300BB downswings are not normal. PERIOD!

[John Ho]

Totally wrong. A 300bb downswing is ridiculous when playing well. It just doesn't happen as often as people are saying. Put it this way...if you put in 4 bets every round and lost you would have to lose well over 20 hands to get to 300bb.

I would wager most people who convince themselves a 300bb downswing is normal are somewhat competent players in games too tough for them to handle comfortably.

[John Ho]

I think I would agree with you if you are talking about the typical flounder filled live games you see at major cardrooms. But online poker is a much tougher game with, imo, less information to work on (even using statistical analysis I think seeing players in person gives you a better read on how they are playing at that time than long term stats). Your game has to be pretty sharp to beat the online games - if you get out of line for even half an hour you are probably going to get hammered pretty hard by tables where half of the players are multitabling and only playing super premium hands.

[DcifrThs]

[ QUOTE ]
[ QUOTE ]

This is just plain wrong! The math dictates that a fraction of winning players will experience such downswings, not that it is to be expected!

[/ QUOTE ]

Show me the math. I'm serious.

I've personally only had 2 250's and no 300's but I've probably only played 400-500k hands.

[/ QUOTE ]

Ive had exactly 1 250bb downswing. it happened when party opened the 30/60 game.

the fact is though if you play enough hands it will happen. period. the fact that it is to be "expected" i definately disagree with.

if you do have a 300bb downswing its certainly cause for concern because the probability of it being pure chance is still very small. that is what rig is saying.

what i believe happens a lot of the time is exactly what is expected and written about by feeney, sklansky, malmuth et al. players play winning poker for a time and run good. they overestimate their ability and start loosening up in a few spots. this brings their win rate down and standard devation up once you start abusing this "edge." this in and of itselt increases the likelihood of a large downswing. what very likely happens is that its a combination of the factors above.

the fact that it seems to happen so often to people on this forum is that we all strive to increase our edge and push small edges.its a delicate balance though.

the bottom line is that while 300bb downswings are rare. they DEFINATELY do occur.some likely caused by more bad play than others.

-Barron

[koolmoe]

[ QUOTE ]
Here is another example that illustrates how likely a 300BB downswing is. Lets say I have a 5% risk of ruin with a 300BB bankroll.

Let also assume that every time I win 100BB, I cash out and round my bankroll down to 300BB.

Each time I do this I start out with a 5% risk of ruin. What happens if I do this 30 times?

The chance of my not going broke each time is .95, so the chance that I don't go broke in all 30 trials is .95 to the 30th power which equals .21.

So there is a 79% chance that I will go broke (and experience a downswing of 300BB or more) during one of the 30 trials. Wow, that is depressing.

[/ QUOTE ]

This is a nice example, but there is a slight problem with it. In each trial, the endpoint is either +100 BB or -300 BB. The ROR = 5% would necessarily include some examples which include an initial swing of +100 BB or more eventually ending at net -300BB. So for the trials as you have set them up, the chances of -300 BB should be slightly less than 5%. A decent estimate of the true percentage might be ROR(300 BB) - ROR(400 BB), though that would probably be too small.

Now, the one thing missing in each trial is the number of hands it takes to reach the end of the trial. So even if we compute that we are 79% (or however) likely to experience a -300BB downswing over 30 trials, we have no idea from this approach how many hands are included (on average slightly less than 10000 for each +100 BB trial).

However, it should be clear that as we play more hands, the likelihood of experiencing a -300 BB downswing are increased to the point of almost certainty, even for a winning player (by construction).

It is a really interesting way to look at the question.

[chuddo]

completely not applicable and i really have nothing of content to add, but being primarily a NL player i enjoy reading these discussions.

nothing better than playing deep, and something goes amiss and boom, instant 300 BB downswing.

run bad a little longer and suddenly in a 1000 BB downswing.
thankfully these funks are just as quick to pop out of.

on subject: the math shows it is possible. long-time winning players and respected posters say it is possible. what is there to argue?

however i also agree with rigoletto in that during such a slump the person probably begins to play less optimally. which is very easy to do when it seems like you just cannot win.

[Duke]

All hell breaks loose at infinity.

~D

[kiddo]

http://www.amazon.com/exec/obidos/tg...s&n=507846

[doppler]

Can you please edit the link?

[ISF]

[ QUOTE ]
My SD/100 at 6 handed and less games is ~17 and for full games it is ~15(~150k at 6-max and 100 at full ring), so I wouldn't expect a substantial difference in the size of the swings given identical winrates.

[/ QUOTE ]

The reason that shorhanded downswings are much larger then would be predicted by our winrates and sd, is that these simple formulas people are using are assume that earnings follow a simple brownian motion where the white noise term is normally distributed and iid. It is not at all in full games and is much less so in shorthanded games as the other players matter more. So our SD is higher in short games, and the our results are much more autocorrelated leading to larger swings then we would have if our edge were constant and results followed a normal distribution.
In essence these simulations only would hold under the assumption that we played the same, and against the same players whom also played the same all of the time. This assumption turns out to be much less valid in SH games making the swings larger relative to our PT SD/100 then people realize.

My guess is that if we multiplied the probability that an x down swing would occur given any winrate and SD by 1.5 or so that it would be a more realistic estimate of how often it is really likely to happen.

[Ulysses]

[ QUOTE ]
earnings follow a simple brownian motion

[/ QUOTE ]
[ QUOTE ]
the white noise term is normally distributed and iid.

[/ QUOTE ]
[ QUOTE ]
results are much more autocorrelated

[/ QUOTE ]

I think this guy wins.