On randomness and streaks

[Lert]

We have all seen threads where people post losing streaks to 'prove' MM isn't random because if it were truly random then 10+ streaks couldn't happen, right?

So I asked a friend of mine who can write code to write a smol program to test exactly that. To flip a coin ten thousand times, report the results and count the streaks for either result.

Link here. Just click 'run'.

You can even change the odds yourself to simulate a worse or a better skilled player, by changing the 'odds' value to whichever winrate you want to simulate. You can tweak the number of matches simulated too - in fact, the code is right there for you to check and modify to test other things.

[ArIskandir]

Ok, I feel lucky. Worst streak I got running your sim was 16. My worst I think so far is 6... Hide your simulation from WG, I don't want that into MM 

[Navalpride33]

18 minutes ago, Lert said:

We have all seen threads where people post losing streaks to 'prove' MM isn't random because if it were truly random then 10+ streaks couldn't happen, right?

So I asked a friend of mine who can write code to write a smol program to test exactly that. To flip a coin ten thousand times, report the results and count the streaks for either result.

Link here. Just click 'run'.

You can even change the odds yourself to simulate a worse or a better skilled player, by changing the 'odds' value to whichever winrate you want to simulate. You can tweak the number of matches simulated too - in fact, the code is right there for you to check and modify to test other things.

The principles of math correspond with your conclusion there our forum black cat...

I find the program highly entertaining .

[Depraved_Miscreant]

16 minutes ago, Lert said:

We have all seen threads where people post losing streaks to 'prove' MM isn't random because if it were truly random then 10+ streaks couldn't happen, right?

So I asked a friend of mine who can write code to write a smol program to test exactly that. To flip a coin ten thousand times, report the results and count the streaks for either result.

Link here. Just click 'run'.

You can even change the odds yourself to simulate a worse or a better skilled player, by changing the 'odds' value to whichever winrate you want to simulate. You can tweak the number of matches simulated too - in fact, the code is right there for you to check and modify to test other things.

I worked out what causes those streaks and it is just the congestion of bad players sitting there waiting for a game. 

[Lert]

Just now, Salvo_Creative said:

I worked out what causes those streaks and it is just the congestion of bad players sitting there waiting for a game. 

I believe there's a psychological effect as well. As a player gets frustrated they tend to play worse, which I believe has a small but measurable influence. Plus, the fact is that the majority of players are below average - even if only slightly - which also skews results.

But that is all just my personal theory, I have no proof either way.

[DocWalker]

Thanks and thanks to your friend.  We should at least recognize (or admit) by now that we humans are pattern seekers.  It's how we survived so long.

And for us humans, the concept of randomness very hard to understand.  'We're making it less random to make it feel more random'  ...  Quote by: (do you know?)

Spoiler

Steve Jobs changed the shuffle feature on the iPod after complaints from users

 

[Depraved_Miscreant]

27 minutes ago, Lert said:

I believe there's a psychological effect as well. As a player gets frustrated they tend to play worse, which I believe has a small but measurable influence. Plus, the fact is that the majority of players are below average - even if only slightly - which also skews results.

But that is all just my personal theory, I have no proof either way.

You're absolutely right, this won't effect everyone but it does have an effect on those whom are prone to it. 

[NutzForBoltz]

If the player base is mostly bad players should randomness have bad players on both sides. I seem to hear that the bad players only happen on their team and not the red team or has fewer bad players.

[eviltane]

3 minutes ago, TwoLate51 said:

If the player base is mostly bad players should randomness have bad players on both sides. I seem to hear that the bad players only happen on their team and not the red team or has fewer bad players.

Hence Lerts example of streaks of in random distribution.  It also applies to player distribution on teams.  

[Lert]

4 minutes ago, TwoLate51 said:

If the player base is mostly bad players should randomness have bad players on both sides. I seem to hear that the bad players only happen on their team and not the red team or has fewer bad players.

But for people on the red team, who are just as human as you are, you are the other team.

[Depraved_Miscreant]

Just now, eviltane said:

Hence Lerts example of streaks of in random distribution.  It also applies to player distribution on teams.  

I managed to semi measure when you will have these bad streaks, it is consistent when there is a large volume of players online, which ,makes sense if the vast majority of the players are actually bad. 

[USMC2145]

6 minutes ago, TwoLate51 said:

If the player base is mostly bad players should randomness have bad players on both sides. I seem to hear that the bad players only happen on their team and not the red team or has fewer bad players.

Bad players make up the most of both teams over the last year it seems. It used to be which team had the best good players now it is which team has the worst bad players who will start the downward steam-roll first. 

[SteelRain_Rifleman]

I only go streaking on Saturdays at night when no one is around.

[ThisIsClassic]

Thank your friend for providing that.  I ran 21 simulations at 50% and 55%.  Results below:

As  you can see:

The spread for win streaks is 10 for 55% and 4 for 50%

The spread for losing streaks was 4 for 55% and 7 for 50%

The AVG win streak for 55% is 2.5 more than 50%

The AVG loss streak for 50% is 1.6 higher than 55%

 

WR = 55			WR = 50
Win	Loss		Win	Loss
14	10		11	17
13	11		11	11
19	10		11	10
13	11		11	10
12	12		14	11
12	11		12	13
11	11		13	13
14	11		12	13
12	12		11	11
20	9		13	13
12	10		14	10
22	10		14	13
13	9		12	11
16	10		14	12
14	12		11	13
13	8		11	11
19	11		15	12
19	9		12	13
13	10		11	12
13	13		11	10
15	8		11	13
22	13	<-MAX->	15	17
11	8	<-MIN->	11	10
14.71	10.38	<-AVG->	12.14	12

[BrushWolf]

1 hour ago, Lert said:

I believe there's a psychological effect as well. As a player gets frustrated they tend to play worse, which I believe has a small but measurable influence. Plus, the fact is that the majority of players are below average - even if only slightly - which also skews results.

But that is all just my personal theory, I have no proof either way.

There is definitely a psychological effect because as you get frustrated you start making unforced errors. If you are becoming frustrated it is best to at least get out of randoms and probably better to take a break.

[Ericson38]

http://shiny.calpoly.sh/Longest_Run/

Seems reasonable. I have had 11 game loosing streaks in Moskva, as well as 10s, 9s, 8s.....

With 630 games and 39% win rate....well there it is. Since I'm not close to a 50% player in terms of average damage, I spend a lot of time in the minus run column, no more then 2 or 3 wins in a sequence in the win column. I can't CIT with that boat, or set fires of a notable consequence.

When you see a T10 cruiser or BB sail behind an island, with them broadside and you angled in Moskva, and they back up to let you have it anyway, its gets you thinking. Petro and Smol and Stalingrad will hunt you down at all costs, if they are given the chance.

 

 

 

[NutzForBoltz]

Is it bad players or less skill? I notice clan wars certain clans always do well, other clans don't. In that level the clans always losing are they bad players or less skill in that level?

[NutzForBoltz]

23 minutes ago, SteelRain_Rifleman said:

I only go streaking on Saturdays at night when no one is around.

Now that is a true streak.

[Navalpride33]

3 minutes ago, TwoLate51 said:

Is it bad players or less skill? I notice clan wars certain clans always do well, other clans don't. In that level the clans always losing are they bad players or less skill in that level?

The population in Clan wars is minuscule compared to Random mod population...

In Clan wars case IMO, low population numbers are to blame. The differences in clans skill or abilities are a lot more noticeable as a result.

[SteelRain_Rifleman]

5 minutes ago, TwoLate51 said:

Now that is a true streak.

Some even curse a blue streak. And often say: "Somebody get that man a kilt before that thing pokes an eye out!"

(The true reason Scotsmen wear kilts) 

[Wombatmetal]

2 hours ago, Lert said:

We have all seen threads where people post losing streaks to 'prove' MM isn't random because if it were truly random then 10+ streaks couldn't happen, right?

So I asked a friend of mine who can write code to write a smol program to test exactly that. To flip a coin ten thousand times, report the results and count the streaks for either result.

Link here. Just click 'run'.

You can even change the odds yourself to simulate a worse or a better skilled player, by changing the 'odds' value to whichever winrate you want to simulate. You can tweak the number of matches simulated too - in fact, the code is right there for you to check and modify to test other things.

There was a mathematician in the 1830s, Simeon Poisson I believe, who worked out the math on randomness. It is counterintuitive, but randomness is streaky. The streaks are called a Poisson Distribution.

I don't believe losing streaks are caused by MM alone, I think once you get a certain number of losses you start to push, and make poor decisions, and it snowballs. But the idea that the existence of a streak shows that something is not random is false on the face of it. Math says otherwise. 

 

[tcbaker777]

ok, but this program doesnt take into account the 11 other people just deciding to throw what should have been a won match

[Lert]

1 hour ago, tcbaker777 said:

ok, but this program doesnt take into account the 11 other people just deciding to throw what should have been a won match

Feel free to write one that does. This is just for people to play around with probabilities and streaks in a pure single value random environment.

[monpetitloup]

@Lert 140 hits, 2 fires.

Chappy fire chance: 8%

so statistically how many fires should i have set?

statistics/math may not lie, yet liars use statistics.

oh by the way there's a fire starting mission on:

coincidence?

[Lert]

3 minutes ago, monpetitloup said:

Chappy fire chance: 8%

And an innate fire chance reduction in each of the ships you were shooting at, plus possible (and seeing as it's high tier, very probable) survivability builds with fire chance reductions.