Let's prove the AI is fair and reasonable

This one’s not as unlikely as you might expect. It’s a weird vagary of event chests that a glory shop troop has 3 times the normal chance of being pulled. So this week, Barghast has 3 times the chance of Pride, as they’re both Epic. So you were unlucky, but not surprisingly so. The odds of pulling no Pride from 250 event keys is around 1 in 500. 18 Barghasts are exactly the average number expected from 250 keys.

2 Likes

I hate to say it, but…um…this premise already shows that you’re misunderstanding the statistics. :fireworks: Irony! :fireworks:

If something “should only happen every few thousand games”, and thousands of people are each playing several games every day, then that “rare” event will happen all the time, overall. “Every few thousand games” is more common than you think - again, overall - when there are “many, many thousands of games” happening.

2 Likes

Actually the probability I “misunderstand” is this rare event is happening 8 or 9 times a day in maybe 20-30 games I tend to play and I’d like a lot of mythic ingots. The point is it is exceedingly common, but every time someone complains it’s dismissed as “a rarity”.

1 Like

I don’t think anything ever gets “proven” by collecting exxagerated reports of events. People claiming that TDS ressurected 7 times in a row as if that happened regularily and all that…
But seeing some posts in here i guess this is the official exxageration thread anyways so i guess that fits well.

1 Like

Tossing a coin 50 times could give the same result 50 times, and yet it doesn’t means the universe is favoring that particular result. It’s an oddity as event sure, but it’s still possible.

A player playing 50 battles during a day could see these events 50 times and it would still be inside the realm of possibility/probability without meaning that the A.I is actively luckier than a player.

Again, since there is a heavy personnal investiment from the players about the only acceptable outcome that validates the egocentric view that he/she/it must win everytime, and any discrepancy to this fantasy caused by the reality tends to birth such threads claiming the A.I cheats.

This is the post I base it on: Short testing of resurrection rates - #7 by akots

I haven’t seen @akots post an analysis of stuff in a while but I always looked forward to the math posts. I myself argued with this analysis at first, but when I went to look up the algorithms in question I found that outside sources verify these tests are proven to determine, from a small sample size, whether an event is independent.

This is important because if events are not independent, normal probability does not apply and things do not behave the way most people understand probability.

I imagine akots quit posting such analysis because “shrug” is about the best response I ever saw to it. Most people looked at it, decided it was wrong, and didn’t even bother to go read about the math and whether people who do statistics for a living trust them.

Anyway, the only “change” I want to push for (because fixing an RNG is a difficult prospect) is I’d still love to see free turns capped or a mode without them. Again: they make me angry far more often than they make me happy.

2 Likes

Yeah i saw his test and i found it to be inconclusive with this small samplesize and him making his assumptions from a sample set with significantly above average resurrection procs aswell as proc streaks if i compare it to my own experience with the troop.
His math may be sound and his conclusion from his sample set right but it did not fit with how the troop behaved for me in over half a year of playing nothing other than TDS teams, in a time when you would also meet TDS very often.

1 Like

I made peace with the RNG.
While my kraken may never devour i sit in comfort knowing that someone in this world has a kraken with a 100% devour rate :laughing:

Been playing daily for over a year. You can make all the claims you want devs but your testing is inadequate. The AI cheats. A lot. Period.

One thing to note is that there are two separate pseudorandom number generators (pRNGs) at play here. The client pRNG is responsible for things that occur within a match — gem creation and fills, AI decision making, respawn/devour/etc. proc chances. The server pRNG is responsible for matchmaking and chest opening.

The most likely implementation of their client-side pRNG algorithm is either System.Random (since they’re a C# app) or UnityEngine.Random. The former is available for perusal here:

The latter is not open-source; it does not forward to System.Random, but rather to Unity proprietary native code, which you can see by inspecting the Unity C# bindings here. I highly doubt IP2 are using any other Random algorithm as they have no need for cryptographically-secure guarantees and those are slower, to boot.

Their server-side pRNG library is anyone’s guess, since we can’t be sure what language and architecture are supporting their web services.

3 Likes

:thinking:
Jokes aside, i would like to ask something Lyya: In cases with streaks and repetitions of events/drops can it be explained as a failure to load/create a new event/drop and the system simply uses the last result calculated or something in this sense?

I have a “small collection” of great drops in the times i used to farm Arcane Traitstones like 3 consecutive Arcanes and 4 consecutive Celestials and 3 consecutive Treasure Gnomes during a normal week with no Vault Event.

(I also found money on the street today, so i bought Divine Protector. Thanks RNG!!! :sunglasses:)

Without seeing how they use the results of a pRNG call, that’s certainly possible. Code like the following could theoretically exist:

void IAmABadMethod()
{
    GemColor color = GemColor.Red;
    while (BoardNeedsMoreGems())
    {
        try
        {
            DoSomethingThatMightThrow();  // Could jump to the finally block in an exceptional case and skip GetRandomColor()
            color = GetRandomColor();
        }
        catch {}
        finally
        {
            AddGemToBoard(color);
        }
    }
}

or perhaps something like this:

void IAmABadMethodAlso()
{
    GemColor color = GemColor.Red;
    while (BoardNeedsMoreGems())
    {
        if (ItsTimeForANewColor()) // Could return false and skip GetRandomColor()
        {
            color = GetRandomColor();
        }
        AddGemToBoard(color);
    }
}

In both cases the calling patterns would allow for color to not be updated in the loop based on exceptional or conditional behavior. There’s really no way to know, short of seeing the source.

2 Likes

See this is the thing, The devs TOLD us it’s fair and reasonable, They TOLD us they ran cpu vs cpu tests, they TOLD us that it’s “Luck Based” Unles they programed in “Luck” for the cpu…Ut’s total Horse S#1t for the simple fact they TOLD us that’s what they did. I still don’t buy it…because they have offered up proof of a lot of other things…like bugs and how to get around them…But not once have they released any footage of these cpu vs cpu matches they told us about…That is fishy in my opinion. And through all of this MANY sheeple has stated “Get over it, it IS luck because the devs told us so!” Besides this is a mute point that i am trying to make anyway…because nothing will change and I am at a level now where Most of this game is EXTREMELY easy…But still enjoyable.

1 Like

If they would provide such footages would you believe them or even consider they were running a different A.I vs A.I just to trick you/us?

One thing is to have suspicions about something that feels sketchy, and other thing is to have such a disposition to an engraved idea in your brain about the devs purpposedly using a cheating code for the A.I. or that they simply doesn’t understand their own game code in a way that it gives the A.I. some sort of advantage…

This is the sort of thing that you and others should examine in yourselves to effectively conclude if anything the devs say/show will or won’t have any impact in your experience.

3 Likes

I sure hope you’re not holding your breath waiting for the source code. As a thought exercise, what would it take to convince you that the code is “fair?”

Edit: sniped!

1 Like

The format around each type of card in the previous versions left them with a more beautiful style and differentiated according to the level of each one. So why do not the developers of the game come back with the old model of the cards instead of just leaving them with two colored lines at their ends?

Exactly this is the issue. The RNG in this game is very streaky. The RNG doesn’t favour the AI over the human player (both are equally (un)lucky), but it is human nature to focus on the negative so we remember the bad RNG streaks more clearly (recall bias). It’s just the streakiness of the RNG that makes players notice the bad RNG rolls more…

3 Likes

Why would I care about said code? I am not a programmer. and I wouldn’t know what to do with it if I had it.
And to convince me? set up a Player vs Computer and a Computer vs Computer and let it run out 50 to 100 matches each…and either livestream it or put it up on youtube after the fact…Visual Proof is MUCH easier to believe than the word of a corporation.

1 Like

Sometimes it’s just different experiences being reported. Opening Event Keys for Sin of Maraj troops mostly everyone on the forums got more copies of Lust than Wrath… and then there is me:

My streakiness seems more inclined about being unlucky at getting Pride… But well, cest la vie

I’ve also got more copies of Wrath than of Lust…

1 Like