winning player, but lots of losing streaks. what does it mean?

[hhboy77]

there's been a lot of discussion on this board about what the longest streak players have encountered. with this in mind i was looking over my results from my first 170 100+9 sng's.

in just this small sample of play, i've had two 12-tounament losing streaks, a 9-tournament losing streak (which i just ended) and a 7-tournament losing streak. i haven't done any statistical analysis on it, but to me this seems very streaky.

i've been wondering if there is information to be gleaned for results of this nature. perhaps, it says something about
my style of play that i need to modify or change to further optimize my results.

fyi, my roi is around 20% (ugh) and itm about 37% right now. any insights or people who have similar hot/cold results i'd love to hear what you have to say.

[KJ o]

The one thing I would check is the time and day of week of winning and losing games. I think the average quality of opponents vary quite a bit, even if I can't prove that yet.

I wouldn't be surprised if one player could barely break even at some periods and make 40%+ ROI on other times (fri and sat nights?).

[Solitare]

I've got around 150 $10 SnGs tracked with results very similar to yours -- 38% ITM, 22% ROI. I have 5 losing streaks ranging from 6 to 8 tournies. But the worst stretch was around 20 trournies with only 3 ITM, all 3rd place.

[Vuron00]

I went through one last month. For the previous 100 $50+5's, I had a 51% ITM, with a great ROI. Then, I went on a streak of 22 with only 4 ITM.

I sat down and poured over my hand histories and I noticed that as the streak went further along, I started bluffing more and making more and bigger bets. I think that I had started trying to force the issue by driving people out of hands. It only took 1 or 2 hands per tourney to knock me down because of my aggression.

After I noticed this and corrected the problem, I was ITM the next 7 in a row with 4 wins.

I'd suggest just sitting down and reading through your hand histories closely and see if you notice anything out of the ordinary.

[rachelwxm]

what's your autocorrelation, is it statistically significant?

[parappa]

[ QUOTE ]
what's your autocorrelation, is it statistically significant?

[/ QUOTE ]

Rachel,

Would you mind explaining for a non-expert how to figure out if your autocorrelation is statistically significant? It sounds like a very good thing to look at.

[LinusKS]

20% on $100 is fantastic.

Losing streaks are normal.

If you change your strategy in response to bad luck (assuming it is bad luck), you'll lose money in the long run.

[scott1]

All of these posts sound more like variance than anything else. A 12 tourney non-win streak is balanced by a period when you won more than average (like 3 of 6 or 7). It will balance out over time. Your #s look good. I'm trying to get up to that playing $10+1 tourneys.

Read some of the other streak posts on here. Yours will seem like nothing.

[hhboy77]

thanks for the responses.

i'm definitely not complaining about my results because i like to think of myself as someone who takes the lumps and moves on.

what i'm saying is that with a 40% itm, a streak of 12 becomes a statistical certainty at 460 tournaments, so i find it a little bit strange to have had two within 170 tournaments. i imagine the chances of this happening randomly are pretty small so i was curious to see if there was information to glean from it before i dismiss it as a statistical anomaly.

vuron, i think you make a good point, because the natural tendency is to start bluffing when the cards aren't running in your favor. and though my tendency is to start bullying when i'm going bad, i do consciously try to tighten up when things aren't going my way.

i should add that perhaps my streakiness is due to my belief in the 9x or less blinds, push theory. and it's become not that uncommon for players to berate me for being a maniac, which i find strange because i'm considered very tight by those who play against me the most often.

i too have no idea what autocorrelation is, and though i don't think 20% is anything to scoff at, i do believe my true roi is close to 25% and ideally i'd like to get around 30%. anything higher than that i feel would be difficult to sustain for anyone without esp (i.e. me).

[DonkeyKong]

I think the autocorrelation comment was a joke...

Autocorrelation is a problem in statistical analysis if you are running a regression. Basically, if there is a pattern in the data, such as a tendency for data point 2 to be somehow correlated with data point 1 --- then your regression will be inaccurate and you will have to run a time-series (Autoregressive) model. Most time-series data have autocorrelation. If you are running a standard regression, the relationship that you figure out between the 2 variables will be misstated if there is autocorrelation.

Statistically, the error term in the regression model cannot have an 'expected value'... if it does, your model does not do a good job of describing the relationship and you have to then run an autoregressive model...

the point is; if there are patterns of errors in your play over time, then you aren't playing well.... duh...

[parappa]

[ QUOTE ]
the point is; if there are patterns of errors in your play over time, then you aren't playing well.... duh...

[/ QUOTE ]

This isn't what I think. I think that you observe patterns of bad results in your play and then are wondering whether it's luck or whether there's a leak in your game. My understanding would be that if there was significant autocorrelation in your streaks of bad results, it would/could indicate that you were playing badly.

If you already know that there is a pattern of errors in your play, then of course it's "duh". But the issue is when there are a pattern of bad results in your play over time and you don't know whether they're due to errors or not. My idea about what rachelwxm said was that you could imply it. And since the last thing you want to do during a losing streak is continually try on new strategies, it might be a good way to decide whether to tinker or not.

You clearly understand more about it than I do, so perhaps you're right and it's trivial, but it doesn't immediately seem so to me.

[rachelwxm]

donkeykong,
no offence. In a perfect world, your games should not be positively auto correlated. In real world, it's inevitable. Let's say you are multitabling and the network goes down. Or it's 2am you are not thinking clearly and you are down 3 in a row and you are determined to get it back before going to bed. Things like that are not perfect situations, but it EXISTs in real world.

Now if you lose 4 games in a row before of network problem and you think hmm, how likely is that I lost 4 in a row. Then you are in trouble.

One good way to look at your game and judge if you are play “perfectly” is to calculate the autocorrelation of your games. If you have significantly large number, then that might explain the difference of your losing stream vs. the theoretical value. IMO.

[Eder]

my ROI 20% higher playing early..(Euro time)...guess USA players tougher?

[BigHobo]

[ QUOTE ]
I'd suggest just sitting down and reading through your hand histories closely and see if you notice anything out of the ordinary.

[/ QUOTE ]

I concur with this suggestion. I've noticed some drift in my play after winning streaks that have often contributed to losing streaks. What happens is that during winning streaks I've hit so many flops and draws that I start unknowingly overplaying hands. When the good luck runs out and I start missing the flops and draws it can get really ugly, really fast. Reviewing my play or taking a break from playing can get me back on the right track.

[DonkeyKong]

>>In a perfect world, your games should not be positively auto correlated. In real world, it's inevitable.>>

Yes, this is the nature of time series data... in this case, you are plotting results of sit n go's over time. you should ask yourself by looking at your results; is there a pattern here? is their predictive value based on this data? if the data is in a pattern that looks like data point 2 is related to data point 1, your results are probably serially correlated... Are leaks in your game occuring in stretches and then you figure em out?

My point was not that autocorrelations don't exist, they do; this is a time series and autocorrelations very common in time series data.

By all means, analyze your results and realize that good play might come in streaks and try to think about why this is.... is their seasonality present (very good weekends, mediocre on weekdays)?? if so, you should be comparing weekends to weekends and weekdays to weekdays...

just understand that if you say, my ROI is 20%, you will need to make some statistical adjustments if you want to forecast this into the future due to the correlation in the results (the patterns). that's all.

[hhboy77]

i have nothing to add on the autocorrelation topic, because frankly it's a bit too esoteric for me.

from the stats experts, i'd be curious to know what are the chances of somone having a 40% itm to have 2 separate 12-tournament losing streaks in 170 tournaments.

since on average i will have a 12-tourney loss streak every 460 tournetys, does that mean the chances of having two losing streaks of 12 in 170 tournaments is simply 170/920 (18%)?

also, does this mean that i've been a bit unlucky and my true roi (as i believe) should be a bit higher, or is it simply and indication of my style of play (high risk). does this mean i'm likely overplaying marginal hands/too aggressive on the bubble?

cheers.

[rachelwxm]

18% sounds right to me. /images/graemlins/smile.gif

[rachelwxm]

Here is an interesting link
web page (http://www.itl.nist.gov/div898/handbook/eda/section3/eda35c.htm)
Is is in general discussed in great detail in any regression books.