

Cleary for the % dice that roll if you hit miss or suppress are not just staight up , pure random #'s either, my guess some curve, mean , median etc is used , otherwise you would at least by now have heard of a player losing a 10 sp tiger to a miltia infantry unit that takes no losses. Improbable yet possible if just straight random dice were thrown.

Rugged defence is somewhat of a mystery, at least in terms of what % chance you will trigger it yet it tells you what % chance it might be triggered. That should be enough in assessing the risk in your decision to attck or not, is it not?

It's just that the game needs to stop allowing itself to be taken hostage by the Random Number Generator. It's a game, and a casual one. The game should control the RNG, not the other way around.

This game is not a good place of giving "0-7" estimations and "5-3" results. This is not the place for ambushes resulting in "0-0" results. There should be clear consequences for your actions, allowing the player to be in control and in charge of what risks he's taking.

I'm sure there's a market for fully randomized gameplay, but first off: that's not the Panzer General gameplay we know; and second: let me repeat I'm in no way out to destroy your game - whatever changes are made, the current situation could always be retained as an option (perhaps an "Ironman" checkbox?).

I really dont think this game is Hostage to the RNG

0-0 Ambushes 0-7 estimates 5-3 results : these really dont appear to be happen as much from the RNG you seem to hate, but from the basic game mechanics the roll of initiative etc. Do you know why Ambushes result in 0-0 sometimes?? Because the Ambusher gets NO attack bonus when it ambushes! Its the game mechanics at play, not the RNG

Well, you can say all you want how you dont want to destroy MY game, yet you want it to change to YOUR way and force ME to have to somehow mod it back, Sorry man, almost all games have some random, maybe people are seeing differnt things but I have been watching the actuall rolls vs predictions closely now and the amt of extrem results , in my experiance is minimal and not something that matters over the course of a battle. It would really ruin my game if the preidcitions were 100% accurate, and no they wernt 100% accurate in PG either. Jeez, I dont even use the predictions except as a general warning in case i failed to notice something important , Woah, why 7 - 0, oh woops , that unit is in level 6 entrechment and is backed by an artillery piece that my blurry eyes failed to notice. Maybe it wouild be better if you just got a green yellow or red light instead of #'s /odds

Why the rage mouser?

The key part of Molve's posts was the one where he pointed out it worked in PG - SSI's five star series had consistently reliable combat predictions. They still allowed for some margin of error, it was just much rarer than it is in PzC, where you can truly pretty much reduce it to GREEN, YELLOW and RED, because the finer nuances are worthless - there's simply too high a chance to get a 2-5 instead of the forecasted 0-8 result.

I'm sure nobody want's to throw out the RNG, it's just that in current PzC, it produces too many outliers.
All that's discussed is to broaden the range in which the predicted result actually becomes the end result.
So now, maybe there's a 30% chance for getting the predicted result, both 20% for a slightly better or slightly worse result (this can already mean taking losses with a 0-6 prediction), both 10% for either a disastrous or outstanding result (3-3 or 0-10) and 10% for special occasions (attacked unit doesn't even fire back, your attack completely fizzles and neither side takes any damage, etc.).
Would it really be so unreasonable to, say have a 50% chance of getting the result predicted, only 15% chances for slightly better or worse results, only 5% for disastrous or outstanding result and still 10% for the special occasions, just to make something up?
_____
rezaf

rezaf wrote:SSI's five star series had consistently reliable combat predictions. They still allowed for some margin of error, it was just much rarer than it is in PzC

Your memory must be failing.
Don't you remember especially air combat in Panzer General? It was 50:50, who shot first killed about 50-near 100% of the strength of the opponent.

Longasc wrote:

rezaf wrote:SSI's five star series had consistently reliable combat predictions. They still allowed for some margin of error, it was just much rarer than it is in PzC

Your memory must be failing.
Don't you remember especially air combat in Panzer General? It was 50:50, who shot first killed about 50-near 100% of the strength of the opponent.

Im with Longasc here..the predicted results were no better in PG infact they may have been worse because in PG if you shot 1st any kills and suppression casued did not shoot back, unlike in this game

@ Rezaf
No rage man, just shouldnt be posting in the AM where I havnt had my caffeine, not a morning person by any means.

@ Molve, I do apoligise because i realize my post , especially with the caps highliting things was certainly a lot more harsh , even rude, which not my intent. My bad man.

In the end though i hope you can you see my point , that possibly, its not the predictor nor necasarily the rng but the actual game mechanics at play here that thow the predicted values off? How can you predict what random bonus initiative you can get, which isnt calced before you commit to combat, and initiative plays a much bigger role that in prior Genral games.

Molve, i get your point that if ALL the #'s are predrolled then certainly the predictor could know what your final initiative is, but think of the ramifications. You can cursor over any unit(s) for the predictions but never have to commit to actually doing the combat What if you go out of order? Then the predictors will be off anyways. A general isnt supposed to know how all thing will play out.

I do concede though that maybe I have been lucky in that i dont see these wild results others are, or maybe my play style allows me to stubbornly ignore things, I dunno.

My own beef the engine ( which is making me wonder why bother continuing to play) is another story. Oh well, cheers!

TheGrayMouser wrote: A general isnt supposed to know how all thing will play out.

If he doesn’t – then he should be busted down to a 2nd lieutenant.

Longasc wrote:Your memory must be failing.

Unlikely, as I last played like 2 weeks ago.
Like I said, OF COURSE the results in PG had a definate amount of randomness to them, it's just that they were more likely to match the predictions given, that's all.
_____
rezaf

All I know is that Ike didn't know if OVerlord would result in a 7-0, 0-8, 5-3 or 0-0 result. He just went with his plan!

rezaf wrote:

Longasc wrote:Your memory must be failing.

Unlikely, as I last played like 2 weeks ago.
Like I said, OF COURSE the results in PG had a definate amount of randomness to them, it's just that they were more likely to match the predictions given, that's all.
_____
rezaf

I dont think they were more likely..I'd say it was the other way around especially with air combat

Well, I did some quick checks, and I think the beef of this might be related to the lessened impact of experience in combat and/or overstrength.
"Glorious" battle memories appearently come from gloriously experienced high-end units.

It's nothing bullet-proof, but I did some quick checks:

First, I attacked a Polish Inf with a Wehrmacht 39 in PG, 10 times, identical setup.
With a prediction of 1/6, the actual results were:
7/8; 0/7; 0/8; 0/7; 6/7; 6/0; 6/6; 6/1; 0/6 and 4/2.
After rounding, that's an average results of 4/5, the high friendly casualty ratio caused by some MASSIVE outliers.

Next, I attacked a dug in AT with a PzIV. Prediction 0/4. Results:
1/4; 0/5; 1/1; 0/5; 0/5; 2/2; 5/2; 1/4; 5/0 and 0/5.
That's a mere 2/3 after rounding, which is a somewhat horrible balance, but what should one expect from a frontal assault on AT guns with PzIVs...
Still WAY off the predictions.

Now, I selected something a bit more experienced - a 2 star XP Tiger tank, albeit without overstrength, vs. some russian INF.
Prediction 0/5. Results:
0/6; 1/4; 0/4; 0/5; 1/5; 1/4; 0/7; 0/6; 0/6 and 1/2.
After rounding, that's 0/5. Pretty close to the prediction.

Running tests in PzC is somewhat more complicated as reloading does not reset the RNG seed, but I created a blank map and placed 10 Wehrmacht next to 10 Polish Inf, slightly spaced out.
Combat predictions were 1/3 in all cases. Results:
1/1; 1/3; 0/6; 1/4; 0/6; 2/4; 1/2; 0/6; 2/2 and 0/4.
That's actually an 1/4 after rounding, which isn't only pretty accurate but even favorable to the player.

So, how about facing off a 2 star Tiger vs. weakling polish INF in PzC? Predictions: 0/6. Results:
0/6; 0/6; 0/6; 0/6; 0/4; 0/5; 0/7; 0/8; 0/8; 0/8.
On average, 0/6. Pretty close as well, I guess.

I'll do some tests with overstrength units, but without OS and excessive XP, the RNG in PzC, despite my own claims, doesn't seem so bad after all.
I will definately keep a closer look on this issue when I get a chance to playtests some levels of my campaign again, but I guess I just undermined my own position...
_____
rezaf

Interesting , from your results , it appears the ones most "off" were between units most vulnerable to losing the initiative.

Probobly the best way to test this would be to temporarily mod the data file and Turn OFF the 10(20% ?) bonus or "free attack" that you get for every one point of initiative higher between attacker and defender. That way , as best as possible, the rng will be isolated.

TheGrayMouser wrote:Interesting , from your results , it appears the ones most "off" were between units most vulnerable to losing the initiative.

Probobly the best way to test this would be to temporarily mod the data file and Turn OFF the 10(20% ?) bonus or "free attack" that you get for every one point of initiative higher between attacker and defender. That way , as best as possible, the rng will be isolated.

But the odds take into account initiative difference...otherwise why would they change when you use mass atack?

EDIT: And from my understanding 20% of all kills and suppressed hits dont shoot back for every point difference on iniative...so if you have 5 higher after rolls etc then only unsuppressed points will get to return fire

Because in massed attacks the intiative bonus is a known and fixed amount. What is not accounted for is the random 0 to +2 intiatiave that both attacker and defenders might get and , since that roll doesnt happen until you actually attack, it cant be part of the predictor ( uhh, I think!)

You know this whole discussion kinda ties into the thread that experiance doesnt matter much

Im not sure but based on the data files on how a unit gains values for every bar of experiance ( iether 10% or +1 whichevr is lower) its seams that infantry get the short end of the stick due to rounding. ie units with starting high levels likly get a point for every bar , but due to rounding , units with low starting values skip levels of experiance with no benefit

Example: Infantry get 2 intiative , at what point does an infantry unit get an additional point ? 2 bars 3? at what point 2 additioal , 4 ?
A tsnk with 12 starting intiative likly gets addtitional points more evenly and thus more!

So an infantry attacked by a tank is kinda hosed here, even if the infantry is in close terrain/entrenched and there is an initiative cap. This is because the cap can be bypassed by 2 things A the experiance bonus and B the random 0 thru +2 roll.

So tanks that benefit by getting additional initiative points from experiance , plus the roll often times win the intitiatve which is very very improtant in this game as you highlight above. Unlike other values like attack ground defence etc, where the differnce beteen 8 and 10, say for hard attack is important but isnt huge, having two point higher in intiative IS because the game uses that as an absolote comparative: every point allows 20% hits to take effect before the defender can fire back!.

This is my gripe that I alluded to about this game, partly because of the above and partly other items that make infantry very very weak IMHO.