Mathmetically speaking

David100 · #1 07-19-2005, 07:40 AM

I have played 1000 215sngs and have a 70% roi. Not sustainable.

Why not play the next 1000 sngs 5+1 to equal out the varience for less losses.

Why would this theory not work, or would it?

David

ps. please do not comment on the 70% roi.

pergesu · #2 07-19-2005, 07:42 AM

[ QUOTE ]
I have played 1000 215sngs and have a 70% roi. Not sustainable.

Why not play the next 1000 sngs 5+1 to equal out the varience for less losses.

Why would this theory not work, or would it?

David

ps. please do not comment on the 70% roi etc.

[/ QUOTE ]

I call shenanigans. Fat chance on not having people comment on the ROI.

Anyway, no, that wouldn't work. There's no reason that the next 1000 are going to run you at like -40% or whatever.

Anyway, with your new 150k, I'd invest a lot of it elsewhere and just keep pwning the 215s. Hopefully you don't start running bad and only pull ~30%

David100 · #3 07-19-2005, 07:45 AM

i better say now to avoid further complication, i dont have that roi, its more of a hypothetical situation.

But to have that roi would be extremely extrememly lucky, hence more chance of having a big crash soon?

David

pergesu · #4 07-19-2005, 07:49 AM

I'm not sure you can say that because you had an absurd amount of good luck, you're very likely to soon have an absurd amount of bad luck.

Based on your ROI and finish distributions, over a period of 1000 games your profit should be within a certain range. Anything beyond the bounds of that range is just high variance.

If you've been running well lately and are worried about regression to the mean...don't trip. Just play and let variance take its course. I think if we could control it or at least channel it, Gigabet would have already posted by now [img]/images/graemlins/smile.gif[/img]

David100 · #5 07-19-2005, 08:05 AM

I agree with you completely, just trying to get my head around why not.

If I flipped a coin 100 times and all were heads, the chance of getting a tails in the next throw is exactly the same as before; however over the next hundred throws why not bet a bit more on tails as surely it should even out?

btw, I am running normally and just thinking from a hypothetical point of view.

I know the rule is that you have exactly the same chance as before, but in the long run surely varience is going to level out - however the speed of this regression is not determined and it could be a gradual one as opposed to a sharp one.

Just some crazy thoughts,

Hope that it makes sense, ish

David

David100 · #6 07-19-2005, 08:08 AM

oh, some of the things i was thinking while writing that have been said [img]/images/graemlins/smile.gif[/img]

I do agree with you all completely. I like the roulette example and was watching people play only the other day doing what you said and feeling very sorry for them.

David

pergesu · #7 07-19-2005, 08:25 AM

[ QUOTE ]
If I flipped a coin 100 times and all were heads, the chance of getting a tails in the next throw is exactly the same as before; however over the next hundred throws why not bet a bit more on tails as surely it should even out?

[/ QUOTE ]
I would bet on heads because I have absolutely no reason to believe that tails is a more likely outcome.

HesseJam · #8 07-19-2005, 08:37 AM

It doesn't really "even out". Ther is no pull or push in either direction. The fluctuation of the relative difference between the times you observed event 1 and the times you observed event 2 is getting smaller. Sorry for not using proper mathematical terms here, I know those only in German.

SonnyJay · #9 07-19-2005, 10:02 AM

As a finance major who recently had to take an extensive statistics program, I'll present you something similar to an example that my professor gave me.

Suppose you're playing a game with your friend where you're flipping a coin, where you make $1 for each heads and you lose $1 for each tails. You will flip 50 times, take a break, then flip 50 more. Your EV for the day, obviously (I hope) is 0.

After 50 flips, you happen to be up $5. Now, given that you're up $5, what is your total expectation for the day at this point? It is $5, because you are already up $5 and flipping 50 more times at an EV of 0.

The "law of averages" that everyone thinks of is bogus, that a big upswing is sure to be followed by a big downswing. The real law to be concerned about is the Law of Large Numbers, meaning that as your sample size grows very large, your return will converge towards the EV.

You have played 1000 SNGs at an ROI that clearly is not sustainable, but that does not mean that a downswing is inevitable. Sure you may not maintain that ROI, but all that the Law of Large Numbers means is that as you approach 10,000 SNGs, 100,000 SNGs, and 1,000,000 SNGs, your ROI will begin to approximate your "true" ROI.

In the long run you'll likely win, just not at the rate you currently are winning.

-SonnyJay

HesseJam · #10 07-19-2005, 07:51 AM

Randomness has no memory. If you got dealt AA ten times in a row the chance that you'll get it an 11th time is the same as if you haven't been dealt AA in the last 1000 hands.
Actually, the chance that you get it an 11th time is way higher because the likelihood that you got them 10 times in a row makes it more likely that the randomizer in the system broke down or something and that you are stuck on the AAs.