News from Beta - May Contain Spoilers!

[Deleted User - 1178646]

As someone who does 30-40 provinces for 5-6 rounds every week, some of these changes will drastically impact my interest in maintaining that level of gameplay if I start to hit an impenetrable wall of difficulty just to keep doing what I currently can. What incentive will I have? The 9 extra fellowship chests offering 10 KP each? That's nothing compared to what I currently earn individually.

I think we have to accept that they want us to do less provinces similar to pre spire?
The main issue now is damage control, to make sure a bit less doesn't become none at all.

 

[TheGiftedKes]

Therefore I believe that this formula will be at least as flawed as the spire one, and as a result most likely will make it impossible for me to keep playing and having fun.

I saw the formula on the beta thread, but I am not on beta so couldn't reply there. While I think @MinMax Gamer has done a great job, just wanted to make sure it is clear that the formula is one of infinite solutions possible with the same or even less error. Many of those solutions not needing to have an exponential term at all. I assume the presence of needed variables in the formula is confirmed through a process of changing only one parameter at a time, but when fitting non-linear models with multiple parameters to merely 200 datapoints, it is quite clear that the form of the formula is pretty arbitrary. The true formula need not even be a smooth function.

Anyway, I don't necessarily like the changes proposed for tourney, but wanted to make sure that formula is not treated as ground truth and be extrapolated too much in the arguments

 

[Deleted User - 1178646]

I saw the formula on the beta thread, but I am not on beta so couldn't reply there. While I think @MinMax Gamer has done a great job, just wanted to make sure it is clear that the formula is one of infinite solutions possible with the same or even less error. Many of those solutions not needing to have an exponential term at all. I assume the presence of needed variables in the formula is confirmed through a process of changing only one parameter at a time, but when fitting non-linear models with multiple parameters to merely 200 datapoints, it is quite clear that the form of the formula is pretty arbitrary. The true formula need not even be a smooth function.

Anyway, I don't necessarily like the changes proposed for tourney, but wanted to make sure that formula is not treated as ground truth and be extrapolated too much in the arguments

There are different ways to write the same thing,
The real formula might be the same one differently written, but I am confident dat after so many weeks and with the help of so many people (it was way ways way more than 200 datapoints lol) the formula is at least correct within the margin of error.
It was confirmed by many it actually worked and math doesn't lie.

Lets not try to create chaos by questioning his work, I think your country has enough of those problems already without adding a elvenar formula to the mix.

If it was incorrect, then A people would have found them and B more importantly the elvenar team would have openly told us it was wrong.
Because the conlusions from that formula actually hurts them from a PR standpoint. so if it was incorrect they would have told us so.

 

[Deleted User - 3932582]

but when fitting non-linear models with multiple parameters to merely 200 datapoints, it is quite clear that the form of the formula is pretty arbitrary. The true formula need not even be a smooth function

LOL, really? So you can make a materially different curve through 200+ points with only 5 independent variables, without drastically increasing dimensionality (e.g. overfitting, which will immediately break with any new data points)? OK, give me at least one other (out of "infinite") solution to that.

And yes, exponential term is there for a reason. Once you control for other variables, dependency on research is not polynomial - it is exponential. So let's not just make some statements without backing them up, shall we?

 

[TheGiftedKes]

LOL, really? So you can make a materially different curve through 200+ points with only 5 independent variables, without drastically increasing dimensionality (e.g. overfitting, which will immediately break with any new data points)? OK, give me at least one other (out of "infinite") solution to that.

And yes, exponential term is there for a reason. Once you control for other variables, dependency on research is not polynomial - it is exponential. So let's not just make some statements without backing them up, shall we?

There are different ways to write the same thing,
The real formula might be the same one differently written, but I am confident dat after so many weeks and with the help of so many people (it was way ways way more than 200 datapoints lol) the formula is at least correct within the margin of error.
It was confirmed by many it actually worked and math doesn't lie.

Lets not try to create chaos by questioning his work, I think your country has enough of those problems already without adding a elvenar formula to the mix.

If it was incorrect, then A people would have found them and B more importantly the elvenar team would have openly told us it was wrong.
Because the conlusions from that formula actually hurts them from a PR standpoint. so if it was incorrect they would have told us so.

As I said the formula is a fit, out of many. Your argument was all based on the exact form of it so I cautioned you

 

[Deleted User - 3932582]

As I said the formula is a fit, out of many. Your argument was all based on the exact form of it so I cautioned you

I still haven't seen any single other one out of "many".

 

[TheGiftedKes]

LOL, really? So you can make a materially different curve through 200+ points with only 5 independent variables, without drastically increasing dimensionality (e.g. overfitting, which will immediately break with any new data points)? OK, give me at least one other (out of "infinite") solution to that.

And yes, exponential term is there for a reason. Once you control for other variables, dependency on research is not polynomial - it is exponential. So let's not just make some statements without backing them up, shall we?

I have a PhD in stats and multiple published articles in peer reviewed journals to back me up. As I said you have done a great job, and I didn't mean to disrespect your work. If you insist, you can share the data and I can provide a couple of other example solutions.

 

[Pheryll]

I have a PhD in stats and multiple published articles in peer reviewed journals to back me up. As I said you have done a great job, and I didn't mean to disrespect your work. If you insist, you can share the data and I can provide a couple of other example solutions.

The fit was close enough that @MinMax Gamer could detect accurately when people had made small errors in counting (even off by one). This is not like training speed where the numbers are assigned individually, so that calculating a curve to represent the data would not only be overly complicated but also would have no predictive power for the next number they assign. MinMax Gamer's numbers do have predictive power.

 

[Deleted User - 3932582]

I have a PhD in stats and multiple published articles in peer reviewed journals to back me up.

Without seeing ANY data, someone with PhD in stats wouldn't make statements like "it is clear that the formula is one of infinite solutions possible with the same or even less error. Many of those solutions not needing to have an exponential term at all". The current model literally has zero errors on all observations so far (zero to the extent that we can observe, i.e. less than rounding error), so you can't have less (observable) error than that. And then the only way to make a "better", or at least "equivalent" model would be to have another model with also zero error and fewer or at least the same number of degrees of freedom (keeping it parsimonous). And given that DF is very small...

I am sorry, but your statements do not check out.

 

[TheGiftedKes]

Without seeing ANY data, someone with PhD in stats wouldn't make statements like "it is clear that the formula is one of infinite solutions possible with the same or even less error. Many of those solutions not needing to have an exponential term at all". The current model literally has zero errors on all observations so far (zero to the extent that we can observe, i.e. less than rounding error), so you can't have less (observable) error than that. And then the only way to make a "better", or at least "equivalent" model would be to have another model with also zero error and fewer or at least the same number of degrees of freedom (keeping it parsimonous). And given that DF is very small...

I am sorry, but your statements do not check out.

I thought I saw a bar graph of residuals were there was some level of error. In which case, my statement would follow. If there is literally zero error, then yes within the current range of data you have captured even the correct form of non-linearity (assuming smoothness). I apologize for not looking at the correct version apparently.

Regarding your comment about me being a liar (politely), you are free to believe what you please. You can ask your professor at some point about the validity of my comment under the non-zero error assumption

 

[Pheryll]

I thought I saw a bar graph of residuals were there was some level of error.

The only graph in the Beta thread close to that is Edeba's grid for estimating the enemy's proportional strength. But that grid is estimating the color hue and assigning arbitrary values. What was described in the announcement paragraph above their color coded chart was that the diagonals of 6-1 to 1-6 and 20-1 to 15-6 corresponded to the lower and upper extremes of the system in place before (0.85 to 1.6), and these values do not match the ones Edeba used in the chart.

 

[Pheryll]

You can ask your professor at some point about the validity of my comment under the non-zero error assumption

Let's keep the conversation civil here.

 

[Deleted User - 3932582]

Regarding your comment about me being a liar (politely), you are free to believe what you please. You can ask your professor at some point about the validity of my comment under the non-zero error assumption

I wouldn't be able to do that as most of my professors passed away a long time ago. And I hope that you use higher standards of academic rigor in your publications, rather than using assumptions that are known to be not correct. So not sure why this assumption is relevant here.

 

[TheGiftedKes]

The only graph in the Beta thread close to that is Edeba's grid for estimating the enemy's proportional strength. But that grid is estimating the color hue and assigning arbitrary values. What was described in the announcement paragraph above their color coded chart was that the diagonals of 6-1 to 1-6 and 20-1 to 15-6 corresponded to the lower and upper extremes of the system in place before (0.85 to 1.6), and these values do not match the ones Edeba used in the chart.

I was referring to a graph I saw when I clicked on the link to MinMax's blog. I saw the formula the CrazyWizard had put on beta thread close to it so I assumed that the residuals bar plot was for the model being discussed. Anyway, if there are no residuals then never mind.

 

[TheGiftedKes]

I wouldn't be able to do that as most of my professors passed away a long time ago. And I hope that you use higher standards of academic rigor in your publications, rather than using assumptions that are known to be not correct. So not sure why this assumption is relevant here.

The assumption was indeed wrong and I apologized for it. The relevance was because you said "Without seeing ANY data, someone with PhD in stats wouldn't make statements ...". If there was indeed some error in the fit, there was no need for me to have any more info about the data. If you don't believe that verify it with a statistician, if you care. If your comment was not meant to be sarcastic then it's all good.

 

[Black watch]

I hate to state the obvious here... but ONE of the things that folks asked INNO to do, was to provide at least a thumbnail image of the terrain for helping speed up fighting in the tourneys... a bloody stupid scouting helper to speed things up. Do we have that now? NO- no they revamp the whole damn thing. Can anyone say "throw the baby out with the bathwater?"
Now as far as I can see (and many other's here TOO) this proposed change not only makes most of what's been learned and found to be helpful in growing our cities and fighting in the tourneys- is being thrown out on its ear.
Great- one fight... I like that speed up idea until you figure that One fight is the equivalent of four and is now forcing us to scout to make sure we don't loose each of those fights. So so much for saving time.
Since we apparently won't be able to know what's coming at us, we can't scout with an appropriate unit. So we're getting losses right from the get go.
I used to be a damn fine fighter in this game with an average well passed 5,000. I enjoyed playing this game. But in the last few weeks, the spire has ramped up its difficulty in a huge way that even boosted nicely doesn't cushion the losses. Tomorrow I'll be fighting in the spire without the boosts and I'm expecting HEAVY losses. If the tourneys turn out to be just another spire experince, then I'm probably going to leave the game.

I'm getting tired of having more and more and more COMPLICATED things thrown at me. I'm having to relearn just about everything with each chapter, with adjustments in every fighting area. What the hell do we need that cluster **** of a catering thing thrown at us. Just... never mind, no, never mind... they won't listen anyway...
I'm disgusted and I'm now going back to my clan and shake my head... my response to them is... "Hey, INNO's lost their mind!
I work for a living... I don't have time to keep up with a game that keeps changing the rules and making things more and more complicated for no real reason.
If you have any doubts as to what I'm saying. The folks that posted just before me are posting about data algorithms that are trying to figure things out.
It's only getting worse...

 

[Ashrem]

I hate to state the obvious here... but ONE of the things that folks asked INNO to do, was to provide at least a thumbnail image of the terrain for helping speed up fighting in the tourneys... a bloody stupid scouting helper to speed things up. Do we have that now? NO- no they revamp the whole damn thing. Can anyone say "throw the baby out with the bathwater?"

To be fair, we only asked for the thumbnail view a few months ago. People voted for fewer encounters two years ago, and have been comlaining about optional SS techs and tournaments for longer than that.

Make no mistake, this will be better for retaining new players, and will help goups with small players do better in the tournaments. Generally speaking, it's good changes. There will be some pain, and I'm not content with the penalty hitting players who decided not to take Optional SSs based on giving up strength in province exploration for strength in tournament.

 

[Deleted User - 3932582]

I've made a wild speculation analysis commentary on the incoming changes (and forum comments) for my FS. Instead of typing in a bunch of text that will be obsolete very soon, I decided to make a couple of videos which are really podcasts (e.g. there is not much to see there, you may just listen to it while doing other things). I go through the announcement and some forum comments, and provide my initial reaction. I figured some other people may also find it useful

Spoiler: Tournament Changes 2020

 

[Deleted User - 1178646]

To be fair, we only asked for the thumbnail view a few months ago. People voted for fewer encounters two years ago, and have been comlaining about optional SS techs and tournaments for longer than that.

Make no mistake, this will be better for retaining new players, and will help goups with small players do better in the tournaments. Generally speaking, it's good changes. There will be some pain, and I'm not content with the penalty hitting players who decided not to take Optional SSs based on giving up strength in province exploration for strength in tournament.

I do not think this has anyting to do with player retention.
Nor will it help small players to do better in the tournaments.

A person on the duch forum said the following:

In the past you could learn how to fight, the fights were laser focussed and if you made a mistake in 1 fight you could learn and improve in the next fight. this knowledge could then be applied in the future.
With the new random fights, each fight is a challenge on it's own, instead of learning a portion of fighting you now have to master it all at once.
Since recurring is low, learning from your mistakes is also low as by the time a similar fight returns so much have been forgotten that you no longer have the knowledge of former mistakes.

I think this is a good point that person made.

 

[mikeledo]

Gems will be useless for the tournament. It was mentioned in the video that nothing you could do could make things worse. If you increase your ability 10%, the strength of troops will only go up 5%. I am thinking more will still be better and add those squad sizes when the event asks for a tech tree update. Perhaps a way to get around this is to have less manufacturing and less WS and use more spells and event/evolve buildings. This should make battles easier. The downside is that you will lose ranking points, the same ones you try to get by scoring high. Where are our numbers people?

Folks spent time and money to build a superior city one that caters to INNO's set up. Now INNO is changing the rules to make those who built all their AWS up to 30 gazillion less effective. Between this change and the various event buildings, AWs are becoming less and less important making push accounts less effective.