Almost 2 full years of this probability mumbo jumbo and this single post made my brain explode!!
I FINALLY get it!!!
You know what @Lyya? Somebody should write a song about you… Oh, nevermind! 
2 Likes
So it’s basically just complete randomness and chaos…got it…lol.
So, just to reiterate, we’re not touching any RNG in the battle game outside of 2 areas:
- Random Gem Spawners that start to get too many turns in a row
- Altering AI luck so that the AI can not get too much “luckier” than the human player with regard to gem drops
But, that aside, here’s some food for thought when dealing with large systems.
- Well in excess of a million games of Gems of War are played every day, but let’s assume it’s just one million
- In each one of those games, there are probably on average 20 total turns
- That would leave us with 20,000,000 sets of Gems spawning in
- If the chance of a weird streak of Gems falling were even as high as 1 in 1,000,000 that means it STILL happens 20 times per day, 10 of which are probably against a human player.
- That’s 70 bad spawns in a week against human players.
- How many people report that? Not many - maybe 1 or 2.
- But when you see 1 or 2 posts per week on social/forums, it’s only natural to start to think there might be shenanigans… but when you look at the scale of game, maybe not so much…
Note that is a very general example of scale… I’m not expressly saying the things people see are 1,000,000 chance, or that all games are 20 turns, or any such thing… it’s intended to illustrative by citing a single fictitious situation only.
Tl;DR - If it can happen, it WILL happen.
BUT… We CAN employ various means to try and reduce streakiness, which we might look at in future.
15 Likes
I’ve never had a problem with gem spawns. Sure, the ai will get some good ones every so often, but so do i. My only problem was resurrect with tds when it states 25%, which we all know is not accurate, at least for the defense teams.
Of course a 25% chance can fire five times in a row. It shouldn’t happen often, but it absolutely can happen. By that other logic, nobody can ever win a lottery, because the odds that any one person has a winning ticket are much smaller than 1 in 1000.
Thank you for your answer!
But I was not pointing at bad streaks or cascades - even if I don’t like them, expeciallly when they happen to me, I can grasp the concept of bad luck.
I was pointing at problems with looping, that make a lot of troops meant to work looping unplayable. When I have weeks of randomness overwhelmingly playing in a way followed by weeks of the same troops performing in the opposite way, then I have to wonder. I may point to Dark Troll/Rock Troll loops, and to Lady Anariel and Jarl Firemantle because I play them a lot.
Random Gem Spawners that start to get too many turns in a row
It would be nice to know how many loops are ‘too many’. Surely one cannot be ‘too many’!
True. When in happens too often - and I admit that this can also be a matter of perception - then problems begin.
I might point out also that AI can lose all the battles in the world and not quit the game. Players, on the other hand, can. This is a grinding game, an high level of ansiety/frustration over every single battle is not good IMHO. Massaging the RNG should be considered accordingly…
Of course it can. But I’ve had a 4 AND 5 resurrection streaks happen against me a few days apart. And i can assure you, I’ve never had a streak of more than 4 or 5 fights where an opposing tds with all traits unlocked didn’t resurrect. He resurrects art least 50% of the time and at least once.
I could say the same with Infernal King, but it seems no one else has the same perception.
@Bradt1977 count yourself lucky, I had TDS resurrect 8 times last week in 1 pvp match, that is a new record for me. Thankfully I still managed to win.
That’s what i mean. Somehow there’s a greater chance for that to happen 8 times in a row than it is to NOT happen 8 fights in a row.
Seriously, start marking down what you experience. This is not what I see, so I want to know just how ridiculous it is for you (and anyone else who cares to take records).
2 Likes
The immortal dragon soul thing is a real problem. I actually have no problem with a troop having a summon on death chance, it’s the respawning itself that’s the issue. How many threads have we seen like this one? It pisses people off me included. The game is meant to be fun, if there’s an issue that pisses people off then FIX IT. I get defeated many times a day in PvP and it doesn’t bother me, but at least once a day a TDS respawns multiple times and it’s infuriating to the point of wanting to throw my controller through the tv and stop playing.
There are many ways to fix it, here are a couple. #1 Only let it respawn once. #2 Have it respawn as a random dragon.
Problem solved.
I won a battle tonight against a Worldbreaker Infernus team because my dragon soul resurrected twice. Sometimes it makes all the difference on an invade.
I think you say something that is guaranteed to be accurate. Unfortunately, it has no relationship towards the actual test I performed. I have run simulation with the same pRNG generator which presumably GoW uses and found that the results are perfectly random and independent. The difference between two samples is very substantial.
Does TDS qualify as a random gem spawner? Same is true for Kraken. These two have frequent anomalies reported by various players. This might be due to their popularity though, so IDK. Anyhow, if there is some link between random spawners and their resurrection/devour ability triggers, this might explain the strange results. Specifically, if these two are linked, that would make the events not independent.
Also, you cannot alter the actual pRNG result, it is only possible to alter the outcome. How is this accomplished specifically? Is the new correct outcome forced or a new pRNG output is generated and checked again in a loop to be correct?
From the point of common sense, all these considerations are perfectly valid. However, statistics and Law of large numbers do not work the same way. The array of individual pRNGs cannot be combined to form some larger pool. Local pRNG results presumably originate from a local seed and local seeds might be different for each individual process. Hence, strictly speaking, combining the pools is not possible without making some specific assumptions. However, if, for example, a random seed number is generated by a single specific server through a continuous process, this might work to an extent. Although normally, pRNG seed numbers are generated locally.
Also, as Lyya kindly explained previously, parts of pRNG are run on the server while parts are local and these certainly cannot be combined. It is not even known if the actual pRNG generators are the same.
Some pRNGs may produce more platykurtic results than others ( Kurtosis - Wikipedia ) and thus, incidence of rare and extremely rare events are considerably higher than one would normally expect. Although Mersenne Twister is not known to be particularly platykurtic, it is supposed to actually be quite uniform and passes majority of easier quality tests.
Another probably invalid comparison is GoW versus lottery. Lottery has a single true random event that determines the outcome while GoW has multiple copies of pRNG events. Thus, any low probability event still remains a low probability event no matter what. Roughly speaking, when I get that one in a million event on my personal copy of GoW, it has no relationship with what other players get on their personal copies nor it might increase the chance of this happening.
That might be coupled to ultrare event filtering as well to exclude some real anomalies. Like TDS resurrecting 8 times in a row (1E-5). However, probably, the issue of borderline might need to be balanced. A random game without surprises (pleasant or unpleasant) might be considered boring. I want these four aces on my hand. 
With regard to what specific events are considered “luck”, the list that we have had on the forum from console developers and Sirrian previously Remove Increased Luck Factor for the AI - #25 by Ozball are not specifically very rare. So, actually this type of filtering might change the outcome more frequently and more substantially than any rare event.
1 Like
-
in the spirit of this thread, I’d like to note how reliably Doomclaw manages to Enrage every ally except the first one… surely another example of why traits/effects like that should at least target an ally that doesn’t already have the effect… and for balance, traits like Mab / Jarl / Venoxia etc should do the same…
-
I definitely misread the title of this thread as ‘I call @Shinnanigens’ and thought ‘hmmm call-outs aren’t allowed’…
I’ve never worked in a business settings. My statistical background is mostly medical. Nature does have some occasional streaks but does not have really rare probability events. A person cannot be 1 inch in waist or half an inch tall or have blood pressure of 1000 mm Hg. Considering that there are 8 billion people on the planet, this should be possible any time soon, we just have to wait a little longer.
Yeah well neither did I 
Oh, how many circus-zoos I’ve worked at that called themselves businesses lol…
1 Like
We should get near to a Black Hole for that.
1 Like
Well if it’s possible for resurrection to hit 8 times in a row at 25% possibility, then there should be streaks of like 40 somewhere where it doesn’t hit. I guarantee that’s never happened.