PLEASE PLEASE PLEASE Fix the odds predictor

[Longasc]

That's my solution ,,as a matter of fact my computer almost went to jump school yesterday . I came real close to throwing it out the window

I wonder what the P.L.F would be for a computer .......

Wait till you lose all your Stukas in one scenario and a ton of tanks and infantry as well. This might happen to you in an upcoming DLC.
At this point you will agree that those Hitler Rage videos on Youtube are not exaggerated.

[boatie]

In my view, this whole issue is not about randomness but about trust. If the player does not trust the RNG (as stated in many comments in this and other threads), they do not trust the underlying mechanics of the game.

If they can't trust the mechanics of the game, this destroys their sense of immersion and suspension of disbelief. So when results that appear intuitively stupid occur, this can ruin the player's overall enjoyment of the game. Once that happens, its hard to get it back and the alternative is not to accept the rules as it, but to find a better game. I am in this camp. If the game does not "feel" fun, but seem more like I am fighting it to have fun, something is wrong.

The point of the game should be to provide enjoyment. It's obvious that the current RNG is not living up to that standard for a significant number of players - how many threads has been started on this already?

Also, given the scaling implemented in the game, each unit does not represent a single unit - one tank unit in the game actually represent a division (or smaller size), extreme variance should be perceive as being counter intuitive. Add in the turn restrictions and the variance in the RNG can make the game understandably frustrating.

[robc04_1]

In my view, this whole issue is not about randomness but about trust. If the player does not trust the RNG (as stated in many comments in this and other threads), they do not trust the underlying mechanics of the game.

If they can't trust the mechanics of the game, this destroys their sense of immersion and suspension of disbelief. So when results that appear intuitively stupid occur, this can ruin the player's overall enjoyment of the game. Once that happens, its hard to get it back and the alternative is not to accept the rules as it, but to find a better game. I am in this camp. If the game does not "feel" fun, but seem more like I am fighting it to have fun, something is wrong.

The point of the game should be to provide enjoyment. It's obvious that the current RNG is not living up to that standard for a significant number of players - how many threads has been started on this already?

Also, given the scaling implemented in the game, each unit does not represent a single unit - one tank unit in the game actually represent a division (or smaller size), extreme variance should be perceive as being counter intuitive. Add in the turn restrictions and the variance in the RNG can make the game understandably frustrating.

Do players actually understand how much variance can be mathmatically expected? For example let's say unit x has a 10% chance of killing unit y. If we performed this combat 50 times we would expect unit x to win 5 times. Using random numbers to model this we probably won't get unit x to win exactly 5 times. Using the formula to compute the variance:
square root(n x probability x (1 - probability))
we get
square root (50 x .1 x (1 - .1)) = 2.12

So 68% of the time we should get unit x killing unit y between 5 - 2 and 5 + 2 (3 to 7) times.
95% of the time we should get unit x killing unit y between 5 - 4 and 5 + 4 (1 to 9) times
99.7% of the time we should get unit x killing unit y between 5 - 6 and 5 + 6 (0 to 11) times

So if you look at the 95% range, that means 5 % of the time we can expect unit x to kill unit y at least 10 times. This is twice the probability of 10%.

I'm not sure if people are experiencing results outside of this range or not, but it isn't unusual to get unexpected results. Also we tend to remember the results we didn't expect instead of the overwhelming majority of those that we did.

In any case, assuming there isn't a bug in the game's calculations Slitherine shouldn't fudge the math if people either don't understand how random numbers work or there are perception problems.

[vasilak]

Do players actually understand how much variance can be mathmatically expected? For example let's say unit x has a 10% chance of killing unit y. If we performed this combat 50 times we would expect unit x to win 5 times. Using random numbers to model this we probably won't get unit x to win exactly 5 times. Using the formula to compute the variance:
square root(n x probability x (1 - probability))
we get
square root (50 x .1 x (1 - .1)) = 2.12

So 68% of the time we should get unit x killing unit y between 5 - 2 and 5 + 2 (3 to 7) times.
95% of the time we should get unit x killing unit y between 5 - 4 and 5 + 4 (1 to 9) times
99.7% of the time we should get unit x killing unit y between 5 - 6 and 5 + 6 (0 to 11) times

So if you look at the 95% range, that means 5 % of the time we can expect unit x to kill unit y at least 10 times. This is twice the probability of 10%.

I'm not sure if people are experiencing results outside of this range or not, but it isn't unusual to get unexpected results. Also we tend to remember the results we didn't expect instead of the overwhelming majority of those that we did.

In any case, assuming there isn't a bug in the game's calculations Slitherine shouldn't fudge the math if people either don't understand how random numbers work or there are perception problems.

Hi.
The issue here is not if users understand the mathematics behind the implementation of a RNG. It is not even "how much variance can be mathematically expected".
From my understanding, the issue is simply what the users expect of the PREDICTION PROVIDED by the game's interface.
For example taking 10 outcomes where the 5 are 0x10 and the other 5 are 10x0 would give an accurate prediction of 5x5.
However, if the user sees a 0x10 and further down the road a 10x0 I bet he/she would not think "OK, the prediction was accurate, so everything is fine".
Instead, the user would most probably think that the prediction, even if mathematically sound, was simply misleading.
In this case, how much the outcome will vary from the prediction given is as important (sometimes even more so) as the prediction itself.

[VPaulus]

I'm not sure if people are experiencing results outside of this range or not, but it isn't unusual to get unexpected results. Also we tend to remember the results we didn't expect instead of the overwhelming majority of those that we did.

This was already told a lot of times, even by Rudankort. He have done thousands of tests and simulations, and the results are not outside the range expected.

[IainMcNeil]

Good explanation - It is all working as intended as we have said a number of times and we do not plan to change it as the results would dramatically change the way the game plays. Probabilty is a strange thing.

If you have a 10% chance of something happening, and you try 5 times what is your chance of success.

50%?

No.

The only way to work it out is to work out the chance of it not happening. On the first go its 90%. On the seconds go 90% etc. This gives you a 90%x90%x90%x90%x90% chance of it not happening = 59%. This gives you a 41% chance of it happening. This is just a simple yes/no test.

This gives an idea of how hard it is to predict results in your head and how hard it would be to show them to the player.

Take away the randomness and you make a unit always hit or always miss so results are 10-0 every time with even the slightest difference in stats. Add the randomness in and you end up back where we are now And this is why we are where we are now and will not be changing!

[vasilak]

Good explanation - It is all working as intended as we have said a number of times and we do not plan to change it as the results would dramatically change the way the game plays. Probabilty is a strange thing.

If you have a 10% chance of something happening, and you try 5 times what is your chance of success.

50%?

No.

The only way to work it out is to work out the chance of it not happening. On the first go its 90%. On the seconds go 90% etc. This gives you a 90%x90%x90%x90%x90% chance of it not happening = 59%. This gives you a 41% chance of it happening. This is just a simple yes/no test.

This gives an idea of how hard it is to predict results in your head and how hard it would be to show them to the player.

Take away the randomness and you make a unit always hit or always miss so results are 10-0 every time with even the slightest difference in stats. Add the randomness in and you end up back where we are now And this is why we are where we are now and will not be changing!

Hi Ian.
First let me say that I understand that this is closed, no argument there.
My personal view as I explained is not the the system does not work as it would be mathematically expected, and therefore incorrectly.
Rather, it is a matter of what the user expects given the prediction by the system. So first and foremost, to me, it's a UI issue of not giving the user to fully understand what to expect.
And in this case, it is vital to let the user know not only the average outcome, but also how much the actual outcome is likely to vary from the prediction.

[IainMcNeil]

Thats a bit subjective! I don't think it works how you think it works.

If you wanted you could ask a specific question and get a % answer, but you need a very specific question.

E.g. do you want to know the chance of your casualty prediction being correct? Do you want to know the chance it is within 1 of being right? Do you want to know if it is correct or better, correct or worse, just worse, just better? If better, better by 1 or 2 or etc. You can only ask one of these questions. That doesnt even think about whether you care about enemy casualties

There are an almost inifinte number of questions you could ask and they will all have a diffrent probability, and you can only ask one.

Then you would have to try and explain to players which question it is you are trying to answer.

And some of them woudl want to know why you hadn't chosen a different question and be in the same position we are now

[vasilak]

Thats a bit subjective! I don't think it works how you think it works.

If you wanted you could ask a specific question and get a % answer, but you need a very specific question.

E.g. do you want to know the chance of your casualty prediction being correct? Do you want to know the chance it is within 1 of being right? Do you want to know if it is correct or better, correct or worse, just worse, just better? If better, better by 1 or 2 or etc. You can only ask one of these questions. That doesnt even think about whether you care about enemy casualties

There are an almost inifinte number of questions you could ask and they will all have a diffrent probability, and you can only ask one.

Then you would have to try and explain to players which question it is you are trying to answer.

And some of them woudl want to know why you hadn't chosen a different question and be in the same position we are now

I understand what you are saying and this is exactly where the issue is.
Indeed you can only show the answer to only 1 or possibly 2 or 3 (if done intuitively) of these questions to the users - I am not saying this is easy to do in an intuitive way.
However, I think a player would not be interested in 10 questions. But he/she would probably be interested in 3! I am not sure which 3 of course, and you are not either I guess.
But I think we can make an educated guess with the help of the community.
A first try would be:
1: Give a prediction of what's going to happen (already done).
2. Give me the maximum casualties that could be inflicted on both friendly and enemy units.
3. Give me the probability of the outcome being close to the prediction. Obviously, this is too general. But this could be for example +-1 casualty for each of the two units. I think this might be the most important statistic besides the prediction because it gives an idea of how probable an outcome close to the prediction is.

[robc04_1]

I'm not sure if people are experiencing results outside of this range or not, but it isn't unusual to get unexpected results. Also we tend to remember the results we didn't expect instead of the overwhelming majority of those that we did.

This was already told a lot of times, even by Rudankort. He have done thousands of tests and simulations, and the results are not outside the range expected.

I believe it. I don't have any complaints with the system.

[brettz123]

Any game that uses random numbers is going to have times where there is significant deviation from the norm. For example in a computer baseball game, if a batter has an expected .333 batting average in the simulation engine, they only will hit between .313. and .353 68% of the time in 600 at bats if random numbers are used to generate the results. That leaves room for a lot of unexpected results. It is just the nature of random numbers.

See to me this is HIGHLY disingenuous. The problem is that there is variation. The problem isn't even that there can be wild variations. The problem is that the combat predictor is WRONG much more often than it is right. This makes no sense. The combat predictor should be close to right most of the time but instead it is wrong by at leat 2 damage more than 50% of the time.

[brettz123]

Holy cow, player from day one and NOT used to random results.

Yeah, it was already debated ad nauseam.

How about not throwing a tantrum when things go wrong and planning with the possibility of bad results in mind?

No one is throwing a tantrum what we are saying is that the predictor should be at least close to correct more than not. And frankly many of us got the game well after this ad nauseum you speak of and would like to talk about it again (for the first time ). While I have no problem even with wild swings they should not be the rule but the exeption.

[airbornemongo101]

I have a suggestion(s)/thoughts on the RNG and the "wild results"

Think of the "wild swings" as the result of enemy action.

In any conflict there are heroes on both sides (and both sides know their side is the right side),,thus within the enemy formations there will be soldiers who go above and beyond.

History is replete of examples where troops held out against all odds or even tho they succumbed they damaged the opposing force out of proportion to their numbers,,or troops who won on the offensive when every thing and everyone said they had no chance.

Don't forget this either:

No matter how well laid a plan ,,Mr. Murphy and the Laws of Chaos (no, not the Warhammer Chaos ) are going to rear their ugly heads.

IE: You tell your Pz. Rgt to engage a a enemy Inf. Rgt. in an open hex there are no enemy units near it,,so it has with no support.

Now bear with me,, imagine the below:

While trying to find the obj. a whole battalion of your Pz. Rgt gets lost . Within that enemy Inf. Rgt. there just happens to be a bunch of grunts who are very adept at anti-tank tactics. Ouch..your 4 (you) to 1 (Inf.) prediction just got turned around and you just lost 40% of that Pz. Rgt.,,,meanwhile the grunts of that enemy Rgt. are laughing their little digital 4th pts. of contact off at you.

That's the way I've always looked at the random number generator.That in real life there is always actions that you cannot control . You plan and you give orders and then it's up to the individual unit commanders to implement your plans well.

I personally have noticed that it seems the wild swings in the AI's favor happen less as your troops gain more experience,,again it seems that way to me,,which makes sense in the fact that your troops are going to make less mistakes (getting lost..etc) and recognize enemy plans before the enemy can implement them.

This way of thinking has helped me when I see a result that really,,really pisses me off. I think of the above (or similiar thoughts) and I plan again.

The RNG lives up to it's name,,it gives random results,which imho ,while not truly reflecting true life circumstances/variables are close enough.

Besides for those who want the numbers to come out "right" every time there is the "chess cheat" that can be implmented.

Just my $0.02

[brettz123]

I have a suggestion(s)/thoughts on the RNG and the "wild results"

Think of the "wild swings" as the result of enemy action.

In any conflict there are heroes on both sides (and both sides know their side is the right side),,thus within the enemy formations there will be soldiers who go above and beyond.

History is replete of examples where troops held out against all odds or even tho they succumbed they damaged the opposing force out of proportion to their numbers,,or troops who won on the offensive when every thing and everyone said they had no chance.

Don't forget this either:

No matter how well laid a plan ,,Mr. Murphy and the Laws of Chaos (no, not the Warhammer Chaos ) are going to rear their ugly heads.

IE: You tell your Pz. Rgt to engage a a enemy Inf. Rgt. in an open hex there are no enemy units near it,,so it has with no support.

Now bear with me,, imagine the below:

While trying to find the obj. a whole battalion of your Pz. Rgt gets lost . Within that enemy Inf. Rgt. there just happens to be a bunch of grunts who are very adept at anti-tank tactics. Ouch..your 4 (you) to 1 (Inf.) prediction just got turned around and you just lost 40% of that Pz. Rgt.,,,meanwhile the grunts of that enemy Rgt. are laughing their little digital 4th pts. of contact off at you.

That's the way I've always looked at the random number generator.That in real life there is always actions that you cannot control . You plan and you give orders and then it's up to the individual unit commanders to implement your plans well.

I personally have noticed that it seems the wild swings in the AI's favor happen less as your troops gain more experience,,again it seems that way to me,,which makes sense in the fact that your troops are going to make less mistakes (getting lost..etc) and recognize enemy plans before the enemy can implement them.

This way of thinking has helped me when I see a result that really,,really pisses me off. I think of the above (or similiar thoughts) and I plan again.

The RNG lives up to it's name,,it gives random results,which imho ,while not truly reflecting true life circumstances/variables are close enough.

Besides for those who want the numbers to come out "right" every time there is the "chess cheat" that can be implmented.

Just my $0.02

Yes, I agree with the basic point of your post but that isn't what we are talking about. I think everyone understands the logic behind wild swings and I think most of us agree there should be a possibility of these things happening because as youy pointed out it is realistic. the porblem is that the predictor should tell you the most likely outcome (at least in my opinion. otherwise it is fairly useless. Now it isn't totally useless because even though it is most often wrong it is usually fairly close. But still it should be right more often than wrong. And to be honest I have not noticed it being right more often with my experienced troops. Though I find complete whiffs where no damage is done much more prevalent than one unit or the other getting massacred, which does still occasionally happen.

Still interesting conversation.

[airbornemongo101]

Thanks brettzz

I failed to mention this in my post,,I had actually typed the below in and I guess deleted it

The predictions would follow the same logic,,the results shown would be if everything went as planned

Throw in Mr Murphy,,unexpected enemy actions,,and other variables (inexperience..etc) and you get the different predictions/results.

Thus the failure of "true" predictions sometimes.

[IainMcNeil]

The problem is that any conbat has a chance of 10 hits for either side so showing the max hits would always be the same - 10.

The chance of getting the predicted results is very low because its one possible out come of 121. While it is the most likely single outcome, the number of other possible outcomes means it is more often than not going to be wrong.

I still think there is a fundamental misunderstanding of how probability works.

[brettz123]

The problem is that any conbat has a chance of 10 hits for either side so showing the max hits would always be the same - 10.

The chance of getting the predicted results is very low because its one possible out come of 121. While it is the most likely single outcome, the number of other possible outcomes means it is more often than not going to be wrong.

I still think there is a fundamental misunderstanding of how probability works.

I think we understand it but the way it works literally makes it almost useless unfortunately. I don't really think it is a big deal and most of us learn to deal with it but it just nags at the back of the brain how it is usually wrong.

[dragos]

I can't understand how the random combat results is a show stopper for some players. Is the game too hard? Can't be, because with proper planning on FM level you can complete scenarios without losing a single unit. The AI is very predictable and if you don't make risky moves you can suffer minimum casualties. IMO the game is imbalanced in favor of human player. Look at the casualties window at the end of a campaign to see how disproportionate are the losses of the AI compared to human player.

[vasilak]

The problem is that any conbat has a chance of 10 hits for either side so showing the max hits would always be the same - 10.

The chance of getting the predicted results is very low because its one possible out come of 121. While it is the most likely single outcome, the number of other possible outcomes means it is more often than not going to be wrong.

I still think there is a fundamental misunderstanding of how probability works.

Personally I think I understand how probability works (I have a degree in Computer Science after all ).
The point we are trying to make is not how the RNG should work.
Probably it works fine, but in any case that's something you need to think about who understand what has been implemented.
The point is how useful the information provided to the user is, and that's the only feedback we as users can provide.
A lot of people's feedback simply says: to the average user the prediction provided is simply not trustworthy enough.
We could play the game with no prediction at all. The prediction simply makes it more user friendly.
So the point goes: the current prediction system simply does not make the game as user friendly as some people would like.

[brettz123]

Personally I think I understand how probability works (I have a degree in Computer Science after all ).
The point we are trying to make is not how the RNG should work.
Probably it works fine, but in any case that's something you need to think about who understand what has been implemented.
The point is how useful the information provided to the user is, and that's the only feedback we as users can provide.
A lot of people's feedback simply says: to the average user the prediction provided is simply not trustworthy enough.
We could play the game with no prediction at all. The prediction simply makes it more user friendly.
So the point goes: the current prediction system simply does not make the game as user friendly as some people would like.

Very well said.