How random is the random function when selecting race?

[Oddnerd]

After playing mostly 3v3 and some on-and-off 1v1 as IG, I've decided to start doing some 1v1 as random to improve my micro and to improve my short-term decision making (instead of falling back on a few memorized builds, I'd rather learn to build on the fly). It's been rough initially, but the worst part is that the random button isn't actually random, it is about 50% WSE, 20% OM, 20% tyranids, and 10% everything else.

Does anyone know what hidden mechanics are behind the random function? Does it give preferential selection to races/commanders you don't play? I haven't chosen to play WSE since before chaos rising came out, so I wondered if that was somehow skewing my random selections.

[egewithin]

Here is how the random mechanic actually works sir, by an old meme from me.

[Toilailee]

In my experience it seems to favour the heroes you've played recently and there seems to fairly high chance of getting the hero you played in the previous game.

Soom once documented what he got by random in 100 games and iirc he got farseer 18 times.

[Indrid]

Pibb documented all his random picks too, and it was a pretty even spread of stuff as you'd expect from a random choice. Even though it felt like he got Nids especially HT way more often,

[Cyris]

Small sample space, confirmation bias, humans do not understand stats. There is of course a small chance the random function is biased a bit, but likely for something mundane like rounding errors.

I remember when I was working on an MMO and a player posted a furious rant about how broken the crating system was. A 80% crt rate for making a new sword, but he failed 5 times in a row! Clearly our system is broken!

No, the opposite is true. If there ISN'T a player out there who rolls 20% 5 times in a row, it means we aren't random. There is something like a .03% chance to fail a 80% roll 5 times in a row. This means if 1000 players each craft 5 times in a row (pretty likely) there is a 74% chance at least 1 person will fail all 5.

So when that guy posts on the forums and complains that random isn't random, he's wrong. Personal Anecdote < Stats

Some other fun with stats that a co-worked just told me while checking my math: If you have a 1% chance of an event happening, it will happen on average after ~70 trials (which is to say, there is a 50% chance that after 70 trials, it happened at least once). So lets say that the 1% event is getting nids 10 games in a row. That will happen to approximately one in every 140 players who selects random 10 times in a row.

So again, Personal Anecdote < Stats

[Cyris]

I'm having some fun brushing up on stats now... Let's go deeper.

There are 6 factions, or 7 if we include OM. Relic likely either coded Random to pick from the commander list, or to pick a faction and then a commander inside. I suspect it's the former, since I feel like I don't get OM that often (I play random almost exclusively these days). So let's assume that it's the former, so we have 19 commanders. I'm gonna fudge numbers and just imagine there is a 20th commander to make the numbers round... so there is a 5% chance of each commander being picked by random.

Assuming all that (20 commanders, even 5% chance for all, no outside factors are included) what can we say?

5% of the time you will get the same commander twice in a row.
.25% of the time you will get the same commander three times in a row. This sounds small, but it means that out of say 10,000 sets of 3 Random games played, we'd expect ~25 of them to be the same commander. And each and every one of those players is gonna think "geez, this can't be random". But they are incredibly wrong. In fact, we'd expect a tripple about every 275 sets of three.
.0125% of the time you will get the same commander four times in a row. That's one expected result out of every ~5,500 sets of four matches. Compared to the number of games of dow2 that have been played, I would bet quite a bit on that happening!


All these numbers are assuming 20 commanders too. If we go back to the days before OM, and look only at faction, then we only have 6 races. The chances of getting repeats are significantly higher! Rounding for easy math, we'll change the actuall probability of a given faction (14.28%) to 15%...

After 30 sets of 3 games, you'd expect to get the same faction 3 times in a row.


Ok, maybe I should go do real work that I'm being paid for X.x

[Asmon]

Truth is we don't know how relic made this thing work. Perhaps it first picks a race then a hero, perhaps it depends on what you've previously played... but it certainly feels poorly implemented at times.

[Paranoid Kamikaze]

Selecting random is a way of picking Nids without letting your opponent know. Truth.