When you change your heroes, does it change? When you switch back, what happens?
I’m curious as a player here.
Take your meds. Wear your special hat. Watch out! They may be listening.
3 Likes
It’s random sometimes it does work in my favor and one hit kills everyone because half the board is purple. Not the point I just wish it was more balanced so if I cleared the board the colors im looking for would appear not in abundance but at least appear.
Yes, That. I really had to laugh today because I specifically set up a team to work against a strong holy defense with a two dark combo team. I spotted a total of 6 dark tiles during the entire match. Fortunately I was able to put all 6 tiles together. It as not enough to even charge the fast mana heroes, but it happened to be just enough along with an obscene number of red tiles to charge Natalya over and over to score a win.
Sometimes the randomness goes in your favor and sometimes you lose to a -700 power team. Shake your head and move on.
1 Like
This horse has been beaten, skinned, tanned and made into its own saddle here:
@rook @Coppersky would you please merge and close?
3 Likes
Merged. And you made me laugh so hard at the saddle comment. TRUE! 
2 Likes
I switched to doing all of my raids with a team that consists of 5 green heroes (Melendor 4/44, Zeline 2/50, Peters 4/40, Caedmon 3/60, and Morgan 2/50) and I have been using this team for 2 weeks. This Green Team has an overall team power is 3200. Using this attack team, I can keep my cups in the 2100 range and win most of my battles (looking in my watch tower, I have 30 wins to 13 losses. I have been rerolling for high ham, not easy battles). If the Dev’s did as you say, I would not win. I can assure you that some boards absolutely rock for this team and some boards are horrid. I seem to always remember the bad boards.
Edit: There are almost always more green tiles at the beginning of the battle than the end…
2 Likes
How do you explain this then?!?!!?!
I’m playing, I just love to tease this thread. I actually won that raid, by the way. @Adar check this one out!
7 Likes
hehehehehehe 
I had some of those as well in my stats !
BS. this is not random. It is not solved.
Do not through thus under the rug.
Even in the war when it auto fills my team and I’m missing a color. The first 3-5 moves can only be the color that is missing. Everytime.
Measure it, record the results and report back.
You won’t, of course, because you’re incapable of the effort. Even if you could do it the results would provide no support for your paranoia, so I guess you’d sit on them anyway.
See the checked box next to this thread’s description? That indicates that this ‘problem’ has been solved. Maybe you could read through the thread to see how.
For better or worse, @Brobb, human nature and statistics don’t mesh well. @Danityjo – you are reaching conclusions based on a few observations. Here is why that’s not helpful:
-
Insufficient sample size. Humans are very good at jumping to conclusions from totally insufficient information. Malcolm Gladwell wrote an interesting book about this, called Blink. The problem is, for hypotheses like “tiles are random”, you need a lot of data. If I flip a coin once, and it lands on heads, can I conclude that the coin is biased to heads? No. What if I get three heads in a row? Ten? One hundred? Your war experience is a sample size of six for a much more complex problem. See posts above where a systematic, large dataset was collected to address the problem.
-
Negative reinforcement. Humans don’t like losing, and when we lose, we look for reasons outside of ourselves. That is, we notice “bad luck” more than “good luck.” You probably have forgotten the starting boards you got that were tremendous and gave you a quick win; it’s human nature to do so.
-
Confirmation bias. It is also human nature to notice data that supports our beliefs, and to overlook/ignore/discount data that contradicts our beliefs. If one starts with the idea that tiles are biased against you, then you notice particularly those boards that confirm this conclusion and (unconsciously) overlook those that do not.
To counter our human’s natural instincts, the science of statistics was developed. Fortunately, someone earlier has posted convincing data, systematically acquired, that allows us to reject the hypothesis “tiles are biased towards colors of missing heroes” with a very high level of confidence.
9 Likes
Great post @Kerridoc
Also something I’ve noticed. When you double on a colour… not only do you immediately notice if there is a paucity in the doubled element’s tiles, but whatever matches there ARE you tend to hungrily hoover up (also human nature). The result will be a continued paucity of the doubled element’s tiles on an ongoing basis during the match, which really reinforces the feeling.
On a similar basis: How often have you noticed when you are lacking a hero for a given element in a board how often the board seems to massively fill up with that colour over time? It’s very tempting to then feel the game is ‘out to get us’ by continuing to feed us that colour, but in actuality if’s only filling with that colour because you tend to avoid playing them (due to the fact that they give the enemy mana without doing damage). So OF COURSE they’ll build up.
Slippery devil this human perception… 
3 Likes
Who exactly is this negativity spewed at?
Huh? What negativity?
After playing this game for several months (started in December) and having worked on software that required a robust random number generator system (we were working with extremely small probabilities), I don’t think that this game has a good system to generate randomness (whether it’s a random number generator, lotto ball machine, or monkeys with darts in the background). I am not in the camp that thinks it’s a conspiracy or related to game play, I just think their randomness generation system is not that robust.
As posted above, I have been raiding exclusively with an all green team, so I am extremely observant of green tiles now. I don’t switch my team (probably a bad choice on my part), and I have a months’ worth of history. In the beginning I was writing it down, now it is just my raiding with the following rule of thumb. (1) If I lose 2 matches in a row due to a lack of green gems, (2) I stop raiding and wait an hour or two, (3) I start back up and I seem to get better tiles. Usually this works sometimes it doesn’t. Similar patterns occur with loot, raid matches, hero pulls, etc.
I tallied about 100 runs through world 8-7 and saw similar “patterns” with streaks of lots of Practice Swords and Adventure Kits followed by oddly long dry spells. I understand that 100 flips of a coin are not going to follow a pattern of heads, tails, heads, tails… But the probability of 10 heads in a row is 1 in 1024. The probability of 10 heads in a row followed by 10 tails in a row is 1 in 1,048,548. I got 4x 4* and 6x 3* ascension items so far in April after getting nothing in March. To me this is either a case of 1/1,000,000 odds or a bad random number generator. My money is on the RNG (literally in the case of 10x hero summons).
2 Likes
Nice work here though it does miss a piece—none of your stats take into account your opponents’ colors or their positions. I really don’t know if that has any effect but I don’t think anyone can categorically say it doesn’t, either. I know such statistics would be massively difficult to collect, even more so than the huge amount of effort you already put forth here, but I don’t think you can call your experiment complete without it.
That said, I can’t believe that no one has pointed out the marked drop in the frequency of blue tiles when fielding 2 blue heroes. If that isn’t confirmation of some kind of bias, I don’t know what is.
I can tell you that what brought me here is anecdotal—fielding 3 red heroes (maxed or near maxed 4*) vs a green Titan, in 3 straight battles, I got a total of 8 red matches, only one of which was more than 3 tiles. It was obvious because one of the heroes was Boldtusk, who was also my only healer, and he only fired his special twice across the 3 boards; the other 2, thrice. I felt strongly that this wasn’t simply my perception. And I think Kahree’s excellent work confirms this.
And don’t be too eager to dismiss this as yet more anecdotal evidence; while I would never say “a million people can’t be wrong”, I will say that there has been enough anecdotal evidence presented in this thread and others that SG should review their board and tile generation algorithms.
This streak observation is interesting. If the RNG is entirely server-side, though, there have been thousands of draws for other players in between any two draws you make from the RNG. This fact would mask any streakiness in the RNG output from any one player’s data.
There’s clearly some amount of tile generation that happens in the client. If you lose internet mid-board, it can still keep playing just fine. I’ve always assumed, though, that the client pulled a random seed from the server RNG at the start of each map level to generate the tiles, and (per the logic above) that seed will not be subject to streaks.
One bit of RNG mechanics I stumbled across: I had a TC20 hero waiting to be pulled from the camp at the moment SG took a snapshot of my account to mirror it in beta. When I pulled that hero from the camp in Beta, it was Joon. Excited, I opened my production account and pulled Renfeld. So, heroes in camps are rolled when they are pulled out, not when they are generated.
2 Likes
This is utterly irrational on multiple levels. It’s pattern seeking, post hoc ergo proctor hoc thinking, attributing agency to the phenomenon of regression to the mean, and leaning into confirmation bias. And that’s just in one paragraph!
All these things are typical of the way people think and the reason we jump to flawed conclusions so often. The tragic bit is that you started out so well: “In the beginning I was writing it down…” Yes! Writing it down is the thing to do! If you write down the actual data - not impressions or emotions - then later you can pull it to bits and find out what it really says.
This is only true if you flip the coin 10 times. If you flip it 100 times, the probability of 10 in a row increases radically. (Do I have to calculate it? It’s tricky.)
Similarly, this is only true if you flip you coin 20 times, not if you look for a sequence of 10 heads then 20 tails in a larger sample.
This is sloppy thinking for two reasons:
-
You have not even attempted to analyse your March vs April results. One in a million odds based on what?
-
Let’s say the game used a horrifically bad RNG, worse than any they are really likely to use. For the sake of it let’s say they just appropriated the digits of Pi - totally predictable, known, the opposite of random. This would work absolutely fine for boards, ascension items and hero summoning - anything in the game, really - and would in no way generate the sort of systematic results you claim to be seeing. So if the problem you think you see was real (it isn’t) it’s cause could not be what you seem to think it is.
1 Like
That’s amazing. The data @Kahree posted showed nothing of the sort, yet that’s what you think you see.
Again, amazing. @Kahree’s excellent work shows the opposite of this (more precisely: provides no evidence in support of this) yet you still think your perception is reinforced by it. Well there certainly is a bias being confirmed here. It’s a confirmation bias.
1 Like