For those who don't think 30+ BI swings are a reality...

BadMongo · #1 12-17-2006, 02:16 AM

...you are very, very wrong. [img]/images/graemlins/wink.gif[/img]

For those of you who haven't read ZBTHorton's great post on variance, check it out here.

Since, the thread was basically hijacked into a discussion about how big a downswing is possible in SSNL, I thought I would make a separate post about this. Some people (myself included) are claiming that 30+ buy-in swings are indeed possible, while others claim this is not possible for a winning player unless massive tilt is the primary factor. So, since we have a whole lot of speculation and anecdotal evidence, but no real data/calculations to back it up, I decided to create a small variance demo in excel to show just how bad things can get.

I pulled up one of Pokey's old posts which contained data on standard deviations from a pool of SSNL players that were surveyed (you can find it here). The lower range of SDs seems to be around 40, while the higher range approaches 70. This allows us to simulate a large sample of hands played for different combinations of SDs and possible winrates. I used a thousand 100 hand samples, so the simulation shows 100K hands each time you run it.

Here are some examples. Note that these were not doctored in any way, they are just random results I selected from the simulation with a few re-starts.

An 8.00 PTBB/100 winner with a SD of 40:

A small ~10 BI swing there, not too bad at all.

An 8.00 PTBB/100 winner with a SD of 50:

Increase the SD slightly, and now we have multiple 10 BI swings and an ugly ~18 BI swing near the top. Overall, not too bad though.

An 8.00 PTBB/100 winner with a SD of 60:

With the SD up to 60, we get a sick 25 BI downsring near the middle, but relatively calm otherwise.

Now let's consider a more marginal winrate...

A 4 PTBB/100 winner with a SD of 40:

Another ~25 BI downswing, along with some gross break-even stretches.

A 4 PTBB/100 winner with a SD of 50:

Ugh, a disgusting 70K break-even stretch.

A 4 PTBB/100 winner with a SD of 60:

The mother of all downswings, the people who experience this have quit poker (or life).

So, a 30+ BI downswing is pretty unlikely for a very good player (8 PTBB/100 or more), but definately not impossible. For more marginal winners (4 PTBB/100 or less) 30+ buy-in swings are not just possible, but likely. Keep in mind that this is only 100K hands and tilt has not been factored in at all.

If anyone is interested, the excel file can be downloaded [link removed]. It's kinda fun to fool around with (in a sick sort of way), just to see how bad things can actually get.

edit: I'm having trouble finding a good place to host this. If anyone can help me out send me a PM.

re-edit: Ok I've got a temporary spot for it. Go to http://pokerfilz.50megs.com/ and click on the variance demo link.

ImsaKidd · #2 12-17-2006, 02:37 AM

very nice work sir. very eye opening.

MyTurn2Raise · #3 12-17-2006, 02:49 AM

[ QUOTE ]
very nice work sir. very eye opening.

[/ QUOTE ]

I suggest this post goes into the sticky/faq

deaders · #4 12-17-2006, 02:50 AM

Nice post!

FoldEqu1ty · #5 12-17-2006, 02:56 AM

Thank you Mongo. The lack of maths in the other post was making my eyes bleed.

ImsaKidd · #6 12-17-2006, 03:04 AM

[ QUOTE ]
Thank you Mongo. The lack of maths in the other post was making my eyes bleed.

[/ QUOTE ]

He didnt show the math but I am willing to assume these are accurate graphical representations.

Please sticky this, so gold. Again OP you did some great work, keep it up.

Isura · #7 12-17-2006, 03:14 AM

Can you run the numbers with bld's stats from the old Party games?

He had a ~ 5PTBB/100 winrate with STD ~90PTBB/100

ajmargarine · #8 12-17-2006, 03:18 AM

Apples and oranges sorta.

Z's was over 22k hands. These are all 10x that.

I don't think anyone is disagreeing that it's possible to lose 30 buy-ins. Most people have bankrolls over that at NL200 and up for a reason. Just not over such a short period of time.

Psydoc · #9 12-17-2006, 03:29 AM

[ QUOTE ]
Apples and oranges sorta.

Z's was over 22k hands. These are all 10x that.

I don't think anyone is disagreeing that it's possible to lose 30 buy-ins. Most people have bankrolls over that at NL200 and up for a reason. Just not over such a short period of time.

[/ QUOTE ]

Doesn't that last graph in the OP show a 60 BI downswing over about 20K hands?

This is a very cool little file...assuming it accurately represents possible trends given the winrate & SD specified...

BadMongo · #10 12-17-2006, 03:29 AM

[ QUOTE ]
Can you run the numbers with bld's stats from the old Party games?

He had a ~ 5PTBB/100 winrate with STD ~90PTBB/100

[/ QUOTE ]

Sure.

Here's 3 interesting graphs:

The last one took several tries to produce, but it's possible nonetheless. As you can see, 100K hands is nothing with those stats.

Anyone who wants to try this out for themselves, or check my numbers, I edited the OP to include a link to the excel demo.